1 @c Copyright (C) 1988-2016 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 * Attribute Syntax:: Formal syntax for attributes.
64 * Function Prototypes:: Prototype declarations and old-style definitions.
65 * C++ Comments:: C++ comments are recognized.
66 * Dollar Signs:: Dollar sign is allowed in identifiers.
67 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
68 * Alignment:: Inquiring about the alignment of a type or variable.
69 * Inline:: Defining inline functions (as fast as macros).
70 * Volatiles:: What constitutes an access to a volatile object.
71 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
72 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
73 * Incomplete Enums:: @code{enum foo;}, with details to follow.
74 * Function Names:: Printable strings which are the name of the current
76 * Return Address:: Getting the return or frame address of a function.
77 * Vector Extensions:: Using vector instructions through built-in functions.
78 * Offsetof:: Special syntax for implementing @code{offsetof}.
79 * __sync Builtins:: Legacy built-in functions for atomic memory access.
80 * __atomic Builtins:: Atomic built-in functions with memory model.
81 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
82 arithmetic overflow checking.
83 * x86 specific memory model extensions for transactional memory:: x86 memory models.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
87 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
88 * Other Builtins:: Other built-in functions.
89 * Target Builtins:: Built-in functions specific to particular targets.
90 * Target Format Checks:: Format checks specific to particular targets.
91 * Pragmas:: Pragmas accepted by GCC.
92 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
93 * Thread-Local:: Per-thread variables.
94 * Binary constants:: Binary constants using the @samp{0b} prefix.
98 @section Statements and Declarations in Expressions
99 @cindex statements inside expressions
100 @cindex declarations inside expressions
101 @cindex expressions containing statements
102 @cindex macros, statements in expressions
104 @c the above section title wrapped and causes an underfull hbox.. i
105 @c changed it from "within" to "in". --mew 4feb93
106 A compound statement enclosed in parentheses may appear as an expression
107 in GNU C@. This allows you to use loops, switches, and local variables
108 within an expression.
110 Recall that a compound statement is a sequence of statements surrounded
111 by braces; in this construct, parentheses go around the braces. For
115 (@{ int y = foo (); int z;
122 is a valid (though slightly more complex than necessary) expression
123 for the absolute value of @code{foo ()}.
125 The last thing in the compound statement should be an expression
126 followed by a semicolon; the value of this subexpression serves as the
127 value of the entire construct. (If you use some other kind of statement
128 last within the braces, the construct has type @code{void}, and thus
129 effectively no value.)
131 This feature is especially useful in making macro definitions ``safe'' (so
132 that they evaluate each operand exactly once). For example, the
133 ``maximum'' function is commonly defined as a macro in standard C as
137 #define max(a,b) ((a) > (b) ? (a) : (b))
141 @cindex side effects, macro argument
142 But this definition computes either @var{a} or @var{b} twice, with bad
143 results if the operand has side effects. In GNU C, if you know the
144 type of the operands (here taken as @code{int}), you can define
145 the macro safely as follows:
148 #define maxint(a,b) \
149 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
152 Embedded statements are not allowed in constant expressions, such as
153 the value of an enumeration constant, the width of a bit-field, or
154 the initial value of a static variable.
156 If you don't know the type of the operand, you can still do this, but you
157 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
159 In G++, the result value of a statement expression undergoes array and
160 function pointer decay, and is returned by value to the enclosing
161 expression. For instance, if @code{A} is a class, then
170 constructs a temporary @code{A} object to hold the result of the
171 statement expression, and that is used to invoke @code{Foo}.
172 Therefore the @code{this} pointer observed by @code{Foo} is not the
175 In a statement expression, any temporaries created within a statement
176 are destroyed at that statement's end. This makes statement
177 expressions inside macros slightly different from function calls. In
178 the latter case temporaries introduced during argument evaluation are
179 destroyed at the end of the statement that includes the function
180 call. In the statement expression case they are destroyed during
181 the statement expression. For instance,
184 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
185 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
195 has different places where temporaries are destroyed. For the
196 @code{macro} case, the temporary @code{X} is destroyed just after
197 the initialization of @code{b}. In the @code{function} case that
198 temporary is destroyed when the function returns.
200 These considerations mean that it is probably a bad idea to use
201 statement expressions of this form in header files that are designed to
202 work with C++. (Note that some versions of the GNU C Library contained
203 header files using statement expressions that lead to precisely this
206 Jumping into a statement expression with @code{goto} or using a
207 @code{switch} statement outside the statement expression with a
208 @code{case} or @code{default} label inside the statement expression is
209 not permitted. Jumping into a statement expression with a computed
210 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
211 Jumping out of a statement expression is permitted, but if the
212 statement expression is part of a larger expression then it is
213 unspecified which other subexpressions of that expression have been
214 evaluated except where the language definition requires certain
215 subexpressions to be evaluated before or after the statement
216 expression. In any case, as with a function call, the evaluation of a
217 statement expression is not interleaved with the evaluation of other
218 parts of the containing expression. For example,
221 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
225 calls @code{foo} and @code{bar1} and does not call @code{baz} but
226 may or may not call @code{bar2}. If @code{bar2} is called, it is
227 called after @code{foo} and before @code{bar1}.
230 @section Locally Declared Labels
232 @cindex macros, local labels
234 GCC allows you to declare @dfn{local labels} in any nested block
235 scope. A local label is just like an ordinary label, but you can
236 only reference it (with a @code{goto} statement, or by taking its
237 address) within the block in which it is declared.
239 A local label declaration looks like this:
242 __label__ @var{label};
249 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
252 Local label declarations must come at the beginning of the block,
253 before any ordinary declarations or statements.
255 The label declaration defines the label @emph{name}, but does not define
256 the label itself. You must do this in the usual way, with
257 @code{@var{label}:}, within the statements of the statement expression.
259 The local label feature is useful for complex macros. If a macro
260 contains nested loops, a @code{goto} can be useful for breaking out of
261 them. However, an ordinary label whose scope is the whole function
262 cannot be used: if the macro can be expanded several times in one
263 function, the label is multiply defined in that function. A
264 local label avoids this problem. For example:
267 #define SEARCH(value, array, target) \
270 typeof (target) _SEARCH_target = (target); \
271 typeof (*(array)) *_SEARCH_array = (array); \
274 for (i = 0; i < max; i++) \
275 for (j = 0; j < max; j++) \
276 if (_SEARCH_array[i][j] == _SEARCH_target) \
277 @{ (value) = i; goto found; @} \
283 This could also be written using a statement expression:
286 #define SEARCH(array, target) \
289 typeof (target) _SEARCH_target = (target); \
290 typeof (*(array)) *_SEARCH_array = (array); \
293 for (i = 0; i < max; i++) \
294 for (j = 0; j < max; j++) \
295 if (_SEARCH_array[i][j] == _SEARCH_target) \
296 @{ value = i; goto found; @} \
303 Local label declarations also make the labels they declare visible to
304 nested functions, if there are any. @xref{Nested Functions}, for details.
306 @node Labels as Values
307 @section Labels as Values
308 @cindex labels as values
309 @cindex computed gotos
310 @cindex goto with computed label
311 @cindex address of a label
313 You can get the address of a label defined in the current function
314 (or a containing function) with the unary operator @samp{&&}. The
315 value has type @code{void *}. This value is a constant and can be used
316 wherever a constant of that type is valid. For example:
324 To use these values, you need to be able to jump to one. This is done
325 with the computed goto statement@footnote{The analogous feature in
326 Fortran is called an assigned goto, but that name seems inappropriate in
327 C, where one can do more than simply store label addresses in label
328 variables.}, @code{goto *@var{exp};}. For example,
335 Any expression of type @code{void *} is allowed.
337 One way of using these constants is in initializing a static array that
338 serves as a jump table:
341 static void *array[] = @{ &&foo, &&bar, &&hack @};
345 Then you can select a label with indexing, like this:
352 Note that this does not check whether the subscript is in bounds---array
353 indexing in C never does that.
355 Such an array of label values serves a purpose much like that of the
356 @code{switch} statement. The @code{switch} statement is cleaner, so
357 use that rather than an array unless the problem does not fit a
358 @code{switch} statement very well.
360 Another use of label values is in an interpreter for threaded code.
361 The labels within the interpreter function can be stored in the
362 threaded code for super-fast dispatching.
364 You may not use this mechanism to jump to code in a different function.
365 If you do that, totally unpredictable things happen. The best way to
366 avoid this is to store the label address only in automatic variables and
367 never pass it as an argument.
369 An alternate way to write the above example is
372 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
374 goto *(&&foo + array[i]);
378 This is more friendly to code living in shared libraries, as it reduces
379 the number of dynamic relocations that are needed, and by consequence,
380 allows the data to be read-only.
381 This alternative with label differences is not supported for the AVR target,
382 please use the first approach for AVR programs.
384 The @code{&&foo} expressions for the same label might have different
385 values if the containing function is inlined or cloned. If a program
386 relies on them being always the same,
387 @code{__attribute__((__noinline__,__noclone__))} should be used to
388 prevent inlining and cloning. If @code{&&foo} is used in a static
389 variable initializer, inlining and cloning is forbidden.
391 @node Nested Functions
392 @section Nested Functions
393 @cindex nested functions
394 @cindex downward funargs
397 A @dfn{nested function} is a function defined inside another function.
398 Nested functions are supported as an extension in GNU C, but are not
399 supported by GNU C++.
401 The nested function's name is local to the block where it is defined.
402 For example, here we define a nested function named @code{square}, and
407 foo (double a, double b)
409 double square (double z) @{ return z * z; @}
411 return square (a) + square (b);
416 The nested function can access all the variables of the containing
417 function that are visible at the point of its definition. This is
418 called @dfn{lexical scoping}. For example, here we show a nested
419 function which uses an inherited variable named @code{offset}:
423 bar (int *array, int offset, int size)
425 int access (int *array, int index)
426 @{ return array[index + offset]; @}
429 for (i = 0; i < size; i++)
430 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
435 Nested function definitions are permitted within functions in the places
436 where variable definitions are allowed; that is, in any block, mixed
437 with the other declarations and statements in the block.
439 It is possible to call the nested function from outside the scope of its
440 name by storing its address or passing the address to another function:
443 hack (int *array, int size)
445 void store (int index, int value)
446 @{ array[index] = value; @}
448 intermediate (store, size);
452 Here, the function @code{intermediate} receives the address of
453 @code{store} as an argument. If @code{intermediate} calls @code{store},
454 the arguments given to @code{store} are used to store into @code{array}.
455 But this technique works only so long as the containing function
456 (@code{hack}, in this example) does not exit.
458 If you try to call the nested function through its address after the
459 containing function exits, all hell breaks loose. If you try
460 to call it after a containing scope level exits, and if it refers
461 to some of the variables that are no longer in scope, you may be lucky,
462 but it's not wise to take the risk. If, however, the nested function
463 does not refer to anything that has gone out of scope, you should be
466 GCC implements taking the address of a nested function using a technique
467 called @dfn{trampolines}. This technique was described in
468 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
469 C++ Conference Proceedings, October 17-21, 1988).
471 A nested function can jump to a label inherited from a containing
472 function, provided the label is explicitly declared in the containing
473 function (@pxref{Local Labels}). Such a jump returns instantly to the
474 containing function, exiting the nested function that did the
475 @code{goto} and any intermediate functions as well. Here is an example:
479 bar (int *array, int offset, int size)
482 int access (int *array, int index)
486 return array[index + offset];
490 for (i = 0; i < size; i++)
491 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
495 /* @r{Control comes here from @code{access}
496 if it detects an error.} */
503 A nested function always has no linkage. Declaring one with
504 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
505 before its definition, use @code{auto} (which is otherwise meaningless
506 for function declarations).
509 bar (int *array, int offset, int size)
512 auto int access (int *, int);
514 int access (int *array, int index)
518 return array[index + offset];
524 @node Constructing Calls
525 @section Constructing Function Calls
526 @cindex constructing calls
527 @cindex forwarding calls
529 Using the built-in functions described below, you can record
530 the arguments a function received, and call another function
531 with the same arguments, without knowing the number or types
534 You can also record the return value of that function call,
535 and later return that value, without knowing what data type
536 the function tried to return (as long as your caller expects
539 However, these built-in functions may interact badly with some
540 sophisticated features or other extensions of the language. It
541 is, therefore, not recommended to use them outside very simple
542 functions acting as mere forwarders for their arguments.
544 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
545 This built-in function returns a pointer to data
546 describing how to perform a call with the same arguments as are passed
547 to the current function.
549 The function saves the arg pointer register, structure value address,
550 and all registers that might be used to pass arguments to a function
551 into a block of memory allocated on the stack. Then it returns the
552 address of that block.
555 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
556 This built-in function invokes @var{function}
557 with a copy of the parameters described by @var{arguments}
560 The value of @var{arguments} should be the value returned by
561 @code{__builtin_apply_args}. The argument @var{size} specifies the size
562 of the stack argument data, in bytes.
564 This function returns a pointer to data describing
565 how to return whatever value is returned by @var{function}. The data
566 is saved in a block of memory allocated on the stack.
568 It is not always simple to compute the proper value for @var{size}. The
569 value is used by @code{__builtin_apply} to compute the amount of data
570 that should be pushed on the stack and copied from the incoming argument
574 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
575 This built-in function returns the value described by @var{result} from
576 the containing function. You should specify, for @var{result}, a value
577 returned by @code{__builtin_apply}.
580 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
581 This built-in function represents all anonymous arguments of an inline
582 function. It can be used only in inline functions that are always
583 inlined, never compiled as a separate function, such as those using
584 @code{__attribute__ ((__always_inline__))} or
585 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
586 It must be only passed as last argument to some other function
587 with variable arguments. This is useful for writing small wrapper
588 inlines for variable argument functions, when using preprocessor
589 macros is undesirable. For example:
591 extern int myprintf (FILE *f, const char *format, ...);
592 extern inline __attribute__ ((__gnu_inline__)) int
593 myprintf (FILE *f, const char *format, ...)
595 int r = fprintf (f, "myprintf: ");
598 int s = fprintf (f, format, __builtin_va_arg_pack ());
606 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
607 This built-in function returns the number of anonymous arguments of
608 an inline function. It can be used only in inline functions that
609 are always inlined, never compiled as a separate function, such
610 as those using @code{__attribute__ ((__always_inline__))} or
611 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
612 For example following does link- or run-time checking of open
613 arguments for optimized code:
616 extern inline __attribute__((__gnu_inline__)) int
617 myopen (const char *path, int oflag, ...)
619 if (__builtin_va_arg_pack_len () > 1)
620 warn_open_too_many_arguments ();
622 if (__builtin_constant_p (oflag))
624 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
626 warn_open_missing_mode ();
627 return __open_2 (path, oflag);
629 return open (path, oflag, __builtin_va_arg_pack ());
632 if (__builtin_va_arg_pack_len () < 1)
633 return __open_2 (path, oflag);
635 return open (path, oflag, __builtin_va_arg_pack ());
642 @section Referring to a Type with @code{typeof}
645 @cindex macros, types of arguments
647 Another way to refer to the type of an expression is with @code{typeof}.
648 The syntax of using of this keyword looks like @code{sizeof}, but the
649 construct acts semantically like a type name defined with @code{typedef}.
651 There are two ways of writing the argument to @code{typeof}: with an
652 expression or with a type. Here is an example with an expression:
659 This assumes that @code{x} is an array of pointers to functions;
660 the type described is that of the values of the functions.
662 Here is an example with a typename as the argument:
669 Here the type described is that of pointers to @code{int}.
671 If you are writing a header file that must work when included in ISO C
672 programs, write @code{__typeof__} instead of @code{typeof}.
673 @xref{Alternate Keywords}.
675 A @code{typeof} construct can be used anywhere a typedef name can be
676 used. For example, you can use it in a declaration, in a cast, or inside
677 of @code{sizeof} or @code{typeof}.
679 The operand of @code{typeof} is evaluated for its side effects if and
680 only if it is an expression of variably modified type or the name of
683 @code{typeof} is often useful in conjunction with
684 statement expressions (@pxref{Statement Exprs}).
685 Here is how the two together can
686 be used to define a safe ``maximum'' macro which operates on any
687 arithmetic type and evaluates each of its arguments exactly once:
691 (@{ typeof (a) _a = (a); \
692 typeof (b) _b = (b); \
693 _a > _b ? _a : _b; @})
696 @cindex underscores in variables in macros
697 @cindex @samp{_} in variables in macros
698 @cindex local variables in macros
699 @cindex variables, local, in macros
700 @cindex macros, local variables in
702 The reason for using names that start with underscores for the local
703 variables is to avoid conflicts with variable names that occur within the
704 expressions that are substituted for @code{a} and @code{b}. Eventually we
705 hope to design a new form of declaration syntax that allows you to declare
706 variables whose scopes start only after their initializers; this will be a
707 more reliable way to prevent such conflicts.
710 Some more examples of the use of @code{typeof}:
714 This declares @code{y} with the type of what @code{x} points to.
721 This declares @code{y} as an array of such values.
728 This declares @code{y} as an array of pointers to characters:
731 typeof (typeof (char *)[4]) y;
735 It is equivalent to the following traditional C declaration:
741 To see the meaning of the declaration using @code{typeof}, and why it
742 might be a useful way to write, rewrite it with these macros:
745 #define pointer(T) typeof(T *)
746 #define array(T, N) typeof(T [N])
750 Now the declaration can be rewritten this way:
753 array (pointer (char), 4) y;
757 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
758 pointers to @code{char}.
761 In GNU C, but not GNU C++, you may also declare the type of a variable
762 as @code{__auto_type}. In that case, the declaration must declare
763 only one variable, whose declarator must just be an identifier, the
764 declaration must be initialized, and the type of the variable is
765 determined by the initializer; the name of the variable is not in
766 scope until after the initializer. (In C++, you should use C++11
767 @code{auto} for this purpose.) Using @code{__auto_type}, the
768 ``maximum'' macro above could be written as:
772 (@{ __auto_type _a = (a); \
773 __auto_type _b = (b); \
774 _a > _b ? _a : _b; @})
777 Using @code{__auto_type} instead of @code{typeof} has two advantages:
780 @item Each argument to the macro appears only once in the expansion of
781 the macro. This prevents the size of the macro expansion growing
782 exponentially when calls to such macros are nested inside arguments of
785 @item If the argument to the macro has variably modified type, it is
786 evaluated only once when using @code{__auto_type}, but twice if
787 @code{typeof} is used.
791 @section Conditionals with Omitted Operands
792 @cindex conditional expressions, extensions
793 @cindex omitted middle-operands
794 @cindex middle-operands, omitted
795 @cindex extensions, @code{?:}
796 @cindex @code{?:} extensions
798 The middle operand in a conditional expression may be omitted. Then
799 if the first operand is nonzero, its value is the value of the conditional
802 Therefore, the expression
809 has the value of @code{x} if that is nonzero; otherwise, the value of
812 This example is perfectly equivalent to
818 @cindex side effect in @code{?:}
819 @cindex @code{?:} side effect
821 In this simple case, the ability to omit the middle operand is not
822 especially useful. When it becomes useful is when the first operand does,
823 or may (if it is a macro argument), contain a side effect. Then repeating
824 the operand in the middle would perform the side effect twice. Omitting
825 the middle operand uses the value already computed without the undesirable
826 effects of recomputing it.
829 @section 128-bit Integers
830 @cindex @code{__int128} data types
832 As an extension the integer scalar type @code{__int128} is supported for
833 targets which have an integer mode wide enough to hold 128 bits.
834 Simply write @code{__int128} for a signed 128-bit integer, or
835 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
836 support in GCC for expressing an integer constant of type @code{__int128}
837 for targets with @code{long long} integer less than 128 bits wide.
840 @section Double-Word Integers
841 @cindex @code{long long} data types
842 @cindex double-word arithmetic
843 @cindex multiprecision arithmetic
844 @cindex @code{LL} integer suffix
845 @cindex @code{ULL} integer suffix
847 ISO C99 supports data types for integers that are at least 64 bits wide,
848 and as an extension GCC supports them in C90 mode and in C++.
849 Simply write @code{long long int} for a signed integer, or
850 @code{unsigned long long int} for an unsigned integer. To make an
851 integer constant of type @code{long long int}, add the suffix @samp{LL}
852 to the integer. To make an integer constant of type @code{unsigned long
853 long int}, add the suffix @samp{ULL} to the integer.
855 You can use these types in arithmetic like any other integer types.
856 Addition, subtraction, and bitwise boolean operations on these types
857 are open-coded on all types of machines. Multiplication is open-coded
858 if the machine supports a fullword-to-doubleword widening multiply
859 instruction. Division and shifts are open-coded only on machines that
860 provide special support. The operations that are not open-coded use
861 special library routines that come with GCC@.
863 There may be pitfalls when you use @code{long long} types for function
864 arguments without function prototypes. If a function
865 expects type @code{int} for its argument, and you pass a value of type
866 @code{long long int}, confusion results because the caller and the
867 subroutine disagree about the number of bytes for the argument.
868 Likewise, if the function expects @code{long long int} and you pass
869 @code{int}. The best way to avoid such problems is to use prototypes.
872 @section Complex Numbers
873 @cindex complex numbers
874 @cindex @code{_Complex} keyword
875 @cindex @code{__complex__} keyword
877 ISO C99 supports complex floating data types, and as an extension GCC
878 supports them in C90 mode and in C++. GCC also supports complex integer data
879 types which are not part of ISO C99. You can declare complex types
880 using the keyword @code{_Complex}. As an extension, the older GNU
881 keyword @code{__complex__} is also supported.
883 For example, @samp{_Complex double x;} declares @code{x} as a
884 variable whose real part and imaginary part are both of type
885 @code{double}. @samp{_Complex short int y;} declares @code{y} to
886 have real and imaginary parts of type @code{short int}; this is not
887 likely to be useful, but it shows that the set of complex types is
890 To write a constant with a complex data type, use the suffix @samp{i} or
891 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
892 has type @code{_Complex float} and @code{3i} has type
893 @code{_Complex int}. Such a constant always has a pure imaginary
894 value, but you can form any complex value you like by adding one to a
895 real constant. This is a GNU extension; if you have an ISO C99
896 conforming C library (such as the GNU C Library), and want to construct complex
897 constants of floating type, you should include @code{<complex.h>} and
898 use the macros @code{I} or @code{_Complex_I} instead.
900 @cindex @code{__real__} keyword
901 @cindex @code{__imag__} keyword
902 To extract the real part of a complex-valued expression @var{exp}, write
903 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
904 extract the imaginary part. This is a GNU extension; for values of
905 floating type, you should use the ISO C99 functions @code{crealf},
906 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
907 @code{cimagl}, declared in @code{<complex.h>} and also provided as
908 built-in functions by GCC@.
910 @cindex complex conjugation
911 The operator @samp{~} performs complex conjugation when used on a value
912 with a complex type. This is a GNU extension; for values of
913 floating type, you should use the ISO C99 functions @code{conjf},
914 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
915 provided as built-in functions by GCC@.
917 GCC can allocate complex automatic variables in a noncontiguous
918 fashion; it's even possible for the real part to be in a register while
919 the imaginary part is on the stack (or vice versa). Only the DWARF
920 debug info format can represent this, so use of DWARF is recommended.
921 If you are using the stabs debug info format, GCC describes a noncontiguous
922 complex variable as if it were two separate variables of noncomplex type.
923 If the variable's actual name is @code{foo}, the two fictitious
924 variables are named @code{foo$real} and @code{foo$imag}. You can
925 examine and set these two fictitious variables with your debugger.
928 @section Additional Floating Types
929 @cindex additional floating types
930 @cindex @code{__float80} data type
931 @cindex @code{__float128} data type
932 @cindex @code{__ibm128} data type
933 @cindex @code{w} floating point suffix
934 @cindex @code{q} floating point suffix
935 @cindex @code{W} floating point suffix
936 @cindex @code{Q} floating point suffix
938 As an extension, GNU C supports additional floating
939 types, @code{__float80} and @code{__float128} to support 80-bit
940 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
941 Support for additional types includes the arithmetic operators:
942 add, subtract, multiply, divide; unary arithmetic operators;
943 relational operators; equality operators; and conversions to and from
944 integer and other floating types. Use a suffix @samp{w} or @samp{W}
945 in a literal constant of type @code{__float80} or type
946 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
948 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
949 types using the corresponding internal complex type, @code{XCmode} for
950 @code{__float80} type and @code{TCmode} for @code{__float128} type:
953 typedef _Complex float __attribute__((mode(TC))) _Complex128;
954 typedef _Complex float __attribute__((mode(XC))) _Complex80;
957 On PowerPC 64-bit Linux systems there are currently problems in using
958 the complex @code{__float128} type. When these problems are fixed,
962 typedef _Complex float __attribute__((mode(KC))) _Complex128;
965 Not all targets support additional floating-point types.
966 @code{__float80} and @code{__float128} types are supported on x86 and
967 IA-64 targets. The @code{__float128} type is supported on hppa HP-UX.
968 The @code{__float128} type is supported on PowerPC 64-bit Linux
969 systems by default if the vector scalar instruction set (VSX) is
972 On the PowerPC, @code{__ibm128} provides access to the IBM extended
973 double format, and it is intended to be used by the library functions
974 that handle conversions if/when long double is changed to be IEEE
975 128-bit floating point.
978 @section Half-Precision Floating Point
979 @cindex half-precision floating point
980 @cindex @code{__fp16} data type
982 On ARM targets, GCC supports half-precision (16-bit) floating point via
983 the @code{__fp16} type. You must enable this type explicitly
984 with the @option{-mfp16-format} command-line option in order to use it.
986 ARM supports two incompatible representations for half-precision
987 floating-point values. You must choose one of the representations and
988 use it consistently in your program.
990 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
991 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
992 There are 11 bits of significand precision, approximately 3
995 Specifying @option{-mfp16-format=alternative} selects the ARM
996 alternative format. This representation is similar to the IEEE
997 format, but does not support infinities or NaNs. Instead, the range
998 of exponents is extended, so that this format can represent normalized
999 values in the range of @math{2^{-14}} to 131008.
1001 The @code{__fp16} type is a storage format only. For purposes
1002 of arithmetic and other operations, @code{__fp16} values in C or C++
1003 expressions are automatically promoted to @code{float}. In addition,
1004 you cannot declare a function with a return value or parameters
1005 of type @code{__fp16}.
1007 Note that conversions from @code{double} to @code{__fp16}
1008 involve an intermediate conversion to @code{float}. Because
1009 of rounding, this can sometimes produce a different result than a
1012 ARM provides hardware support for conversions between
1013 @code{__fp16} and @code{float} values
1014 as an extension to VFP and NEON (Advanced SIMD). GCC generates
1015 code using these hardware instructions if you compile with
1016 options to select an FPU that provides them;
1017 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1018 in addition to the @option{-mfp16-format} option to select
1019 a half-precision format.
1021 Language-level support for the @code{__fp16} data type is
1022 independent of whether GCC generates code using hardware floating-point
1023 instructions. In cases where hardware support is not specified, GCC
1024 implements conversions between @code{__fp16} and @code{float} values
1028 @section Decimal Floating Types
1029 @cindex decimal floating types
1030 @cindex @code{_Decimal32} data type
1031 @cindex @code{_Decimal64} data type
1032 @cindex @code{_Decimal128} data type
1033 @cindex @code{df} integer suffix
1034 @cindex @code{dd} integer suffix
1035 @cindex @code{dl} integer suffix
1036 @cindex @code{DF} integer suffix
1037 @cindex @code{DD} integer suffix
1038 @cindex @code{DL} integer suffix
1040 As an extension, GNU C supports decimal floating types as
1041 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1042 floating types in GCC will evolve as the draft technical report changes.
1043 Calling conventions for any target might also change. Not all targets
1044 support decimal floating types.
1046 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1047 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1048 @code{float}, @code{double}, and @code{long double} whose radix is not
1049 specified by the C standard but is usually two.
1051 Support for decimal floating types includes the arithmetic operators
1052 add, subtract, multiply, divide; unary arithmetic operators;
1053 relational operators; equality operators; and conversions to and from
1054 integer and other floating types. Use a suffix @samp{df} or
1055 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1056 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1059 GCC support of decimal float as specified by the draft technical report
1064 When the value of a decimal floating type cannot be represented in the
1065 integer type to which it is being converted, the result is undefined
1066 rather than the result value specified by the draft technical report.
1069 GCC does not provide the C library functionality associated with
1070 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1071 @file{wchar.h}, which must come from a separate C library implementation.
1072 Because of this the GNU C compiler does not define macro
1073 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1074 the technical report.
1077 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1078 are supported by the DWARF debug information format.
1084 ISO C99 supports floating-point numbers written not only in the usual
1085 decimal notation, such as @code{1.55e1}, but also numbers such as
1086 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1087 supports this in C90 mode (except in some cases when strictly
1088 conforming) and in C++. In that format the
1089 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1090 mandatory. The exponent is a decimal number that indicates the power of
1091 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1098 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1099 is the same as @code{1.55e1}.
1101 Unlike for floating-point numbers in the decimal notation the exponent
1102 is always required in the hexadecimal notation. Otherwise the compiler
1103 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1104 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1105 extension for floating-point constants of type @code{float}.
1108 @section Fixed-Point Types
1109 @cindex fixed-point types
1110 @cindex @code{_Fract} data type
1111 @cindex @code{_Accum} data type
1112 @cindex @code{_Sat} data type
1113 @cindex @code{hr} fixed-suffix
1114 @cindex @code{r} fixed-suffix
1115 @cindex @code{lr} fixed-suffix
1116 @cindex @code{llr} fixed-suffix
1117 @cindex @code{uhr} fixed-suffix
1118 @cindex @code{ur} fixed-suffix
1119 @cindex @code{ulr} fixed-suffix
1120 @cindex @code{ullr} fixed-suffix
1121 @cindex @code{hk} fixed-suffix
1122 @cindex @code{k} fixed-suffix
1123 @cindex @code{lk} fixed-suffix
1124 @cindex @code{llk} fixed-suffix
1125 @cindex @code{uhk} fixed-suffix
1126 @cindex @code{uk} fixed-suffix
1127 @cindex @code{ulk} fixed-suffix
1128 @cindex @code{ullk} fixed-suffix
1129 @cindex @code{HR} fixed-suffix
1130 @cindex @code{R} fixed-suffix
1131 @cindex @code{LR} fixed-suffix
1132 @cindex @code{LLR} fixed-suffix
1133 @cindex @code{UHR} fixed-suffix
1134 @cindex @code{UR} fixed-suffix
1135 @cindex @code{ULR} fixed-suffix
1136 @cindex @code{ULLR} fixed-suffix
1137 @cindex @code{HK} fixed-suffix
1138 @cindex @code{K} fixed-suffix
1139 @cindex @code{LK} fixed-suffix
1140 @cindex @code{LLK} fixed-suffix
1141 @cindex @code{UHK} fixed-suffix
1142 @cindex @code{UK} fixed-suffix
1143 @cindex @code{ULK} fixed-suffix
1144 @cindex @code{ULLK} fixed-suffix
1146 As an extension, GNU C supports fixed-point types as
1147 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1148 types in GCC will evolve as the draft technical report changes.
1149 Calling conventions for any target might also change. Not all targets
1150 support fixed-point types.
1152 The fixed-point types are
1153 @code{short _Fract},
1156 @code{long long _Fract},
1157 @code{unsigned short _Fract},
1158 @code{unsigned _Fract},
1159 @code{unsigned long _Fract},
1160 @code{unsigned long long _Fract},
1161 @code{_Sat short _Fract},
1163 @code{_Sat long _Fract},
1164 @code{_Sat long long _Fract},
1165 @code{_Sat unsigned short _Fract},
1166 @code{_Sat unsigned _Fract},
1167 @code{_Sat unsigned long _Fract},
1168 @code{_Sat unsigned long long _Fract},
1169 @code{short _Accum},
1172 @code{long long _Accum},
1173 @code{unsigned short _Accum},
1174 @code{unsigned _Accum},
1175 @code{unsigned long _Accum},
1176 @code{unsigned long long _Accum},
1177 @code{_Sat short _Accum},
1179 @code{_Sat long _Accum},
1180 @code{_Sat long long _Accum},
1181 @code{_Sat unsigned short _Accum},
1182 @code{_Sat unsigned _Accum},
1183 @code{_Sat unsigned long _Accum},
1184 @code{_Sat unsigned long long _Accum}.
1186 Fixed-point data values contain fractional and optional integral parts.
1187 The format of fixed-point data varies and depends on the target machine.
1189 Support for fixed-point types includes:
1192 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1194 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1196 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1198 binary shift operators (@code{<<}, @code{>>})
1200 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1202 equality operators (@code{==}, @code{!=})
1204 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1205 @code{<<=}, @code{>>=})
1207 conversions to and from integer, floating-point, or fixed-point types
1210 Use a suffix in a fixed-point literal constant:
1212 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1213 @code{_Sat short _Fract}
1214 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1215 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1216 @code{_Sat long _Fract}
1217 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1218 @code{_Sat long long _Fract}
1219 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1220 @code{_Sat unsigned short _Fract}
1221 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1222 @code{_Sat unsigned _Fract}
1223 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1224 @code{_Sat unsigned long _Fract}
1225 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1226 and @code{_Sat unsigned long long _Fract}
1227 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1228 @code{_Sat short _Accum}
1229 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1230 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1231 @code{_Sat long _Accum}
1232 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1233 @code{_Sat long long _Accum}
1234 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1235 @code{_Sat unsigned short _Accum}
1236 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1237 @code{_Sat unsigned _Accum}
1238 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1239 @code{_Sat unsigned long _Accum}
1240 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1241 and @code{_Sat unsigned long long _Accum}
1244 GCC support of fixed-point types as specified by the draft technical report
1249 Pragmas to control overflow and rounding behaviors are not implemented.
1252 Fixed-point types are supported by the DWARF debug information format.
1254 @node Named Address Spaces
1255 @section Named Address Spaces
1256 @cindex Named Address Spaces
1258 As an extension, GNU C supports named address spaces as
1259 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1260 address spaces in GCC will evolve as the draft technical report
1261 changes. Calling conventions for any target might also change. At
1262 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1263 address spaces other than the generic address space.
1265 Address space identifiers may be used exactly like any other C type
1266 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1267 document for more details.
1269 @anchor{AVR Named Address Spaces}
1270 @subsection AVR Named Address Spaces
1272 On the AVR target, there are several address spaces that can be used
1273 in order to put read-only data into the flash memory and access that
1274 data by means of the special instructions @code{LPM} or @code{ELPM}
1275 needed to read from flash.
1277 Per default, any data including read-only data is located in RAM
1278 (the generic address space) so that non-generic address spaces are
1279 needed to locate read-only data in flash memory
1280 @emph{and} to generate the right instructions to access this data
1281 without using (inline) assembler code.
1285 @cindex @code{__flash} AVR Named Address Spaces
1286 The @code{__flash} qualifier locates data in the
1287 @code{.progmem.data} section. Data is read using the @code{LPM}
1288 instruction. Pointers to this address space are 16 bits wide.
1295 @cindex @code{__flash1} AVR Named Address Spaces
1296 @cindex @code{__flash2} AVR Named Address Spaces
1297 @cindex @code{__flash3} AVR Named Address Spaces
1298 @cindex @code{__flash4} AVR Named Address Spaces
1299 @cindex @code{__flash5} AVR Named Address Spaces
1300 These are 16-bit address spaces locating data in section
1301 @code{.progmem@var{N}.data} where @var{N} refers to
1302 address space @code{__flash@var{N}}.
1303 The compiler sets the @code{RAMPZ} segment register appropriately
1304 before reading data by means of the @code{ELPM} instruction.
1307 @cindex @code{__memx} AVR Named Address Spaces
1308 This is a 24-bit address space that linearizes flash and RAM:
1309 If the high bit of the address is set, data is read from
1310 RAM using the lower two bytes as RAM address.
1311 If the high bit of the address is clear, data is read from flash
1312 with @code{RAMPZ} set according to the high byte of the address.
1313 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1315 Objects in this address space are located in @code{.progmemx.data}.
1321 char my_read (const __flash char ** p)
1323 /* p is a pointer to RAM that points to a pointer to flash.
1324 The first indirection of p reads that flash pointer
1325 from RAM and the second indirection reads a char from this
1331 /* Locate array[] in flash memory */
1332 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1338 /* Return 17 by reading from flash memory */
1339 return array[array[i]];
1344 For each named address space supported by avr-gcc there is an equally
1345 named but uppercase built-in macro defined.
1346 The purpose is to facilitate testing if respective address space
1347 support is available or not:
1351 const __flash int var = 1;
1358 #include <avr/pgmspace.h> /* From AVR-LibC */
1360 const int var PROGMEM = 1;
1364 return (int) pgm_read_word (&var);
1366 #endif /* __FLASH */
1370 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1371 locates data in flash but
1372 accesses to these data read from generic address space, i.e.@:
1374 so that you need special accessors like @code{pgm_read_byte}
1375 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1376 together with attribute @code{progmem}.
1379 @b{Limitations and caveats}
1383 Reading across the 64@tie{}KiB section boundary of
1384 the @code{__flash} or @code{__flash@var{N}} address spaces
1385 shows undefined behavior. The only address space that
1386 supports reading across the 64@tie{}KiB flash segment boundaries is
1390 If you use one of the @code{__flash@var{N}} address spaces
1391 you must arrange your linker script to locate the
1392 @code{.progmem@var{N}.data} sections according to your needs.
1395 Any data or pointers to the non-generic address spaces must
1396 be qualified as @code{const}, i.e.@: as read-only data.
1397 This still applies if the data in one of these address
1398 spaces like software version number or calibration lookup table are intended to
1399 be changed after load time by, say, a boot loader. In this case
1400 the right qualification is @code{const} @code{volatile} so that the compiler
1401 must not optimize away known values or insert them
1402 as immediates into operands of instructions.
1405 The following code initializes a variable @code{pfoo}
1406 located in static storage with a 24-bit address:
1408 extern const __memx char foo;
1409 const __memx void *pfoo = &foo;
1413 Such code requires at least binutils 2.23, see
1414 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1418 @subsection M32C Named Address Spaces
1419 @cindex @code{__far} M32C Named Address Spaces
1421 On the M32C target, with the R8C and M16C CPU variants, variables
1422 qualified with @code{__far} are accessed using 32-bit addresses in
1423 order to access memory beyond the first 64@tie{}Ki bytes. If
1424 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1427 @subsection RL78 Named Address Spaces
1428 @cindex @code{__far} RL78 Named Address Spaces
1430 On the RL78 target, variables qualified with @code{__far} are accessed
1431 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1432 addresses. Non-far variables are assumed to appear in the topmost
1433 64@tie{}KiB of the address space.
1435 @subsection SPU Named Address Spaces
1436 @cindex @code{__ea} SPU Named Address Spaces
1438 On the SPU target variables may be declared as
1439 belonging to another address space by qualifying the type with the
1440 @code{__ea} address space identifier:
1447 The compiler generates special code to access the variable @code{i}.
1448 It may use runtime library
1449 support, or generate special machine instructions to access that address
1452 @subsection x86 Named Address Spaces
1453 @cindex x86 named address spaces
1455 On the x86 target, variables may be declared as being relative
1456 to the @code{%fs} or @code{%gs} segments.
1461 @cindex @code{__seg_fs} x86 named address space
1462 @cindex @code{__seg_gs} x86 named address space
1463 The object is accessed with the respective segment override prefix.
1465 The respective segment base must be set via some method specific to
1466 the operating system. Rather than require an expensive system call
1467 to retrieve the segment base, these address spaces are not considered
1468 to be subspaces of the generic (flat) address space. This means that
1469 explicit casts are required to convert pointers between these address
1470 spaces and the generic address space. In practice the application
1471 should cast to @code{uintptr_t} and apply the segment base offset
1472 that it installed previously.
1474 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1475 defined when these address spaces are supported.
1479 @section Arrays of Length Zero
1480 @cindex arrays of length zero
1481 @cindex zero-length arrays
1482 @cindex length-zero arrays
1483 @cindex flexible array members
1485 Zero-length arrays are allowed in GNU C@. They are very useful as the
1486 last element of a structure that is really a header for a variable-length
1495 struct line *thisline = (struct line *)
1496 malloc (sizeof (struct line) + this_length);
1497 thisline->length = this_length;
1500 In ISO C90, you would have to give @code{contents} a length of 1, which
1501 means either you waste space or complicate the argument to @code{malloc}.
1503 In ISO C99, you would use a @dfn{flexible array member}, which is
1504 slightly different in syntax and semantics:
1508 Flexible array members are written as @code{contents[]} without
1512 Flexible array members have incomplete type, and so the @code{sizeof}
1513 operator may not be applied. As a quirk of the original implementation
1514 of zero-length arrays, @code{sizeof} evaluates to zero.
1517 Flexible array members may only appear as the last member of a
1518 @code{struct} that is otherwise non-empty.
1521 A structure containing a flexible array member, or a union containing
1522 such a structure (possibly recursively), may not be a member of a
1523 structure or an element of an array. (However, these uses are
1524 permitted by GCC as extensions.)
1527 Non-empty initialization of zero-length
1528 arrays is treated like any case where there are more initializer
1529 elements than the array holds, in that a suitable warning about ``excess
1530 elements in array'' is given, and the excess elements (all of them, in
1531 this case) are ignored.
1533 GCC allows static initialization of flexible array members.
1534 This is equivalent to defining a new structure containing the original
1535 structure followed by an array of sufficient size to contain the data.
1536 E.g.@: in the following, @code{f1} is constructed as if it were declared
1542 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1545 struct f1 f1; int data[3];
1546 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1550 The convenience of this extension is that @code{f1} has the desired
1551 type, eliminating the need to consistently refer to @code{f2.f1}.
1553 This has symmetry with normal static arrays, in that an array of
1554 unknown size is also written with @code{[]}.
1556 Of course, this extension only makes sense if the extra data comes at
1557 the end of a top-level object, as otherwise we would be overwriting
1558 data at subsequent offsets. To avoid undue complication and confusion
1559 with initialization of deeply nested arrays, we simply disallow any
1560 non-empty initialization except when the structure is the top-level
1561 object. For example:
1564 struct foo @{ int x; int y[]; @};
1565 struct bar @{ struct foo z; @};
1567 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1568 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1569 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1570 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1573 @node Empty Structures
1574 @section Structures with No Members
1575 @cindex empty structures
1576 @cindex zero-size structures
1578 GCC permits a C structure to have no members:
1585 The structure has size zero. In C++, empty structures are part
1586 of the language. G++ treats empty structures as if they had a single
1587 member of type @code{char}.
1589 @node Variable Length
1590 @section Arrays of Variable Length
1591 @cindex variable-length arrays
1592 @cindex arrays of variable length
1595 Variable-length automatic arrays are allowed in ISO C99, and as an
1596 extension GCC accepts them in C90 mode and in C++. These arrays are
1597 declared like any other automatic arrays, but with a length that is not
1598 a constant expression. The storage is allocated at the point of
1599 declaration and deallocated when the block scope containing the declaration
1605 concat_fopen (char *s1, char *s2, char *mode)
1607 char str[strlen (s1) + strlen (s2) + 1];
1610 return fopen (str, mode);
1614 @cindex scope of a variable length array
1615 @cindex variable-length array scope
1616 @cindex deallocating variable length arrays
1617 Jumping or breaking out of the scope of the array name deallocates the
1618 storage. Jumping into the scope is not allowed; you get an error
1621 @cindex variable-length array in a structure
1622 As an extension, GCC accepts variable-length arrays as a member of
1623 a structure or a union. For example:
1629 struct S @{ int x[n]; @};
1633 @cindex @code{alloca} vs variable-length arrays
1634 You can use the function @code{alloca} to get an effect much like
1635 variable-length arrays. The function @code{alloca} is available in
1636 many other C implementations (but not in all). On the other hand,
1637 variable-length arrays are more elegant.
1639 There are other differences between these two methods. Space allocated
1640 with @code{alloca} exists until the containing @emph{function} returns.
1641 The space for a variable-length array is deallocated as soon as the array
1642 name's scope ends, unless you also use @code{alloca} in this scope.
1644 You can also use variable-length arrays as arguments to functions:
1648 tester (int len, char data[len][len])
1654 The length of an array is computed once when the storage is allocated
1655 and is remembered for the scope of the array in case you access it with
1658 If you want to pass the array first and the length afterward, you can
1659 use a forward declaration in the parameter list---another GNU extension.
1663 tester (int len; char data[len][len], int len)
1669 @cindex parameter forward declaration
1670 The @samp{int len} before the semicolon is a @dfn{parameter forward
1671 declaration}, and it serves the purpose of making the name @code{len}
1672 known when the declaration of @code{data} is parsed.
1674 You can write any number of such parameter forward declarations in the
1675 parameter list. They can be separated by commas or semicolons, but the
1676 last one must end with a semicolon, which is followed by the ``real''
1677 parameter declarations. Each forward declaration must match a ``real''
1678 declaration in parameter name and data type. ISO C99 does not support
1679 parameter forward declarations.
1681 @node Variadic Macros
1682 @section Macros with a Variable Number of Arguments.
1683 @cindex variable number of arguments
1684 @cindex macro with variable arguments
1685 @cindex rest argument (in macro)
1686 @cindex variadic macros
1688 In the ISO C standard of 1999, a macro can be declared to accept a
1689 variable number of arguments much as a function can. The syntax for
1690 defining the macro is similar to that of a function. Here is an
1694 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1698 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1699 such a macro, it represents the zero or more tokens until the closing
1700 parenthesis that ends the invocation, including any commas. This set of
1701 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1702 wherever it appears. See the CPP manual for more information.
1704 GCC has long supported variadic macros, and used a different syntax that
1705 allowed you to give a name to the variable arguments just like any other
1706 argument. Here is an example:
1709 #define debug(format, args...) fprintf (stderr, format, args)
1713 This is in all ways equivalent to the ISO C example above, but arguably
1714 more readable and descriptive.
1716 GNU CPP has two further variadic macro extensions, and permits them to
1717 be used with either of the above forms of macro definition.
1719 In standard C, you are not allowed to leave the variable argument out
1720 entirely; but you are allowed to pass an empty argument. For example,
1721 this invocation is invalid in ISO C, because there is no comma after
1728 GNU CPP permits you to completely omit the variable arguments in this
1729 way. In the above examples, the compiler would complain, though since
1730 the expansion of the macro still has the extra comma after the format
1733 To help solve this problem, CPP behaves specially for variable arguments
1734 used with the token paste operator, @samp{##}. If instead you write
1737 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1741 and if the variable arguments are omitted or empty, the @samp{##}
1742 operator causes the preprocessor to remove the comma before it. If you
1743 do provide some variable arguments in your macro invocation, GNU CPP
1744 does not complain about the paste operation and instead places the
1745 variable arguments after the comma. Just like any other pasted macro
1746 argument, these arguments are not macro expanded.
1748 @node Escaped Newlines
1749 @section Slightly Looser Rules for Escaped Newlines
1750 @cindex escaped newlines
1751 @cindex newlines (escaped)
1753 The preprocessor treatment of escaped newlines is more relaxed
1754 than that specified by the C90 standard, which requires the newline
1755 to immediately follow a backslash.
1756 GCC's implementation allows whitespace in the form
1757 of spaces, horizontal and vertical tabs, and form feeds between the
1758 backslash and the subsequent newline. The preprocessor issues a
1759 warning, but treats it as a valid escaped newline and combines the two
1760 lines to form a single logical line. This works within comments and
1761 tokens, as well as between tokens. Comments are @emph{not} treated as
1762 whitespace for the purposes of this relaxation, since they have not
1763 yet been replaced with spaces.
1766 @section Non-Lvalue Arrays May Have Subscripts
1767 @cindex subscripting
1768 @cindex arrays, non-lvalue
1770 @cindex subscripting and function values
1771 In ISO C99, arrays that are not lvalues still decay to pointers, and
1772 may be subscripted, although they may not be modified or used after
1773 the next sequence point and the unary @samp{&} operator may not be
1774 applied to them. As an extension, GNU C allows such arrays to be
1775 subscripted in C90 mode, though otherwise they do not decay to
1776 pointers outside C99 mode. For example,
1777 this is valid in GNU C though not valid in C90:
1781 struct foo @{int a[4];@};
1787 return f().a[index];
1793 @section Arithmetic on @code{void}- and Function-Pointers
1794 @cindex void pointers, arithmetic
1795 @cindex void, size of pointer to
1796 @cindex function pointers, arithmetic
1797 @cindex function, size of pointer to
1799 In GNU C, addition and subtraction operations are supported on pointers to
1800 @code{void} and on pointers to functions. This is done by treating the
1801 size of a @code{void} or of a function as 1.
1803 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1804 and on function types, and returns 1.
1806 @opindex Wpointer-arith
1807 The option @option{-Wpointer-arith} requests a warning if these extensions
1810 @node Pointers to Arrays
1811 @section Pointers to Arrays with Qualifiers Work as Expected
1812 @cindex pointers to arrays
1813 @cindex const qualifier
1815 In GNU C, pointers to arrays with qualifiers work similar to pointers
1816 to other qualified types. For example, a value of type @code{int (*)[5]}
1817 can be used to initialize a variable of type @code{const int (*)[5]}.
1818 These types are incompatible in ISO C because the @code{const} qualifier
1819 is formally attached to the element type of the array and not the
1824 transpose (int N, int M, double out[M][N], const double in[N][M]);
1828 transpose(3, 2, y, x);
1832 @section Non-Constant Initializers
1833 @cindex initializers, non-constant
1834 @cindex non-constant initializers
1836 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1837 automatic variable are not required to be constant expressions in GNU C@.
1838 Here is an example of an initializer with run-time varying elements:
1841 foo (float f, float g)
1843 float beat_freqs[2] = @{ f-g, f+g @};
1848 @node Compound Literals
1849 @section Compound Literals
1850 @cindex constructor expressions
1851 @cindex initializations in expressions
1852 @cindex structures, constructor expression
1853 @cindex expressions, constructor
1854 @cindex compound literals
1855 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1857 ISO C99 supports compound literals. A compound literal looks like
1858 a cast containing an initializer. Its value is an object of the
1859 type specified in the cast, containing the elements specified in
1860 the initializer; it is an lvalue. As an extension, GCC supports
1861 compound literals in C90 mode and in C++, though the semantics are
1862 somewhat different in C++.
1864 Usually, the specified type is a structure. Assume that
1865 @code{struct foo} and @code{structure} are declared as shown:
1868 struct foo @{int a; char b[2];@} structure;
1872 Here is an example of constructing a @code{struct foo} with a compound literal:
1875 structure = ((struct foo) @{x + y, 'a', 0@});
1879 This is equivalent to writing the following:
1883 struct foo temp = @{x + y, 'a', 0@};
1888 You can also construct an array, though this is dangerous in C++, as
1889 explained below. If all the elements of the compound literal are
1890 (made up of) simple constant expressions, suitable for use in
1891 initializers of objects of static storage duration, then the compound
1892 literal can be coerced to a pointer to its first element and used in
1893 such an initializer, as shown here:
1896 char **foo = (char *[]) @{ "x", "y", "z" @};
1899 Compound literals for scalar types and union types are
1900 also allowed, but then the compound literal is equivalent
1903 As a GNU extension, GCC allows initialization of objects with static storage
1904 duration by compound literals (which is not possible in ISO C99, because
1905 the initializer is not a constant).
1906 It is handled as if the object is initialized only with the bracket
1907 enclosed list if the types of the compound literal and the object match.
1908 The initializer list of the compound literal must be constant.
1909 If the object being initialized has array type of unknown size, the size is
1910 determined by compound literal size.
1913 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1914 static int y[] = (int []) @{1, 2, 3@};
1915 static int z[] = (int [3]) @{1@};
1919 The above lines are equivalent to the following:
1921 static struct foo x = @{1, 'a', 'b'@};
1922 static int y[] = @{1, 2, 3@};
1923 static int z[] = @{1, 0, 0@};
1926 In C, a compound literal designates an unnamed object with static or
1927 automatic storage duration. In C++, a compound literal designates a
1928 temporary object, which only lives until the end of its
1929 full-expression. As a result, well-defined C code that takes the
1930 address of a subobject of a compound literal can be undefined in C++,
1931 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1932 For instance, if the array compound literal example above appeared
1933 inside a function, any subsequent use of @samp{foo} in C++ has
1934 undefined behavior because the lifetime of the array ends after the
1935 declaration of @samp{foo}.
1937 As an optimization, the C++ compiler sometimes gives array compound
1938 literals longer lifetimes: when the array either appears outside a
1939 function or has const-qualified type. If @samp{foo} and its
1940 initializer had elements of @samp{char *const} type rather than
1941 @samp{char *}, or if @samp{foo} were a global variable, the array
1942 would have static storage duration. But it is probably safest just to
1943 avoid the use of array compound literals in code compiled as C++.
1945 @node Designated Inits
1946 @section Designated Initializers
1947 @cindex initializers with labeled elements
1948 @cindex labeled elements in initializers
1949 @cindex case labels in initializers
1950 @cindex designated initializers
1952 Standard C90 requires the elements of an initializer to appear in a fixed
1953 order, the same as the order of the elements in the array or structure
1956 In ISO C99 you can give the elements in any order, specifying the array
1957 indices or structure field names they apply to, and GNU C allows this as
1958 an extension in C90 mode as well. This extension is not
1959 implemented in GNU C++.
1961 To specify an array index, write
1962 @samp{[@var{index}] =} before the element value. For example,
1965 int a[6] = @{ [4] = 29, [2] = 15 @};
1972 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1976 The index values must be constant expressions, even if the array being
1977 initialized is automatic.
1979 An alternative syntax for this that has been obsolete since GCC 2.5 but
1980 GCC still accepts is to write @samp{[@var{index}]} before the element
1981 value, with no @samp{=}.
1983 To initialize a range of elements to the same value, write
1984 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1985 extension. For example,
1988 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1992 If the value in it has side-effects, the side-effects happen only once,
1993 not for each initialized field by the range initializer.
1996 Note that the length of the array is the highest value specified
1999 In a structure initializer, specify the name of a field to initialize
2000 with @samp{.@var{fieldname} =} before the element value. For example,
2001 given the following structure,
2004 struct point @{ int x, y; @};
2008 the following initialization
2011 struct point p = @{ .y = yvalue, .x = xvalue @};
2018 struct point p = @{ xvalue, yvalue @};
2021 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2022 @samp{@var{fieldname}:}, as shown here:
2025 struct point p = @{ y: yvalue, x: xvalue @};
2028 Omitted field members are implicitly initialized the same as objects
2029 that have static storage duration.
2032 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2033 @dfn{designator}. You can also use a designator (or the obsolete colon
2034 syntax) when initializing a union, to specify which element of the union
2035 should be used. For example,
2038 union foo @{ int i; double d; @};
2040 union foo f = @{ .d = 4 @};
2044 converts 4 to a @code{double} to store it in the union using
2045 the second element. By contrast, casting 4 to type @code{union foo}
2046 stores it into the union as the integer @code{i}, since it is
2047 an integer. (@xref{Cast to Union}.)
2049 You can combine this technique of naming elements with ordinary C
2050 initialization of successive elements. Each initializer element that
2051 does not have a designator applies to the next consecutive element of the
2052 array or structure. For example,
2055 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2062 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2065 Labeling the elements of an array initializer is especially useful
2066 when the indices are characters or belong to an @code{enum} type.
2071 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2072 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2075 @cindex designator lists
2076 You can also write a series of @samp{.@var{fieldname}} and
2077 @samp{[@var{index}]} designators before an @samp{=} to specify a
2078 nested subobject to initialize; the list is taken relative to the
2079 subobject corresponding to the closest surrounding brace pair. For
2080 example, with the @samp{struct point} declaration above:
2083 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2087 If the same field is initialized multiple times, it has the value from
2088 the last initialization. If any such overridden initialization has
2089 side-effect, it is unspecified whether the side-effect happens or not.
2090 Currently, GCC discards them and issues a warning.
2093 @section Case Ranges
2095 @cindex ranges in case statements
2097 You can specify a range of consecutive values in a single @code{case} label,
2101 case @var{low} ... @var{high}:
2105 This has the same effect as the proper number of individual @code{case}
2106 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2108 This feature is especially useful for ranges of ASCII character codes:
2114 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2115 it may be parsed wrong when you use it with integer values. For example,
2130 @section Cast to a Union Type
2131 @cindex cast to a union
2132 @cindex union, casting to a
2134 A cast to union type is similar to other casts, except that the type
2135 specified is a union type. You can specify the type either with
2136 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2137 a constructor, not a cast, and hence does not yield an lvalue like
2138 normal casts. (@xref{Compound Literals}.)
2140 The types that may be cast to the union type are those of the members
2141 of the union. Thus, given the following union and variables:
2144 union foo @{ int i; double d; @};
2150 both @code{x} and @code{y} can be cast to type @code{union foo}.
2152 Using the cast as the right-hand side of an assignment to a variable of
2153 union type is equivalent to storing in a member of the union:
2158 u = (union foo) x @equiv{} u.i = x
2159 u = (union foo) y @equiv{} u.d = y
2162 You can also use the union cast as a function argument:
2165 void hack (union foo);
2167 hack ((union foo) x);
2170 @node Mixed Declarations
2171 @section Mixed Declarations and Code
2172 @cindex mixed declarations and code
2173 @cindex declarations, mixed with code
2174 @cindex code, mixed with declarations
2176 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2177 within compound statements. As an extension, GNU C also allows this in
2178 C90 mode. For example, you could do:
2187 Each identifier is visible from where it is declared until the end of
2188 the enclosing block.
2190 @node Function Attributes
2191 @section Declaring Attributes of Functions
2192 @cindex function attributes
2193 @cindex declaring attributes of functions
2194 @cindex @code{volatile} applied to function
2195 @cindex @code{const} applied to function
2197 In GNU C, you can use function attributes to declare certain things
2198 about functions called in your program which help the compiler
2199 optimize calls and check your code more carefully. For example, you
2200 can use attributes to declare that a function never returns
2201 (@code{noreturn}), returns a value depending only on its arguments
2202 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2204 You can also use attributes to control memory placement, code
2205 generation options or call/return conventions within the function
2206 being annotated. Many of these attributes are target-specific. For
2207 example, many targets support attributes for defining interrupt
2208 handler functions, which typically must follow special register usage
2209 and return conventions.
2211 Function attributes are introduced by the @code{__attribute__} keyword
2212 on a declaration, followed by an attribute specification inside double
2213 parentheses. You can specify multiple attributes in a declaration by
2214 separating them by commas within the double parentheses or by
2215 immediately following an attribute declaration with another attribute
2216 declaration. @xref{Attribute Syntax}, for the exact rules on
2217 attribute syntax and placement.
2219 GCC also supports attributes on
2220 variable declarations (@pxref{Variable Attributes}),
2221 labels (@pxref{Label Attributes}),
2222 enumerators (@pxref{Enumerator Attributes}),
2223 and types (@pxref{Type Attributes}).
2225 There is some overlap between the purposes of attributes and pragmas
2226 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2227 found convenient to use @code{__attribute__} to achieve a natural
2228 attachment of attributes to their corresponding declarations, whereas
2229 @code{#pragma} is of use for compatibility with other compilers
2230 or constructs that do not naturally form part of the grammar.
2232 In addition to the attributes documented here,
2233 GCC plugins may provide their own attributes.
2236 * Common Function Attributes::
2237 * AArch64 Function Attributes::
2238 * ARC Function Attributes::
2239 * ARM Function Attributes::
2240 * AVR Function Attributes::
2241 * Blackfin Function Attributes::
2242 * CR16 Function Attributes::
2243 * Epiphany Function Attributes::
2244 * H8/300 Function Attributes::
2245 * IA-64 Function Attributes::
2246 * M32C Function Attributes::
2247 * M32R/D Function Attributes::
2248 * m68k Function Attributes::
2249 * MCORE Function Attributes::
2250 * MeP Function Attributes::
2251 * MicroBlaze Function Attributes::
2252 * Microsoft Windows Function Attributes::
2253 * MIPS Function Attributes::
2254 * MSP430 Function Attributes::
2255 * NDS32 Function Attributes::
2256 * Nios II Function Attributes::
2257 * Nvidia PTX Function Attributes::
2258 * PowerPC Function Attributes::
2259 * RL78 Function Attributes::
2260 * RX Function Attributes::
2261 * S/390 Function Attributes::
2262 * SH Function Attributes::
2263 * SPU Function Attributes::
2264 * Symbian OS Function Attributes::
2265 * V850 Function Attributes::
2266 * Visium Function Attributes::
2267 * x86 Function Attributes::
2268 * Xstormy16 Function Attributes::
2271 @node Common Function Attributes
2272 @subsection Common Function Attributes
2274 The following attributes are supported on most targets.
2277 @c Keep this table alphabetized by attribute name. Treat _ as space.
2279 @item alias ("@var{target}")
2280 @cindex @code{alias} function attribute
2281 The @code{alias} attribute causes the declaration to be emitted as an
2282 alias for another symbol, which must be specified. For instance,
2285 void __f () @{ /* @r{Do something.} */; @}
2286 void f () __attribute__ ((weak, alias ("__f")));
2290 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2291 mangled name for the target must be used. It is an error if @samp{__f}
2292 is not defined in the same translation unit.
2294 This attribute requires assembler and object file support,
2295 and may not be available on all targets.
2297 @item aligned (@var{alignment})
2298 @cindex @code{aligned} function attribute
2299 This attribute specifies a minimum alignment for the function,
2302 You cannot use this attribute to decrease the alignment of a function,
2303 only to increase it. However, when you explicitly specify a function
2304 alignment this overrides the effect of the
2305 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2308 Note that the effectiveness of @code{aligned} attributes may be
2309 limited by inherent limitations in your linker. On many systems, the
2310 linker is only able to arrange for functions to be aligned up to a
2311 certain maximum alignment. (For some linkers, the maximum supported
2312 alignment may be very very small.) See your linker documentation for
2313 further information.
2315 The @code{aligned} attribute can also be used for variables and fields
2316 (@pxref{Variable Attributes}.)
2319 @cindex @code{alloc_align} function attribute
2320 The @code{alloc_align} attribute is used to tell the compiler that the
2321 function return value points to memory, where the returned pointer minimum
2322 alignment is given by one of the functions parameters. GCC uses this
2323 information to improve pointer alignment analysis.
2325 The function parameter denoting the allocated alignment is specified by
2326 one integer argument, whose number is the argument of the attribute.
2327 Argument numbering starts at one.
2332 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2336 declares that @code{my_memalign} returns memory with minimum alignment
2337 given by parameter 1.
2340 @cindex @code{alloc_size} function attribute
2341 The @code{alloc_size} attribute is used to tell the compiler that the
2342 function return value points to memory, where the size is given by
2343 one or two of the functions parameters. GCC uses this
2344 information to improve the correctness of @code{__builtin_object_size}.
2346 The function parameter(s) denoting the allocated size are specified by
2347 one or two integer arguments supplied to the attribute. The allocated size
2348 is either the value of the single function argument specified or the product
2349 of the two function arguments specified. Argument numbering starts at
2355 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2356 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2360 declares that @code{my_calloc} returns memory of the size given by
2361 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2362 of the size given by parameter 2.
2365 @cindex @code{always_inline} function attribute
2366 Generally, functions are not inlined unless optimization is specified.
2367 For functions declared inline, this attribute inlines the function
2368 independent of any restrictions that otherwise apply to inlining.
2369 Failure to inline such a function is diagnosed as an error.
2370 Note that if such a function is called indirectly the compiler may
2371 or may not inline it depending on optimization level and a failure
2372 to inline an indirect call may or may not be diagnosed.
2375 @cindex @code{artificial} function attribute
2376 This attribute is useful for small inline wrappers that if possible
2377 should appear during debugging as a unit. Depending on the debug
2378 info format it either means marking the function as artificial
2379 or using the caller location for all instructions within the inlined
2382 @item assume_aligned
2383 @cindex @code{assume_aligned} function attribute
2384 The @code{assume_aligned} attribute is used to tell the compiler that the
2385 function return value points to memory, where the returned pointer minimum
2386 alignment is given by the first argument.
2387 If the attribute has two arguments, the second argument is misalignment offset.
2392 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2393 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2397 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2398 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2401 @item bnd_instrument
2402 @cindex @code{bnd_instrument} function attribute
2403 The @code{bnd_instrument} attribute on functions is used to inform the
2404 compiler that the function should be instrumented when compiled
2405 with the @option{-fchkp-instrument-marked-only} option.
2408 @cindex @code{bnd_legacy} function attribute
2409 @cindex Pointer Bounds Checker attributes
2410 The @code{bnd_legacy} attribute on functions is used to inform the
2411 compiler that the function should not be instrumented when compiled
2412 with the @option{-fcheck-pointer-bounds} option.
2415 @cindex @code{cold} function attribute
2416 The @code{cold} attribute on functions is used to inform the compiler that
2417 the function is unlikely to be executed. The function is optimized for
2418 size rather than speed and on many targets it is placed into a special
2419 subsection of the text section so all cold functions appear close together,
2420 improving code locality of non-cold parts of program. The paths leading
2421 to calls of cold functions within code are marked as unlikely by the branch
2422 prediction mechanism. It is thus useful to mark functions used to handle
2423 unlikely conditions, such as @code{perror}, as cold to improve optimization
2424 of hot functions that do call marked functions in rare occasions.
2426 When profile feedback is available, via @option{-fprofile-use}, cold functions
2427 are automatically detected and this attribute is ignored.
2430 @cindex @code{const} function attribute
2431 @cindex functions that have no side effects
2432 Many functions do not examine any values except their arguments, and
2433 have no effects except the return value. Basically this is just slightly
2434 more strict class than the @code{pure} attribute below, since function is not
2435 allowed to read global memory.
2437 @cindex pointer arguments
2438 Note that a function that has pointer arguments and examines the data
2439 pointed to must @emph{not} be declared @code{const}. Likewise, a
2440 function that calls a non-@code{const} function usually must not be
2441 @code{const}. It does not make sense for a @code{const} function to
2446 @itemx constructor (@var{priority})
2447 @itemx destructor (@var{priority})
2448 @cindex @code{constructor} function attribute
2449 @cindex @code{destructor} function attribute
2450 The @code{constructor} attribute causes the function to be called
2451 automatically before execution enters @code{main ()}. Similarly, the
2452 @code{destructor} attribute causes the function to be called
2453 automatically after @code{main ()} completes or @code{exit ()} is
2454 called. Functions with these attributes are useful for
2455 initializing data that is used implicitly during the execution of
2458 You may provide an optional integer priority to control the order in
2459 which constructor and destructor functions are run. A constructor
2460 with a smaller priority number runs before a constructor with a larger
2461 priority number; the opposite relationship holds for destructors. So,
2462 if you have a constructor that allocates a resource and a destructor
2463 that deallocates the same resource, both functions typically have the
2464 same priority. The priorities for constructor and destructor
2465 functions are the same as those specified for namespace-scope C++
2466 objects (@pxref{C++ Attributes}).
2468 These attributes are not currently implemented for Objective-C@.
2471 @itemx deprecated (@var{msg})
2472 @cindex @code{deprecated} function attribute
2473 The @code{deprecated} attribute results in a warning if the function
2474 is used anywhere in the source file. This is useful when identifying
2475 functions that are expected to be removed in a future version of a
2476 program. The warning also includes the location of the declaration
2477 of the deprecated function, to enable users to easily find further
2478 information about why the function is deprecated, or what they should
2479 do instead. Note that the warnings only occurs for uses:
2482 int old_fn () __attribute__ ((deprecated));
2484 int (*fn_ptr)() = old_fn;
2488 results in a warning on line 3 but not line 2. The optional @var{msg}
2489 argument, which must be a string, is printed in the warning if
2492 The @code{deprecated} attribute can also be used for variables and
2493 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2495 @item error ("@var{message}")
2496 @itemx warning ("@var{message}")
2497 @cindex @code{error} function attribute
2498 @cindex @code{warning} function attribute
2499 If the @code{error} or @code{warning} attribute
2500 is used on a function declaration and a call to such a function
2501 is not eliminated through dead code elimination or other optimizations,
2502 an error or warning (respectively) that includes @var{message} is diagnosed.
2504 for compile-time checking, especially together with @code{__builtin_constant_p}
2505 and inline functions where checking the inline function arguments is not
2506 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2508 While it is possible to leave the function undefined and thus invoke
2509 a link failure (to define the function with
2510 a message in @code{.gnu.warning*} section),
2511 when using these attributes the problem is diagnosed
2512 earlier and with exact location of the call even in presence of inline
2513 functions or when not emitting debugging information.
2515 @item externally_visible
2516 @cindex @code{externally_visible} function attribute
2517 This attribute, attached to a global variable or function, nullifies
2518 the effect of the @option{-fwhole-program} command-line option, so the
2519 object remains visible outside the current compilation unit.
2521 If @option{-fwhole-program} is used together with @option{-flto} and
2522 @command{gold} is used as the linker plugin,
2523 @code{externally_visible} attributes are automatically added to functions
2524 (not variable yet due to a current @command{gold} issue)
2525 that are accessed outside of LTO objects according to resolution file
2526 produced by @command{gold}.
2527 For other linkers that cannot generate resolution file,
2528 explicit @code{externally_visible} attributes are still necessary.
2531 @cindex @code{flatten} function attribute
2532 Generally, inlining into a function is limited. For a function marked with
2533 this attribute, every call inside this function is inlined, if possible.
2534 Whether the function itself is considered for inlining depends on its size and
2535 the current inlining parameters.
2537 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2538 @cindex @code{format} function attribute
2539 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2541 The @code{format} attribute specifies that a function takes @code{printf},
2542 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2543 should be type-checked against a format string. For example, the
2548 my_printf (void *my_object, const char *my_format, ...)
2549 __attribute__ ((format (printf, 2, 3)));
2553 causes the compiler to check the arguments in calls to @code{my_printf}
2554 for consistency with the @code{printf} style format string argument
2557 The parameter @var{archetype} determines how the format string is
2558 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2559 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2560 @code{strfmon}. (You can also use @code{__printf__},
2561 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2562 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2563 @code{ms_strftime} are also present.
2564 @var{archetype} values such as @code{printf} refer to the formats accepted
2565 by the system's C runtime library,
2566 while values prefixed with @samp{gnu_} always refer
2567 to the formats accepted by the GNU C Library. On Microsoft Windows
2568 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2569 @file{msvcrt.dll} library.
2570 The parameter @var{string-index}
2571 specifies which argument is the format string argument (starting
2572 from 1), while @var{first-to-check} is the number of the first
2573 argument to check against the format string. For functions
2574 where the arguments are not available to be checked (such as
2575 @code{vprintf}), specify the third parameter as zero. In this case the
2576 compiler only checks the format string for consistency. For
2577 @code{strftime} formats, the third parameter is required to be zero.
2578 Since non-static C++ methods have an implicit @code{this} argument, the
2579 arguments of such methods should be counted from two, not one, when
2580 giving values for @var{string-index} and @var{first-to-check}.
2582 In the example above, the format string (@code{my_format}) is the second
2583 argument of the function @code{my_print}, and the arguments to check
2584 start with the third argument, so the correct parameters for the format
2585 attribute are 2 and 3.
2587 @opindex ffreestanding
2588 @opindex fno-builtin
2589 The @code{format} attribute allows you to identify your own functions
2590 that take format strings as arguments, so that GCC can check the
2591 calls to these functions for errors. The compiler always (unless
2592 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2593 for the standard library functions @code{printf}, @code{fprintf},
2594 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2595 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2596 warnings are requested (using @option{-Wformat}), so there is no need to
2597 modify the header file @file{stdio.h}. In C99 mode, the functions
2598 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2599 @code{vsscanf} are also checked. Except in strictly conforming C
2600 standard modes, the X/Open function @code{strfmon} is also checked as
2601 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2602 @xref{C Dialect Options,,Options Controlling C Dialect}.
2604 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2605 recognized in the same context. Declarations including these format attributes
2606 are parsed for correct syntax, however the result of checking of such format
2607 strings is not yet defined, and is not carried out by this version of the
2610 The target may also provide additional types of format checks.
2611 @xref{Target Format Checks,,Format Checks Specific to Particular
2614 @item format_arg (@var{string-index})
2615 @cindex @code{format_arg} function attribute
2616 @opindex Wformat-nonliteral
2617 The @code{format_arg} attribute specifies that a function takes a format
2618 string for a @code{printf}, @code{scanf}, @code{strftime} or
2619 @code{strfmon} style function and modifies it (for example, to translate
2620 it into another language), so the result can be passed to a
2621 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2622 function (with the remaining arguments to the format function the same
2623 as they would have been for the unmodified string). For example, the
2628 my_dgettext (char *my_domain, const char *my_format)
2629 __attribute__ ((format_arg (2)));
2633 causes the compiler to check the arguments in calls to a @code{printf},
2634 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2635 format string argument is a call to the @code{my_dgettext} function, for
2636 consistency with the format string argument @code{my_format}. If the
2637 @code{format_arg} attribute had not been specified, all the compiler
2638 could tell in such calls to format functions would be that the format
2639 string argument is not constant; this would generate a warning when
2640 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2641 without the attribute.
2643 The parameter @var{string-index} specifies which argument is the format
2644 string argument (starting from one). Since non-static C++ methods have
2645 an implicit @code{this} argument, the arguments of such methods should
2646 be counted from two.
2648 The @code{format_arg} attribute allows you to identify your own
2649 functions that modify format strings, so that GCC can check the
2650 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2651 type function whose operands are a call to one of your own function.
2652 The compiler always treats @code{gettext}, @code{dgettext}, and
2653 @code{dcgettext} in this manner except when strict ISO C support is
2654 requested by @option{-ansi} or an appropriate @option{-std} option, or
2655 @option{-ffreestanding} or @option{-fno-builtin}
2656 is used. @xref{C Dialect Options,,Options
2657 Controlling C Dialect}.
2659 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2660 @code{NSString} reference for compatibility with the @code{format} attribute
2663 The target may also allow additional types in @code{format-arg} attributes.
2664 @xref{Target Format Checks,,Format Checks Specific to Particular
2668 @cindex @code{gnu_inline} function attribute
2669 This attribute should be used with a function that is also declared
2670 with the @code{inline} keyword. It directs GCC to treat the function
2671 as if it were defined in gnu90 mode even when compiling in C99 or
2674 If the function is declared @code{extern}, then this definition of the
2675 function is used only for inlining. In no case is the function
2676 compiled as a standalone function, not even if you take its address
2677 explicitly. Such an address becomes an external reference, as if you
2678 had only declared the function, and had not defined it. This has
2679 almost the effect of a macro. The way to use this is to put a
2680 function definition in a header file with this attribute, and put
2681 another copy of the function, without @code{extern}, in a library
2682 file. The definition in the header file causes most calls to the
2683 function to be inlined. If any uses of the function remain, they
2684 refer to the single copy in the library. Note that the two
2685 definitions of the functions need not be precisely the same, although
2686 if they do not have the same effect your program may behave oddly.
2688 In C, if the function is neither @code{extern} nor @code{static}, then
2689 the function is compiled as a standalone function, as well as being
2690 inlined where possible.
2692 This is how GCC traditionally handled functions declared
2693 @code{inline}. Since ISO C99 specifies a different semantics for
2694 @code{inline}, this function attribute is provided as a transition
2695 measure and as a useful feature in its own right. This attribute is
2696 available in GCC 4.1.3 and later. It is available if either of the
2697 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2698 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2699 Function is As Fast As a Macro}.
2701 In C++, this attribute does not depend on @code{extern} in any way,
2702 but it still requires the @code{inline} keyword to enable its special
2706 @cindex @code{hot} function attribute
2707 The @code{hot} attribute on a function is used to inform the compiler that
2708 the function is a hot spot of the compiled program. The function is
2709 optimized more aggressively and on many targets it is placed into a special
2710 subsection of the text section so all hot functions appear close together,
2713 When profile feedback is available, via @option{-fprofile-use}, hot functions
2714 are automatically detected and this attribute is ignored.
2716 @item ifunc ("@var{resolver}")
2717 @cindex @code{ifunc} function attribute
2718 @cindex indirect functions
2719 @cindex functions that are dynamically resolved
2720 The @code{ifunc} attribute is used to mark a function as an indirect
2721 function using the STT_GNU_IFUNC symbol type extension to the ELF
2722 standard. This allows the resolution of the symbol value to be
2723 determined dynamically at load time, and an optimized version of the
2724 routine can be selected for the particular processor or other system
2725 characteristics determined then. To use this attribute, first define
2726 the implementation functions available, and a resolver function that
2727 returns a pointer to the selected implementation function. The
2728 implementation functions' declarations must match the API of the
2729 function being implemented, the resolver's declaration is be a
2730 function returning pointer to void function returning void:
2733 void *my_memcpy (void *dst, const void *src, size_t len)
2738 static void (*resolve_memcpy (void)) (void)
2740 return my_memcpy; // we'll just always select this routine
2745 The exported header file declaring the function the user calls would
2749 extern void *memcpy (void *, const void *, size_t);
2753 allowing the user to call this as a regular function, unaware of the
2754 implementation. Finally, the indirect function needs to be defined in
2755 the same translation unit as the resolver function:
2758 void *memcpy (void *, const void *, size_t)
2759 __attribute__ ((ifunc ("resolve_memcpy")));
2762 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2763 and GNU C Library version 2.11.1 are required to use this feature.
2766 @itemx interrupt_handler
2767 Many GCC back ends support attributes to indicate that a function is
2768 an interrupt handler, which tells the compiler to generate function
2769 entry and exit sequences that differ from those from regular
2770 functions. The exact syntax and behavior are target-specific;
2771 refer to the following subsections for details.
2774 @cindex @code{leaf} function attribute
2775 Calls to external functions with this attribute must return to the
2776 current compilation unit only by return or by exception handling. In
2777 particular, a leaf function is not allowed to invoke callback functions
2778 passed to it from the current compilation unit, directly call functions
2779 exported by the unit, or @code{longjmp} into the unit. Leaf functions
2780 might still call functions from other compilation units and thus they
2781 are not necessarily leaf in the sense that they contain no function
2784 The attribute is intended for library functions to improve dataflow
2785 analysis. The compiler takes the hint that any data not escaping the
2786 current compilation unit cannot be used or modified by the leaf
2787 function. For example, the @code{sin} function is a leaf function, but
2788 @code{qsort} is not.
2790 Note that leaf functions might indirectly run a signal handler defined
2791 in the current compilation unit that uses static variables. Similarly,
2792 when lazy symbol resolution is in effect, leaf functions might invoke
2793 indirect functions whose resolver function or implementation function is
2794 defined in the current compilation unit and uses static variables. There
2795 is no standard-compliant way to write such a signal handler, resolver
2796 function, or implementation function, and the best that you can do is to
2797 remove the @code{leaf} attribute or mark all such static variables
2798 @code{volatile}. Lastly, for ELF-based systems that support symbol
2799 interposition, care should be taken that functions defined in the
2800 current compilation unit do not unexpectedly interpose other symbols
2801 based on the defined standards mode and defined feature test macros;
2802 otherwise an inadvertent callback would be added.
2804 The attribute has no effect on functions defined within the current
2805 compilation unit. This is to allow easy merging of multiple compilation
2806 units into one, for example, by using the link-time optimization. For
2807 this reason the attribute is not allowed on types to annotate indirect
2811 @cindex @code{malloc} function attribute
2812 @cindex functions that behave like malloc
2813 This tells the compiler that a function is @code{malloc}-like, i.e.,
2814 that the pointer @var{P} returned by the function cannot alias any
2815 other pointer valid when the function returns, and moreover no
2816 pointers to valid objects occur in any storage addressed by @var{P}.
2818 Using this attribute can improve optimization. Functions like
2819 @code{malloc} and @code{calloc} have this property because they return
2820 a pointer to uninitialized or zeroed-out storage. However, functions
2821 like @code{realloc} do not have this property, as they can return a
2822 pointer to storage containing pointers.
2825 @cindex @code{no_icf} function attribute
2826 This function attribute prevents a functions from being merged with another
2827 semantically equivalent function.
2829 @item no_instrument_function
2830 @cindex @code{no_instrument_function} function attribute
2831 @opindex finstrument-functions
2832 If @option{-finstrument-functions} is given, profiling function calls are
2833 generated at entry and exit of most user-compiled functions.
2834 Functions with this attribute are not so instrumented.
2837 @cindex @code{no_reorder} function attribute
2838 Do not reorder functions or variables marked @code{no_reorder}
2839 against each other or top level assembler statements the executable.
2840 The actual order in the program will depend on the linker command
2841 line. Static variables marked like this are also not removed.
2842 This has a similar effect
2843 as the @option{-fno-toplevel-reorder} option, but only applies to the
2846 @item no_sanitize_address
2847 @itemx no_address_safety_analysis
2848 @cindex @code{no_sanitize_address} function attribute
2849 The @code{no_sanitize_address} attribute on functions is used
2850 to inform the compiler that it should not instrument memory accesses
2851 in the function when compiling with the @option{-fsanitize=address} option.
2852 The @code{no_address_safety_analysis} is a deprecated alias of the
2853 @code{no_sanitize_address} attribute, new code should use
2854 @code{no_sanitize_address}.
2856 @item no_sanitize_thread
2857 @cindex @code{no_sanitize_thread} function attribute
2858 The @code{no_sanitize_thread} attribute on functions is used
2859 to inform the compiler that it should not instrument memory accesses
2860 in the function when compiling with the @option{-fsanitize=thread} option.
2862 @item no_sanitize_undefined
2863 @cindex @code{no_sanitize_undefined} function attribute
2864 The @code{no_sanitize_undefined} attribute on functions is used
2865 to inform the compiler that it should not check for undefined behavior
2866 in the function when compiling with the @option{-fsanitize=undefined} option.
2868 @item no_split_stack
2869 @cindex @code{no_split_stack} function attribute
2870 @opindex fsplit-stack
2871 If @option{-fsplit-stack} is given, functions have a small
2872 prologue which decides whether to split the stack. Functions with the
2873 @code{no_split_stack} attribute do not have that prologue, and thus
2874 may run with only a small amount of stack space available.
2876 @item no_stack_limit
2877 @cindex @code{no_stack_limit} function attribute
2878 This attribute locally overrides the @option{-fstack-limit-register}
2879 and @option{-fstack-limit-symbol} command-line options; it has the effect
2880 of disabling stack limit checking in the function it applies to.
2883 @cindex @code{noclone} function attribute
2884 This function attribute prevents a function from being considered for
2885 cloning---a mechanism that produces specialized copies of functions
2886 and which is (currently) performed by interprocedural constant
2890 @cindex @code{noinline} function attribute
2891 This function attribute prevents a function from being considered for
2893 @c Don't enumerate the optimizations by name here; we try to be
2894 @c future-compatible with this mechanism.
2895 If the function does not have side-effects, there are optimizations
2896 other than inlining that cause function calls to be optimized away,
2897 although the function call is live. To keep such calls from being
2904 (@pxref{Extended Asm}) in the called function, to serve as a special
2907 @item nonnull (@var{arg-index}, @dots{})
2908 @cindex @code{nonnull} function attribute
2909 @cindex functions with non-null pointer arguments
2910 The @code{nonnull} attribute specifies that some function parameters should
2911 be non-null pointers. For instance, the declaration:
2915 my_memcpy (void *dest, const void *src, size_t len)
2916 __attribute__((nonnull (1, 2)));
2920 causes the compiler to check that, in calls to @code{my_memcpy},
2921 arguments @var{dest} and @var{src} are non-null. If the compiler
2922 determines that a null pointer is passed in an argument slot marked
2923 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2924 is issued. The compiler may also choose to make optimizations based
2925 on the knowledge that certain function arguments will never be null.
2927 If no argument index list is given to the @code{nonnull} attribute,
2928 all pointer arguments are marked as non-null. To illustrate, the
2929 following declaration is equivalent to the previous example:
2933 my_memcpy (void *dest, const void *src, size_t len)
2934 __attribute__((nonnull));
2938 @cindex @code{noplt} function attribute
2939 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
2940 Calls to functions marked with this attribute in position-independent code
2945 /* Externally defined function foo. */
2946 int foo () __attribute__ ((noplt));
2949 main (/* @r{@dots{}} */)
2958 The @code{noplt} attribute on function @code{foo}
2959 tells the compiler to assume that
2960 the function @code{foo} is externally defined and that the call to
2961 @code{foo} must avoid the PLT
2962 in position-independent code.
2964 In position-dependent code, a few targets also convert calls to
2965 functions that are marked to not use the PLT to use the GOT instead.
2968 @cindex @code{noreturn} function attribute
2969 @cindex functions that never return
2970 A few standard library functions, such as @code{abort} and @code{exit},
2971 cannot return. GCC knows this automatically. Some programs define
2972 their own functions that never return. You can declare them
2973 @code{noreturn} to tell the compiler this fact. For example,
2977 void fatal () __attribute__ ((noreturn));
2980 fatal (/* @r{@dots{}} */)
2982 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2988 The @code{noreturn} keyword tells the compiler to assume that
2989 @code{fatal} cannot return. It can then optimize without regard to what
2990 would happen if @code{fatal} ever did return. This makes slightly
2991 better code. More importantly, it helps avoid spurious warnings of
2992 uninitialized variables.
2994 The @code{noreturn} keyword does not affect the exceptional path when that
2995 applies: a @code{noreturn}-marked function may still return to the caller
2996 by throwing an exception or calling @code{longjmp}.
2998 Do not assume that registers saved by the calling function are
2999 restored before calling the @code{noreturn} function.
3001 It does not make sense for a @code{noreturn} function to have a return
3002 type other than @code{void}.
3005 @cindex @code{nothrow} function attribute
3006 The @code{nothrow} attribute is used to inform the compiler that a
3007 function cannot throw an exception. For example, most functions in
3008 the standard C library can be guaranteed not to throw an exception
3009 with the notable exceptions of @code{qsort} and @code{bsearch} that
3010 take function pointer arguments.
3013 @cindex @code{optimize} function attribute
3014 The @code{optimize} attribute is used to specify that a function is to
3015 be compiled with different optimization options than specified on the
3016 command line. Arguments can either be numbers or strings. Numbers
3017 are assumed to be an optimization level. Strings that begin with
3018 @code{O} are assumed to be an optimization option, while other options
3019 are assumed to be used with a @code{-f} prefix. You can also use the
3020 @samp{#pragma GCC optimize} pragma to set the optimization options
3021 that affect more than one function.
3022 @xref{Function Specific Option Pragmas}, for details about the
3023 @samp{#pragma GCC optimize} pragma.
3025 This can be used for instance to have frequently-executed functions
3026 compiled with more aggressive optimization options that produce faster
3027 and larger code, while other functions can be compiled with less
3031 @cindex @code{pure} function attribute
3032 @cindex functions that have no side effects
3033 Many functions have no effects except the return value and their
3034 return value depends only on the parameters and/or global variables.
3035 Such a function can be subject
3036 to common subexpression elimination and loop optimization just as an
3037 arithmetic operator would be. These functions should be declared
3038 with the attribute @code{pure}. For example,
3041 int square (int) __attribute__ ((pure));
3045 says that the hypothetical function @code{square} is safe to call
3046 fewer times than the program says.
3048 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3049 Interesting non-pure functions are functions with infinite loops or those
3050 depending on volatile memory or other system resource, that may change between
3051 two consecutive calls (such as @code{feof} in a multithreading environment).
3053 @item returns_nonnull
3054 @cindex @code{returns_nonnull} function attribute
3055 The @code{returns_nonnull} attribute specifies that the function
3056 return value should be a non-null pointer. For instance, the declaration:
3060 mymalloc (size_t len) __attribute__((returns_nonnull));
3064 lets the compiler optimize callers based on the knowledge
3065 that the return value will never be null.
3068 @cindex @code{returns_twice} function attribute
3069 @cindex functions that return more than once
3070 The @code{returns_twice} attribute tells the compiler that a function may
3071 return more than one time. The compiler ensures that all registers
3072 are dead before calling such a function and emits a warning about
3073 the variables that may be clobbered after the second return from the
3074 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3075 The @code{longjmp}-like counterpart of such function, if any, might need
3076 to be marked with the @code{noreturn} attribute.
3078 @item section ("@var{section-name}")
3079 @cindex @code{section} function attribute
3080 @cindex functions in arbitrary sections
3081 Normally, the compiler places the code it generates in the @code{text} section.
3082 Sometimes, however, you need additional sections, or you need certain
3083 particular functions to appear in special sections. The @code{section}
3084 attribute specifies that a function lives in a particular section.
3085 For example, the declaration:
3088 extern void foobar (void) __attribute__ ((section ("bar")));
3092 puts the function @code{foobar} in the @code{bar} section.
3094 Some file formats do not support arbitrary sections so the @code{section}
3095 attribute is not available on all platforms.
3096 If you need to map the entire contents of a module to a particular
3097 section, consider using the facilities of the linker instead.
3100 @cindex @code{sentinel} function attribute
3101 This function attribute ensures that a parameter in a function call is
3102 an explicit @code{NULL}. The attribute is only valid on variadic
3103 functions. By default, the sentinel is located at position zero, the
3104 last parameter of the function call. If an optional integer position
3105 argument P is supplied to the attribute, the sentinel must be located at
3106 position P counting backwards from the end of the argument list.
3109 __attribute__ ((sentinel))
3111 __attribute__ ((sentinel(0)))
3114 The attribute is automatically set with a position of 0 for the built-in
3115 functions @code{execl} and @code{execlp}. The built-in function
3116 @code{execle} has the attribute set with a position of 1.
3118 A valid @code{NULL} in this context is defined as zero with any pointer
3119 type. If your system defines the @code{NULL} macro with an integer type
3120 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3121 with a copy that redefines NULL appropriately.
3123 The warnings for missing or incorrect sentinels are enabled with
3127 @itemx simd("@var{mask}")
3128 @cindex @code{simd} function attribute
3129 This attribute enables creation of one or more function versions that
3130 can process multiple arguments using SIMD instructions from a
3131 single invocation. Specifying this attribute allows compiler to
3132 assume that such versions are available at link time (provided
3133 in the same or another translation unit). Generated versions are
3134 target-dependent and described in the corresponding Vector ABI document. For
3135 x86_64 target this document can be found
3136 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3138 The optional argument @var{mask} may have the value
3139 @code{notinbranch} or @code{inbranch},
3140 and instructs the compiler to generate non-masked or masked
3141 clones correspondingly. By default, all clones are generated.
3143 The attribute should not be used together with Cilk Plus @code{vector}
3144 attribute on the same function.
3146 If the attribute is specified and @code{#pragma omp declare simd} is
3147 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3148 switch is specified, then the attribute is ignored.
3151 @cindex @code{stack_protect} function attribute
3152 This attribute adds stack protection code to the function if
3153 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3154 or @option{-fstack-protector-explicit} are set.
3156 @item target (@var{options})
3157 @cindex @code{target} function attribute
3158 Multiple target back ends implement the @code{target} attribute
3159 to specify that a function is to
3160 be compiled with different target options than specified on the
3161 command line. This can be used for instance to have functions
3162 compiled with a different ISA (instruction set architecture) than the
3163 default. You can also use the @samp{#pragma GCC target} pragma to set
3164 more than one function to be compiled with specific target options.
3165 @xref{Function Specific Option Pragmas}, for details about the
3166 @samp{#pragma GCC target} pragma.
3168 For instance, on an x86, you could declare one function with the
3169 @code{target("sse4.1,arch=core2")} attribute and another with
3170 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3171 compiling the first function with @option{-msse4.1} and
3172 @option{-march=core2} options, and the second function with
3173 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3174 to make sure that a function is only invoked on a machine that
3175 supports the particular ISA it is compiled for (for example by using
3176 @code{cpuid} on x86 to determine what feature bits and architecture
3180 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3181 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3184 You can either use multiple
3185 strings separated by commas to specify multiple options,
3186 or separate the options with a comma (@samp{,}) within a single string.
3188 The options supported are specific to each target; refer to @ref{x86
3189 Function Attributes}, @ref{PowerPC Function Attributes},
3190 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3193 @item target_clones (@var{options})
3194 @cindex @code{target_clones} function attribute
3195 The @code{target_clones} attribute is used to specify that a function
3196 be cloned into multiple versions compiled with different target options
3197 than specified on the command line. The supported options and restrictions
3198 are the same as for @code{target} attribute.
3200 For instance, on an x86, you could compile a function with
3201 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3202 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3203 It also creates a resolver function (see the @code{ifunc} attribute
3204 above) that dynamically selects a clone suitable for current architecture.
3207 @cindex @code{unused} function attribute
3208 This attribute, attached to a function, means that the function is meant
3209 to be possibly unused. GCC does not produce a warning for this
3213 @cindex @code{used} function attribute
3214 This attribute, attached to a function, means that code must be emitted
3215 for the function even if it appears that the function is not referenced.
3216 This is useful, for example, when the function is referenced only in
3219 When applied to a member function of a C++ class template, the
3220 attribute also means that the function is instantiated if the
3221 class itself is instantiated.
3223 @item visibility ("@var{visibility_type}")
3224 @cindex @code{visibility} function attribute
3225 This attribute affects the linkage of the declaration to which it is attached.
3226 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3227 (@pxref{Common Type Attributes}) as well as functions.
3229 There are four supported @var{visibility_type} values: default,
3230 hidden, protected or internal visibility.
3233 void __attribute__ ((visibility ("protected")))
3234 f () @{ /* @r{Do something.} */; @}
3235 int i __attribute__ ((visibility ("hidden")));
3238 The possible values of @var{visibility_type} correspond to the
3239 visibility settings in the ELF gABI.
3242 @c keep this list of visibilities in alphabetical order.
3245 Default visibility is the normal case for the object file format.
3246 This value is available for the visibility attribute to override other
3247 options that may change the assumed visibility of entities.
3249 On ELF, default visibility means that the declaration is visible to other
3250 modules and, in shared libraries, means that the declared entity may be
3253 On Darwin, default visibility means that the declaration is visible to
3256 Default visibility corresponds to ``external linkage'' in the language.
3259 Hidden visibility indicates that the entity declared has a new
3260 form of linkage, which we call ``hidden linkage''. Two
3261 declarations of an object with hidden linkage refer to the same object
3262 if they are in the same shared object.
3265 Internal visibility is like hidden visibility, but with additional
3266 processor specific semantics. Unless otherwise specified by the
3267 psABI, GCC defines internal visibility to mean that a function is
3268 @emph{never} called from another module. Compare this with hidden
3269 functions which, while they cannot be referenced directly by other
3270 modules, can be referenced indirectly via function pointers. By
3271 indicating that a function cannot be called from outside the module,
3272 GCC may for instance omit the load of a PIC register since it is known
3273 that the calling function loaded the correct value.
3276 Protected visibility is like default visibility except that it
3277 indicates that references within the defining module bind to the
3278 definition in that module. That is, the declared entity cannot be
3279 overridden by another module.
3283 All visibilities are supported on many, but not all, ELF targets
3284 (supported when the assembler supports the @samp{.visibility}
3285 pseudo-op). Default visibility is supported everywhere. Hidden
3286 visibility is supported on Darwin targets.
3288 The visibility attribute should be applied only to declarations that
3289 would otherwise have external linkage. The attribute should be applied
3290 consistently, so that the same entity should not be declared with
3291 different settings of the attribute.
3293 In C++, the visibility attribute applies to types as well as functions
3294 and objects, because in C++ types have linkage. A class must not have
3295 greater visibility than its non-static data member types and bases,
3296 and class members default to the visibility of their class. Also, a
3297 declaration without explicit visibility is limited to the visibility
3300 In C++, you can mark member functions and static member variables of a
3301 class with the visibility attribute. This is useful if you know a
3302 particular method or static member variable should only be used from
3303 one shared object; then you can mark it hidden while the rest of the
3304 class has default visibility. Care must be taken to avoid breaking
3305 the One Definition Rule; for example, it is usually not useful to mark
3306 an inline method as hidden without marking the whole class as hidden.
3308 A C++ namespace declaration can also have the visibility attribute.
3311 namespace nspace1 __attribute__ ((visibility ("protected")))
3312 @{ /* @r{Do something.} */; @}
3315 This attribute applies only to the particular namespace body, not to
3316 other definitions of the same namespace; it is equivalent to using
3317 @samp{#pragma GCC visibility} before and after the namespace
3318 definition (@pxref{Visibility Pragmas}).
3320 In C++, if a template argument has limited visibility, this
3321 restriction is implicitly propagated to the template instantiation.
3322 Otherwise, template instantiations and specializations default to the
3323 visibility of their template.
3325 If both the template and enclosing class have explicit visibility, the
3326 visibility from the template is used.
3328 @item warn_unused_result
3329 @cindex @code{warn_unused_result} function attribute
3330 The @code{warn_unused_result} attribute causes a warning to be emitted
3331 if a caller of the function with this attribute does not use its
3332 return value. This is useful for functions where not checking
3333 the result is either a security problem or always a bug, such as
3337 int fn () __attribute__ ((warn_unused_result));
3340 if (fn () < 0) return -1;
3347 results in warning on line 5.
3350 @cindex @code{weak} function attribute
3351 The @code{weak} attribute causes the declaration to be emitted as a weak
3352 symbol rather than a global. This is primarily useful in defining
3353 library functions that can be overridden in user code, though it can
3354 also be used with non-function declarations. Weak symbols are supported
3355 for ELF targets, and also for a.out targets when using the GNU assembler
3359 @itemx weakref ("@var{target}")
3360 @cindex @code{weakref} function attribute
3361 The @code{weakref} attribute marks a declaration as a weak reference.
3362 Without arguments, it should be accompanied by an @code{alias} attribute
3363 naming the target symbol. Optionally, the @var{target} may be given as
3364 an argument to @code{weakref} itself. In either case, @code{weakref}
3365 implicitly marks the declaration as @code{weak}. Without a
3366 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3367 @code{weakref} is equivalent to @code{weak}.
3370 static int x() __attribute__ ((weakref ("y")));
3371 /* is equivalent to... */
3372 static int x() __attribute__ ((weak, weakref, alias ("y")));
3374 static int x() __attribute__ ((weakref));
3375 static int x() __attribute__ ((alias ("y")));
3378 A weak reference is an alias that does not by itself require a
3379 definition to be given for the target symbol. If the target symbol is
3380 only referenced through weak references, then it becomes a @code{weak}
3381 undefined symbol. If it is directly referenced, however, then such
3382 strong references prevail, and a definition is required for the
3383 symbol, not necessarily in the same translation unit.
3385 The effect is equivalent to moving all references to the alias to a
3386 separate translation unit, renaming the alias to the aliased symbol,
3387 declaring it as weak, compiling the two separate translation units and
3388 performing a reloadable link on them.
3390 At present, a declaration to which @code{weakref} is attached can
3391 only be @code{static}.
3396 @c This is the end of the target-independent attribute table
3398 @node AArch64 Function Attributes
3399 @subsection AArch64 Function Attributes
3401 The following target-specific function attributes are available for the
3402 AArch64 target. For the most part, these options mirror the behavior of
3403 similar command-line options (@pxref{AArch64 Options}), but on a
3407 @item general-regs-only
3408 @cindex @code{general-regs-only} function attribute, AArch64
3409 Indicates that no floating-point or Advanced SIMD registers should be
3410 used when generating code for this function. If the function explicitly
3411 uses floating-point code, then the compiler gives an error. This is
3412 the same behavior as that of the command-line option
3413 @option{-mgeneral-regs-only}.
3415 @item fix-cortex-a53-835769
3416 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3417 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3418 applied to this function. To explicitly disable the workaround for this
3419 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3420 This corresponds to the behavior of the command line options
3421 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3424 @cindex @code{cmodel=} function attribute, AArch64
3425 Indicates that code should be generated for a particular code model for
3426 this function. The behavior and permissible arguments are the same as
3427 for the command line option @option{-mcmodel=}.
3430 @cindex @code{strict-align} function attribute, AArch64
3431 Indicates that the compiler should not assume that unaligned memory references
3432 are handled by the system. The behavior is the same as for the command-line
3433 option @option{-mstrict-align}.
3435 @item omit-leaf-frame-pointer
3436 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3437 Indicates that the frame pointer should be omitted for a leaf function call.
3438 To keep the frame pointer, the inverse attribute
3439 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3440 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3441 and @option{-mno-omit-leaf-frame-pointer}.
3444 @cindex @code{tls-dialect=} function attribute, AArch64
3445 Specifies the TLS dialect to use for this function. The behavior and
3446 permissible arguments are the same as for the command-line option
3447 @option{-mtls-dialect=}.
3450 @cindex @code{arch=} function attribute, AArch64
3451 Specifies the architecture version and architectural extensions to use
3452 for this function. The behavior and permissible arguments are the same as
3453 for the @option{-march=} command-line option.
3456 @cindex @code{tune=} function attribute, AArch64
3457 Specifies the core for which to tune the performance of this function.
3458 The behavior and permissible arguments are the same as for the @option{-mtune=}
3459 command-line option.
3462 @cindex @code{cpu=} function attribute, AArch64
3463 Specifies the core for which to tune the performance of this function and also
3464 whose architectural features to use. The behavior and valid arguments are the
3465 same as for the @option{-mcpu=} command-line option.
3469 The above target attributes can be specified as follows:
3472 __attribute__((target("@var{attr-string}")))
3480 where @code{@var{attr-string}} is one of the attribute strings specified above.
3482 Additionally, the architectural extension string may be specified on its
3483 own. This can be used to turn on and off particular architectural extensions
3484 without having to specify a particular architecture version or core. Example:
3487 __attribute__((target("+crc+nocrypto")))
3495 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3496 extension and disables the @code{crypto} extension for the function @code{foo}
3497 without modifying an existing @option{-march=} or @option{-mcpu} option.
3499 Multiple target function attributes can be specified by separating them with
3500 a comma. For example:
3502 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3510 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3511 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3513 @subsubsection Inlining rules
3514 Specifying target attributes on individual functions or performing link-time
3515 optimization across translation units compiled with different target options
3516 can affect function inlining rules:
3518 In particular, a caller function can inline a callee function only if the
3519 architectural features available to the callee are a subset of the features
3520 available to the caller.
3521 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3522 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3523 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3524 because the all the architectural features that function @code{bar} requires
3525 are available to function @code{foo}. Conversely, function @code{bar} cannot
3526 inline function @code{foo}.
3528 Additionally inlining a function compiled with @option{-mstrict-align} into a
3529 function compiled without @code{-mstrict-align} is not allowed.
3530 However, inlining a function compiled without @option{-mstrict-align} into a
3531 function compiled with @option{-mstrict-align} is allowed.
3533 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3534 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3535 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3536 architectural feature rules specified above.
3538 @node ARC Function Attributes
3539 @subsection ARC Function Attributes
3541 These function attributes are supported by the ARC back end:
3545 @cindex @code{interrupt} function attribute, ARC
3546 Use this attribute to indicate
3547 that the specified function is an interrupt handler. The compiler generates
3548 function entry and exit sequences suitable for use in an interrupt handler
3549 when this attribute is present.
3551 On the ARC, you must specify the kind of interrupt to be handled
3552 in a parameter to the interrupt attribute like this:
3555 void f () __attribute__ ((interrupt ("ilink1")));
3558 Permissible values for this parameter are: @w{@code{ilink1}} and
3564 @cindex @code{long_call} function attribute, ARC
3565 @cindex @code{medium_call} function attribute, ARC
3566 @cindex @code{short_call} function attribute, ARC
3567 @cindex indirect calls, ARC
3568 These attributes specify how a particular function is called.
3569 These attributes override the
3570 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3571 command-line switches and @code{#pragma long_calls} settings.
3573 For ARC, a function marked with the @code{long_call} attribute is
3574 always called using register-indirect jump-and-link instructions,
3575 thereby enabling the called function to be placed anywhere within the
3576 32-bit address space. A function marked with the @code{medium_call}
3577 attribute will always be close enough to be called with an unconditional
3578 branch-and-link instruction, which has a 25-bit offset from
3579 the call site. A function marked with the @code{short_call}
3580 attribute will always be close enough to be called with a conditional
3581 branch-and-link instruction, which has a 21-bit offset from
3585 @node ARM Function Attributes
3586 @subsection ARM Function Attributes
3588 These function attributes are supported for ARM targets:
3592 @cindex @code{interrupt} function attribute, ARM
3593 Use this attribute to indicate
3594 that the specified function is an interrupt handler. The compiler generates
3595 function entry and exit sequences suitable for use in an interrupt handler
3596 when this attribute is present.
3598 You can specify the kind of interrupt to be handled by
3599 adding an optional parameter to the interrupt attribute like this:
3602 void f () __attribute__ ((interrupt ("IRQ")));
3606 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3607 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3609 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3610 may be called with a word-aligned stack pointer.
3613 @cindex @code{isr} function attribute, ARM
3614 Use this attribute on ARM to write Interrupt Service Routines. This is an
3615 alias to the @code{interrupt} attribute above.
3619 @cindex @code{long_call} function attribute, ARM
3620 @cindex @code{short_call} function attribute, ARM
3621 @cindex indirect calls, ARM
3622 These attributes specify how a particular function is called.
3623 These attributes override the
3624 @option{-mlong-calls} (@pxref{ARM Options})
3625 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3626 @code{long_call} attribute indicates that the function might be far
3627 away from the call site and require a different (more expensive)
3628 calling sequence. The @code{short_call} attribute always places
3629 the offset to the function from the call site into the @samp{BL}
3630 instruction directly.
3633 @cindex @code{naked} function attribute, ARM
3634 This attribute allows the compiler to construct the
3635 requisite function declaration, while allowing the body of the
3636 function to be assembly code. The specified function will not have
3637 prologue/epilogue sequences generated by the compiler. Only basic
3638 @code{asm} statements can safely be included in naked functions
3639 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3640 basic @code{asm} and C code may appear to work, they cannot be
3641 depended upon to work reliably and are not supported.
3644 @cindex @code{pcs} function attribute, ARM
3646 The @code{pcs} attribute can be used to control the calling convention
3647 used for a function on ARM. The attribute takes an argument that specifies
3648 the calling convention to use.
3650 When compiling using the AAPCS ABI (or a variant of it) then valid
3651 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3652 order to use a variant other than @code{"aapcs"} then the compiler must
3653 be permitted to use the appropriate co-processor registers (i.e., the
3654 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3658 /* Argument passed in r0, and result returned in r0+r1. */
3659 double f2d (float) __attribute__((pcs("aapcs")));
3662 Variadic functions always use the @code{"aapcs"} calling convention and
3663 the compiler rejects attempts to specify an alternative.
3665 @item target (@var{options})
3666 @cindex @code{target} function attribute
3667 As discussed in @ref{Common Function Attributes}, this attribute
3668 allows specification of target-specific compilation options.
3670 On ARM, the following options are allowed:
3674 @cindex @code{target("thumb")} function attribute, ARM
3675 Force code generation in the Thumb (T16/T32) ISA, depending on the
3679 @cindex @code{target("arm")} function attribute, ARM
3680 Force code generation in the ARM (A32) ISA.
3682 Functions from different modes can be inlined in the caller's mode.
3685 @cindex @code{target("fpu=")} function attribute, ARM
3686 Specifies the fpu for which to tune the performance of this function.
3687 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3688 command-line option.
3694 @node AVR Function Attributes
3695 @subsection AVR Function Attributes
3697 These function attributes are supported by the AVR back end:
3701 @cindex @code{interrupt} function attribute, AVR
3702 Use this attribute to indicate
3703 that the specified function is an interrupt handler. The compiler generates
3704 function entry and exit sequences suitable for use in an interrupt handler
3705 when this attribute is present.
3707 On the AVR, the hardware globally disables interrupts when an
3708 interrupt is executed. The first instruction of an interrupt handler
3709 declared with this attribute is a @code{SEI} instruction to
3710 re-enable interrupts. See also the @code{signal} function attribute
3711 that does not insert a @code{SEI} instruction. If both @code{signal} and
3712 @code{interrupt} are specified for the same function, @code{signal}
3713 is silently ignored.
3716 @cindex @code{naked} function attribute, AVR
3717 This attribute allows the compiler to construct the
3718 requisite function declaration, while allowing the body of the
3719 function to be assembly code. The specified function will not have
3720 prologue/epilogue sequences generated by the compiler. Only basic
3721 @code{asm} statements can safely be included in naked functions
3722 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3723 basic @code{asm} and C code may appear to work, they cannot be
3724 depended upon to work reliably and are not supported.
3728 @cindex @code{OS_main} function attribute, AVR
3729 @cindex @code{OS_task} function attribute, AVR
3730 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3731 do not save/restore any call-saved register in their prologue/epilogue.
3733 The @code{OS_main} attribute can be used when there @emph{is
3734 guarantee} that interrupts are disabled at the time when the function
3735 is entered. This saves resources when the stack pointer has to be
3736 changed to set up a frame for local variables.
3738 The @code{OS_task} attribute can be used when there is @emph{no
3739 guarantee} that interrupts are disabled at that time when the function
3740 is entered like for, e@.g@. task functions in a multi-threading operating
3741 system. In that case, changing the stack pointer register is
3742 guarded by save/clear/restore of the global interrupt enable flag.
3744 The differences to the @code{naked} function attribute are:
3746 @item @code{naked} functions do not have a return instruction whereas
3747 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3748 @code{RETI} return instruction.
3749 @item @code{naked} functions do not set up a frame for local variables
3750 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3755 @cindex @code{signal} function attribute, AVR
3756 Use this attribute on the AVR to indicate that the specified
3757 function is an interrupt handler. The compiler generates function
3758 entry and exit sequences suitable for use in an interrupt handler when this
3759 attribute is present.
3761 See also the @code{interrupt} function attribute.
3763 The AVR hardware globally disables interrupts when an interrupt is executed.
3764 Interrupt handler functions defined with the @code{signal} attribute
3765 do not re-enable interrupts. It is save to enable interrupts in a
3766 @code{signal} handler. This ``save'' only applies to the code
3767 generated by the compiler and not to the IRQ layout of the
3768 application which is responsibility of the application.
3770 If both @code{signal} and @code{interrupt} are specified for the same
3771 function, @code{signal} is silently ignored.
3774 @node Blackfin Function Attributes
3775 @subsection Blackfin Function Attributes
3777 These function attributes are supported by the Blackfin back end:
3781 @item exception_handler
3782 @cindex @code{exception_handler} function attribute
3783 @cindex exception handler functions, Blackfin
3784 Use this attribute on the Blackfin to indicate that the specified function
3785 is an exception handler. The compiler generates function entry and
3786 exit sequences suitable for use in an exception handler when this
3787 attribute is present.
3789 @item interrupt_handler
3790 @cindex @code{interrupt_handler} function attribute, Blackfin
3791 Use this attribute to
3792 indicate that the specified function is an interrupt handler. The compiler
3793 generates function entry and exit sequences suitable for use in an
3794 interrupt handler when this attribute is present.
3797 @cindex @code{kspisusp} function attribute, Blackfin
3798 @cindex User stack pointer in interrupts on the Blackfin
3799 When used together with @code{interrupt_handler}, @code{exception_handler}
3800 or @code{nmi_handler}, code is generated to load the stack pointer
3801 from the USP register in the function prologue.
3804 @cindex @code{l1_text} function attribute, Blackfin
3805 This attribute specifies a function to be placed into L1 Instruction
3806 SRAM@. The function is put into a specific section named @code{.l1.text}.
3807 With @option{-mfdpic}, function calls with a such function as the callee
3808 or caller uses inlined PLT.
3811 @cindex @code{l2} function attribute, Blackfin
3812 This attribute specifies a function to be placed into L2
3813 SRAM. The function is put into a specific section named
3814 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3819 @cindex indirect calls, Blackfin
3820 @cindex @code{longcall} function attribute, Blackfin
3821 @cindex @code{shortcall} function attribute, Blackfin
3822 The @code{longcall} attribute
3823 indicates that the function might be far away from the call site and
3824 require a different (more expensive) calling sequence. The
3825 @code{shortcall} attribute indicates that the function is always close
3826 enough for the shorter calling sequence to be used. These attributes
3827 override the @option{-mlongcall} switch.
3830 @cindex @code{nesting} function attribute, Blackfin
3831 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3832 Use this attribute together with @code{interrupt_handler},
3833 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3834 entry code should enable nested interrupts or exceptions.
3837 @cindex @code{nmi_handler} function attribute, Blackfin
3838 @cindex NMI handler functions on the Blackfin processor
3839 Use this attribute on the Blackfin to indicate that the specified function
3840 is an NMI handler. The compiler generates function entry and
3841 exit sequences suitable for use in an NMI handler when this
3842 attribute is present.
3845 @cindex @code{saveall} function attribute, Blackfin
3846 @cindex save all registers on the Blackfin
3847 Use this attribute to indicate that
3848 all registers except the stack pointer should be saved in the prologue
3849 regardless of whether they are used or not.
3852 @node CR16 Function Attributes
3853 @subsection CR16 Function Attributes
3855 These function attributes are supported by the CR16 back end:
3859 @cindex @code{interrupt} function attribute, CR16
3860 Use this attribute to indicate
3861 that the specified function is an interrupt handler. The compiler generates
3862 function entry and exit sequences suitable for use in an interrupt handler
3863 when this attribute is present.
3866 @node Epiphany Function Attributes
3867 @subsection Epiphany Function Attributes
3869 These function attributes are supported by the Epiphany back end:
3873 @cindex @code{disinterrupt} function attribute, Epiphany
3874 This attribute causes the compiler to emit
3875 instructions to disable interrupts for the duration of the given
3878 @item forwarder_section
3879 @cindex @code{forwarder_section} function attribute, Epiphany
3880 This attribute modifies the behavior of an interrupt handler.
3881 The interrupt handler may be in external memory which cannot be
3882 reached by a branch instruction, so generate a local memory trampoline
3883 to transfer control. The single parameter identifies the section where
3884 the trampoline is placed.
3887 @cindex @code{interrupt} function attribute, Epiphany
3888 Use this attribute to indicate
3889 that the specified function is an interrupt handler. The compiler generates
3890 function entry and exit sequences suitable for use in an interrupt handler
3891 when this attribute is present. It may also generate
3892 a special section with code to initialize the interrupt vector table.
3894 On Epiphany targets one or more optional parameters can be added like this:
3897 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3900 Permissible values for these parameters are: @w{@code{reset}},
3901 @w{@code{software_exception}}, @w{@code{page_miss}},
3902 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3903 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3904 Multiple parameters indicate that multiple entries in the interrupt
3905 vector table should be initialized for this function, i.e.@: for each
3906 parameter @w{@var{name}}, a jump to the function is emitted in
3907 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3908 entirely, in which case no interrupt vector table entry is provided.
3910 Note that interrupts are enabled inside the function
3911 unless the @code{disinterrupt} attribute is also specified.
3913 The following examples are all valid uses of these attributes on
3916 void __attribute__ ((interrupt)) universal_handler ();
3917 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3918 void __attribute__ ((interrupt ("dma0, dma1")))
3919 universal_dma_handler ();
3920 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3921 fast_timer_handler ();
3922 void __attribute__ ((interrupt ("dma0, dma1"),
3923 forwarder_section ("tramp")))
3924 external_dma_handler ();
3929 @cindex @code{long_call} function attribute, Epiphany
3930 @cindex @code{short_call} function attribute, Epiphany
3931 @cindex indirect calls, Epiphany
3932 These attributes specify how a particular function is called.
3933 These attributes override the
3934 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3935 command-line switch and @code{#pragma long_calls} settings.
3939 @node H8/300 Function Attributes
3940 @subsection H8/300 Function Attributes
3942 These function attributes are available for H8/300 targets:
3945 @item function_vector
3946 @cindex @code{function_vector} function attribute, H8/300
3947 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3948 that the specified function should be called through the function vector.
3949 Calling a function through the function vector reduces code size; however,
3950 the function vector has a limited size (maximum 128 entries on the H8/300
3951 and 64 entries on the H8/300H and H8S)
3952 and shares space with the interrupt vector.
3954 @item interrupt_handler
3955 @cindex @code{interrupt_handler} function attribute, H8/300
3956 Use this attribute on the H8/300, H8/300H, and H8S to
3957 indicate that the specified function is an interrupt handler. The compiler
3958 generates function entry and exit sequences suitable for use in an
3959 interrupt handler when this attribute is present.
3962 @cindex @code{saveall} function attribute, H8/300
3963 @cindex save all registers on the H8/300, H8/300H, and H8S
3964 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3965 all registers except the stack pointer should be saved in the prologue
3966 regardless of whether they are used or not.
3969 @node IA-64 Function Attributes
3970 @subsection IA-64 Function Attributes
3972 These function attributes are supported on IA-64 targets:
3975 @item syscall_linkage
3976 @cindex @code{syscall_linkage} function attribute, IA-64
3977 This attribute is used to modify the IA-64 calling convention by marking
3978 all input registers as live at all function exits. This makes it possible
3979 to restart a system call after an interrupt without having to save/restore
3980 the input registers. This also prevents kernel data from leaking into
3984 @cindex @code{version_id} function attribute, IA-64
3985 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
3986 symbol to contain a version string, thus allowing for function level
3987 versioning. HP-UX system header files may use function level versioning
3988 for some system calls.
3991 extern int foo () __attribute__((version_id ("20040821")));
3995 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
3998 @node M32C Function Attributes
3999 @subsection M32C Function Attributes
4001 These function attributes are supported by the M32C back end:
4005 @cindex @code{bank_switch} function attribute, M32C
4006 When added to an interrupt handler with the M32C port, causes the
4007 prologue and epilogue to use bank switching to preserve the registers
4008 rather than saving them on the stack.
4010 @item fast_interrupt
4011 @cindex @code{fast_interrupt} function attribute, M32C
4012 Use this attribute on the M32C port to indicate that the specified
4013 function is a fast interrupt handler. This is just like the
4014 @code{interrupt} attribute, except that @code{freit} is used to return
4015 instead of @code{reit}.
4017 @item function_vector
4018 @cindex @code{function_vector} function attribute, M16C/M32C
4019 On M16C/M32C targets, the @code{function_vector} attribute declares a
4020 special page subroutine call function. Use of this attribute reduces
4021 the code size by 2 bytes for each call generated to the
4022 subroutine. The argument to the attribute is the vector number entry
4023 from the special page vector table which contains the 16 low-order
4024 bits of the subroutine's entry address. Each vector table has special
4025 page number (18 to 255) that is used in @code{jsrs} instructions.
4026 Jump addresses of the routines are generated by adding 0x0F0000 (in
4027 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4028 2-byte addresses set in the vector table. Therefore you need to ensure
4029 that all the special page vector routines should get mapped within the
4030 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4033 In the following example 2 bytes are saved for each call to
4034 function @code{foo}.
4037 void foo (void) __attribute__((function_vector(0x18)));
4048 If functions are defined in one file and are called in another file,
4049 then be sure to write this declaration in both files.
4051 This attribute is ignored for R8C target.
4054 @cindex @code{interrupt} function attribute, M32C
4055 Use this attribute to indicate
4056 that the specified function is an interrupt handler. The compiler generates
4057 function entry and exit sequences suitable for use in an interrupt handler
4058 when this attribute is present.
4061 @node M32R/D Function Attributes
4062 @subsection M32R/D Function Attributes
4064 These function attributes are supported by the M32R/D back end:
4068 @cindex @code{interrupt} function attribute, M32R/D
4069 Use this attribute to indicate
4070 that the specified function is an interrupt handler. The compiler generates
4071 function entry and exit sequences suitable for use in an interrupt handler
4072 when this attribute is present.
4074 @item model (@var{model-name})
4075 @cindex @code{model} function attribute, M32R/D
4076 @cindex function addressability on the M32R/D
4078 On the M32R/D, use this attribute to set the addressability of an
4079 object, and of the code generated for a function. The identifier
4080 @var{model-name} is one of @code{small}, @code{medium}, or
4081 @code{large}, representing each of the code models.
4083 Small model objects live in the lower 16MB of memory (so that their
4084 addresses can be loaded with the @code{ld24} instruction), and are
4085 callable with the @code{bl} instruction.
4087 Medium model objects may live anywhere in the 32-bit address space (the
4088 compiler generates @code{seth/add3} instructions to load their addresses),
4089 and are callable with the @code{bl} instruction.
4091 Large model objects may live anywhere in the 32-bit address space (the
4092 compiler generates @code{seth/add3} instructions to load their addresses),
4093 and may not be reachable with the @code{bl} instruction (the compiler
4094 generates the much slower @code{seth/add3/jl} instruction sequence).
4097 @node m68k Function Attributes
4098 @subsection m68k Function Attributes
4100 These function attributes are supported by the m68k back end:
4104 @itemx interrupt_handler
4105 @cindex @code{interrupt} function attribute, m68k
4106 @cindex @code{interrupt_handler} function attribute, m68k
4107 Use this attribute to
4108 indicate that the specified function is an interrupt handler. The compiler
4109 generates function entry and exit sequences suitable for use in an
4110 interrupt handler when this attribute is present. Either name may be used.
4112 @item interrupt_thread
4113 @cindex @code{interrupt_thread} function attribute, fido
4114 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4115 that the specified function is an interrupt handler that is designed
4116 to run as a thread. The compiler omits generate prologue/epilogue
4117 sequences and replaces the return instruction with a @code{sleep}
4118 instruction. This attribute is available only on fido.
4121 @node MCORE Function Attributes
4122 @subsection MCORE Function Attributes
4124 These function attributes are supported by the MCORE back end:
4128 @cindex @code{naked} function attribute, MCORE
4129 This attribute allows the compiler to construct the
4130 requisite function declaration, while allowing the body of the
4131 function to be assembly code. The specified function will not have
4132 prologue/epilogue sequences generated by the compiler. Only basic
4133 @code{asm} statements can safely be included in naked functions
4134 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4135 basic @code{asm} and C code may appear to work, they cannot be
4136 depended upon to work reliably and are not supported.
4139 @node MeP Function Attributes
4140 @subsection MeP Function Attributes
4142 These function attributes are supported by the MeP back end:
4146 @cindex @code{disinterrupt} function attribute, MeP
4147 On MeP targets, this attribute causes the compiler to emit
4148 instructions to disable interrupts for the duration of the given
4152 @cindex @code{interrupt} function attribute, MeP
4153 Use this attribute to indicate
4154 that the specified function is an interrupt handler. The compiler generates
4155 function entry and exit sequences suitable for use in an interrupt handler
4156 when this attribute is present.
4159 @cindex @code{near} function attribute, MeP
4160 This attribute causes the compiler to assume the called
4161 function is close enough to use the normal calling convention,
4162 overriding the @option{-mtf} command-line option.
4165 @cindex @code{far} function attribute, MeP
4166 On MeP targets this causes the compiler to use a calling convention
4167 that assumes the called function is too far away for the built-in
4171 @cindex @code{vliw} function attribute, MeP
4172 The @code{vliw} attribute tells the compiler to emit
4173 instructions in VLIW mode instead of core mode. Note that this
4174 attribute is not allowed unless a VLIW coprocessor has been configured
4175 and enabled through command-line options.
4178 @node MicroBlaze Function Attributes
4179 @subsection MicroBlaze Function Attributes
4181 These function attributes are supported on MicroBlaze targets:
4184 @item save_volatiles
4185 @cindex @code{save_volatiles} function attribute, MicroBlaze
4186 Use this attribute to indicate that the function is
4187 an interrupt handler. All volatile registers (in addition to non-volatile
4188 registers) are saved in the function prologue. If the function is a leaf
4189 function, only volatiles used by the function are saved. A normal function
4190 return is generated instead of a return from interrupt.
4193 @cindex @code{break_handler} function attribute, MicroBlaze
4194 @cindex break handler functions
4195 Use this attribute to indicate that
4196 the specified function is a break handler. The compiler generates function
4197 entry and exit sequences suitable for use in an break handler when this
4198 attribute is present. The return from @code{break_handler} is done through
4199 the @code{rtbd} instead of @code{rtsd}.
4202 void f () __attribute__ ((break_handler));
4205 @item interrupt_handler
4206 @itemx fast_interrupt
4207 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4208 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4209 These attributes indicate that the specified function is an interrupt
4210 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4211 used in low-latency interrupt mode, and @code{interrupt_handler} for
4212 interrupts that do not use low-latency handlers. In both cases, GCC
4213 emits appropriate prologue code and generates a return from the handler
4214 using @code{rtid} instead of @code{rtsd}.
4217 @node Microsoft Windows Function Attributes
4218 @subsection Microsoft Windows Function Attributes
4220 The following attributes are available on Microsoft Windows and Symbian OS
4225 @cindex @code{dllexport} function attribute
4226 @cindex @code{__declspec(dllexport)}
4227 On Microsoft Windows targets and Symbian OS targets the
4228 @code{dllexport} attribute causes the compiler to provide a global
4229 pointer to a pointer in a DLL, so that it can be referenced with the
4230 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4231 name is formed by combining @code{_imp__} and the function or variable
4234 You can use @code{__declspec(dllexport)} as a synonym for
4235 @code{__attribute__ ((dllexport))} for compatibility with other
4238 On systems that support the @code{visibility} attribute, this
4239 attribute also implies ``default'' visibility. It is an error to
4240 explicitly specify any other visibility.
4242 GCC's default behavior is to emit all inline functions with the
4243 @code{dllexport} attribute. Since this can cause object file-size bloat,
4244 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4245 ignore the attribute for inlined functions unless the
4246 @option{-fkeep-inline-functions} flag is used instead.
4248 The attribute is ignored for undefined symbols.
4250 When applied to C++ classes, the attribute marks defined non-inlined
4251 member functions and static data members as exports. Static consts
4252 initialized in-class are not marked unless they are also defined
4255 For Microsoft Windows targets there are alternative methods for
4256 including the symbol in the DLL's export table such as using a
4257 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4258 the @option{--export-all} linker flag.
4261 @cindex @code{dllimport} function attribute
4262 @cindex @code{__declspec(dllimport)}
4263 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4264 attribute causes the compiler to reference a function or variable via
4265 a global pointer to a pointer that is set up by the DLL exporting the
4266 symbol. The attribute implies @code{extern}. On Microsoft Windows
4267 targets, the pointer name is formed by combining @code{_imp__} and the
4268 function or variable name.
4270 You can use @code{__declspec(dllimport)} as a synonym for
4271 @code{__attribute__ ((dllimport))} for compatibility with other
4274 On systems that support the @code{visibility} attribute, this
4275 attribute also implies ``default'' visibility. It is an error to
4276 explicitly specify any other visibility.
4278 Currently, the attribute is ignored for inlined functions. If the
4279 attribute is applied to a symbol @emph{definition}, an error is reported.
4280 If a symbol previously declared @code{dllimport} is later defined, the
4281 attribute is ignored in subsequent references, and a warning is emitted.
4282 The attribute is also overridden by a subsequent declaration as
4285 When applied to C++ classes, the attribute marks non-inlined
4286 member functions and static data members as imports. However, the
4287 attribute is ignored for virtual methods to allow creation of vtables
4290 On the SH Symbian OS target the @code{dllimport} attribute also has
4291 another affect---it can cause the vtable and run-time type information
4292 for a class to be exported. This happens when the class has a
4293 dllimported constructor or a non-inline, non-pure virtual function
4294 and, for either of those two conditions, the class also has an inline
4295 constructor or destructor and has a key function that is defined in
4296 the current translation unit.
4298 For Microsoft Windows targets the use of the @code{dllimport}
4299 attribute on functions is not necessary, but provides a small
4300 performance benefit by eliminating a thunk in the DLL@. The use of the
4301 @code{dllimport} attribute on imported variables can be avoided by passing the
4302 @option{--enable-auto-import} switch to the GNU linker. As with
4303 functions, using the attribute for a variable eliminates a thunk in
4306 One drawback to using this attribute is that a pointer to a
4307 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4308 address. However, a pointer to a @emph{function} with the
4309 @code{dllimport} attribute can be used as a constant initializer; in
4310 this case, the address of a stub function in the import lib is
4311 referenced. On Microsoft Windows targets, the attribute can be disabled
4312 for functions by setting the @option{-mnop-fun-dllimport} flag.
4315 @node MIPS Function Attributes
4316 @subsection MIPS Function Attributes
4318 These function attributes are supported by the MIPS back end:
4322 @cindex @code{interrupt} function attribute, MIPS
4323 Use this attribute to indicate that the specified function is an interrupt
4324 handler. The compiler generates function entry and exit sequences suitable
4325 for use in an interrupt handler when this attribute is present.
4326 An optional argument is supported for the interrupt attribute which allows
4327 the interrupt mode to be described. By default GCC assumes the external
4328 interrupt controller (EIC) mode is in use, this can be explicitly set using
4329 @code{eic}. When interrupts are non-masked then the requested Interrupt
4330 Priority Level (IPL) is copied to the current IPL which has the effect of only
4331 enabling higher priority interrupts. To use vectored interrupt mode use
4332 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4333 the behavior of the non-masked interrupt support and GCC will arrange to mask
4334 all interrupts from sw0 up to and including the specified interrupt vector.
4336 You can use the following attributes to modify the behavior
4337 of an interrupt handler:
4339 @item use_shadow_register_set
4340 @cindex @code{use_shadow_register_set} function attribute, MIPS
4341 Assume that the handler uses a shadow register set, instead of
4342 the main general-purpose registers. An optional argument @code{intstack} is
4343 supported to indicate that the shadow register set contains a valid stack
4346 @item keep_interrupts_masked
4347 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4348 Keep interrupts masked for the whole function. Without this attribute,
4349 GCC tries to reenable interrupts for as much of the function as it can.
4351 @item use_debug_exception_return
4352 @cindex @code{use_debug_exception_return} function attribute, MIPS
4353 Return using the @code{deret} instruction. Interrupt handlers that don't
4354 have this attribute return using @code{eret} instead.
4357 You can use any combination of these attributes, as shown below:
4359 void __attribute__ ((interrupt)) v0 ();
4360 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4361 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4362 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4363 void __attribute__ ((interrupt, use_shadow_register_set,
4364 keep_interrupts_masked)) v4 ();
4365 void __attribute__ ((interrupt, use_shadow_register_set,
4366 use_debug_exception_return)) v5 ();
4367 void __attribute__ ((interrupt, keep_interrupts_masked,
4368 use_debug_exception_return)) v6 ();
4369 void __attribute__ ((interrupt, use_shadow_register_set,
4370 keep_interrupts_masked,
4371 use_debug_exception_return)) v7 ();
4372 void __attribute__ ((interrupt("eic"))) v8 ();
4373 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4379 @cindex indirect calls, MIPS
4380 @cindex @code{long_call} function attribute, MIPS
4381 @cindex @code{near} function attribute, MIPS
4382 @cindex @code{far} function attribute, MIPS
4383 These attributes specify how a particular function is called on MIPS@.
4384 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4385 command-line switch. The @code{long_call} and @code{far} attributes are
4386 synonyms, and cause the compiler to always call
4387 the function by first loading its address into a register, and then using
4388 the contents of that register. The @code{near} attribute has the opposite
4389 effect; it specifies that non-PIC calls should be made using the more
4390 efficient @code{jal} instruction.
4394 @cindex @code{mips16} function attribute, MIPS
4395 @cindex @code{nomips16} function attribute, MIPS
4397 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4398 function attributes to locally select or turn off MIPS16 code generation.
4399 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4400 while MIPS16 code generation is disabled for functions with the
4401 @code{nomips16} attribute. These attributes override the
4402 @option{-mips16} and @option{-mno-mips16} options on the command line
4403 (@pxref{MIPS Options}).
4405 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4406 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4407 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4408 may interact badly with some GCC extensions such as @code{__builtin_apply}
4409 (@pxref{Constructing Calls}).
4411 @item micromips, MIPS
4412 @itemx nomicromips, MIPS
4413 @cindex @code{micromips} function attribute
4414 @cindex @code{nomicromips} function attribute
4416 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4417 function attributes to locally select or turn off microMIPS code generation.
4418 A function with the @code{micromips} attribute is emitted as microMIPS code,
4419 while microMIPS code generation is disabled for functions with the
4420 @code{nomicromips} attribute. These attributes override the
4421 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4422 (@pxref{MIPS Options}).
4424 When compiling files containing mixed microMIPS and non-microMIPS code, the
4425 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4427 not that within individual functions. Mixed microMIPS and non-microMIPS code
4428 may interact badly with some GCC extensions such as @code{__builtin_apply}
4429 (@pxref{Constructing Calls}).
4432 @cindex @code{nocompression} function attribute, MIPS
4433 On MIPS targets, you can use the @code{nocompression} function attribute
4434 to locally turn off MIPS16 and microMIPS code generation. This attribute
4435 overrides the @option{-mips16} and @option{-mmicromips} options on the
4436 command line (@pxref{MIPS Options}).
4439 @node MSP430 Function Attributes
4440 @subsection MSP430 Function Attributes
4442 These function attributes are supported by the MSP430 back end:
4446 @cindex @code{critical} function attribute, MSP430
4447 Critical functions disable interrupts upon entry and restore the
4448 previous interrupt state upon exit. Critical functions cannot also
4449 have the @code{naked} or @code{reentrant} attributes. They can have
4450 the @code{interrupt} attribute.
4453 @cindex @code{interrupt} function attribute, MSP430
4454 Use this attribute to indicate
4455 that the specified function is an interrupt handler. The compiler generates
4456 function entry and exit sequences suitable for use in an interrupt handler
4457 when this attribute is present.
4459 You can provide an argument to the interrupt
4460 attribute which specifies a name or number. If the argument is a
4461 number it indicates the slot in the interrupt vector table (0 - 31) to
4462 which this handler should be assigned. If the argument is a name it
4463 is treated as a symbolic name for the vector slot. These names should
4464 match up with appropriate entries in the linker script. By default
4465 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4466 @code{reset} for vector 31 are recognized.
4469 @cindex @code{naked} function attribute, MSP430
4470 This attribute allows the compiler to construct the
4471 requisite function declaration, while allowing the body of the
4472 function to be assembly code. The specified function will not have
4473 prologue/epilogue sequences generated by the compiler. Only basic
4474 @code{asm} statements can safely be included in naked functions
4475 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4476 basic @code{asm} and C code may appear to work, they cannot be
4477 depended upon to work reliably and are not supported.
4480 @cindex @code{reentrant} function attribute, MSP430
4481 Reentrant functions disable interrupts upon entry and enable them
4482 upon exit. Reentrant functions cannot also have the @code{naked}
4483 or @code{critical} attributes. They can have the @code{interrupt}
4487 @cindex @code{wakeup} function attribute, MSP430
4488 This attribute only applies to interrupt functions. It is silently
4489 ignored if applied to a non-interrupt function. A wakeup interrupt
4490 function will rouse the processor from any low-power state that it
4491 might be in when the function exits.
4496 @cindex @code{lower} function attribute, MSP430
4497 @cindex @code{upper} function attribute, MSP430
4498 @cindex @code{either} function attribute, MSP430
4499 On the MSP430 target these attributes can be used to specify whether
4500 the function or variable should be placed into low memory, high
4501 memory, or the placement should be left to the linker to decide. The
4502 attributes are only significant if compiling for the MSP430X
4505 The attributes work in conjunction with a linker script that has been
4506 augmented to specify where to place sections with a @code{.lower} and
4507 a @code{.upper} prefix. So, for example, as well as placing the
4508 @code{.data} section, the script also specifies the placement of a
4509 @code{.lower.data} and a @code{.upper.data} section. The intention
4510 is that @code{lower} sections are placed into a small but easier to
4511 access memory region and the upper sections are placed into a larger, but
4512 slower to access, region.
4514 The @code{either} attribute is special. It tells the linker to place
4515 the object into the corresponding @code{lower} section if there is
4516 room for it. If there is insufficient room then the object is placed
4517 into the corresponding @code{upper} section instead. Note that the
4518 placement algorithm is not very sophisticated. It does not attempt to
4519 find an optimal packing of the @code{lower} sections. It just makes
4520 one pass over the objects and does the best that it can. Using the
4521 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4522 options can help the packing, however, since they produce smaller,
4523 easier to pack regions.
4526 @node NDS32 Function Attributes
4527 @subsection NDS32 Function Attributes
4529 These function attributes are supported by the NDS32 back end:
4533 @cindex @code{exception} function attribute
4534 @cindex exception handler functions, NDS32
4535 Use this attribute on the NDS32 target to indicate that the specified function
4536 is an exception handler. The compiler will generate corresponding sections
4537 for use in an exception handler.
4540 @cindex @code{interrupt} function attribute, NDS32
4541 On NDS32 target, this attribute indicates that the specified function
4542 is an interrupt handler. The compiler generates corresponding sections
4543 for use in an interrupt handler. You can use the following attributes
4544 to modify the behavior:
4547 @cindex @code{nested} function attribute, NDS32
4548 This interrupt service routine is interruptible.
4550 @cindex @code{not_nested} function attribute, NDS32
4551 This interrupt service routine is not interruptible.
4553 @cindex @code{nested_ready} function attribute, NDS32
4554 This interrupt service routine is interruptible after @code{PSW.GIE}
4555 (global interrupt enable) is set. This allows interrupt service routine to
4556 finish some short critical code before enabling interrupts.
4558 @cindex @code{save_all} function attribute, NDS32
4559 The system will help save all registers into stack before entering
4562 @cindex @code{partial_save} function attribute, NDS32
4563 The system will help save caller registers into stack before entering
4568 @cindex @code{naked} function attribute, NDS32
4569 This attribute allows the compiler to construct the
4570 requisite function declaration, while allowing the body of the
4571 function to be assembly code. The specified function will not have
4572 prologue/epilogue sequences generated by the compiler. Only basic
4573 @code{asm} statements can safely be included in naked functions
4574 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4575 basic @code{asm} and C code may appear to work, they cannot be
4576 depended upon to work reliably and are not supported.
4579 @cindex @code{reset} function attribute, NDS32
4580 @cindex reset handler functions
4581 Use this attribute on the NDS32 target to indicate that the specified function
4582 is a reset handler. The compiler will generate corresponding sections
4583 for use in a reset handler. You can use the following attributes
4584 to provide extra exception handling:
4587 @cindex @code{nmi} function attribute, NDS32
4588 Provide a user-defined function to handle NMI exception.
4590 @cindex @code{warm} function attribute, NDS32
4591 Provide a user-defined function to handle warm reset exception.
4595 @node Nios II Function Attributes
4596 @subsection Nios II Function Attributes
4598 These function attributes are supported by the Nios II back end:
4601 @item target (@var{options})
4602 @cindex @code{target} function attribute
4603 As discussed in @ref{Common Function Attributes}, this attribute
4604 allows specification of target-specific compilation options.
4606 When compiling for Nios II, the following options are allowed:
4609 @item custom-@var{insn}=@var{N}
4610 @itemx no-custom-@var{insn}
4611 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4612 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4613 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4614 custom instruction with encoding @var{N} when generating code that uses
4615 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4616 the custom instruction @var{insn}.
4617 These target attributes correspond to the
4618 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4619 command-line options, and support the same set of @var{insn} keywords.
4620 @xref{Nios II Options}, for more information.
4622 @item custom-fpu-cfg=@var{name}
4623 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4624 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4625 command-line option, to select a predefined set of custom instructions
4627 @xref{Nios II Options}, for more information.
4631 @node Nvidia PTX Function Attributes
4632 @subsection Nvidia PTX Function Attributes
4634 These function attributes are supported by the Nvidia PTX back end:
4638 @cindex @code{kernel} attribute, Nvidia PTX
4639 This attribute indicates that the corresponding function should be compiled
4640 as a kernel function, which can be invoked from the host via the CUDA RT
4642 By default functions are only callable only from other PTX functions.
4644 Kernel functions must have @code{void} return type.
4647 @node PowerPC Function Attributes
4648 @subsection PowerPC Function Attributes
4650 These function attributes are supported by the PowerPC back end:
4655 @cindex indirect calls, PowerPC
4656 @cindex @code{longcall} function attribute, PowerPC
4657 @cindex @code{shortcall} function attribute, PowerPC
4658 The @code{longcall} attribute
4659 indicates that the function might be far away from the call site and
4660 require a different (more expensive) calling sequence. The
4661 @code{shortcall} attribute indicates that the function is always close
4662 enough for the shorter calling sequence to be used. These attributes
4663 override both the @option{-mlongcall} switch and
4664 the @code{#pragma longcall} setting.
4666 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4667 calls are necessary.
4669 @item target (@var{options})
4670 @cindex @code{target} function attribute
4671 As discussed in @ref{Common Function Attributes}, this attribute
4672 allows specification of target-specific compilation options.
4674 On the PowerPC, the following options are allowed:
4679 @cindex @code{target("altivec")} function attribute, PowerPC
4680 Generate code that uses (does not use) AltiVec instructions. In
4681 32-bit code, you cannot enable AltiVec instructions unless
4682 @option{-mabi=altivec} is used on the command line.
4686 @cindex @code{target("cmpb")} function attribute, PowerPC
4687 Generate code that uses (does not use) the compare bytes instruction
4688 implemented on the POWER6 processor and other processors that support
4689 the PowerPC V2.05 architecture.
4693 @cindex @code{target("dlmzb")} function attribute, PowerPC
4694 Generate code that uses (does not use) the string-search @samp{dlmzb}
4695 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4696 generated by default when targeting those processors.
4700 @cindex @code{target("fprnd")} function attribute, PowerPC
4701 Generate code that uses (does not use) the FP round to integer
4702 instructions implemented on the POWER5+ processor and other processors
4703 that support the PowerPC V2.03 architecture.
4707 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4708 Generate code that uses (does not use) the decimal floating-point
4709 instructions implemented on some POWER processors.
4713 @cindex @code{target("isel")} function attribute, PowerPC
4714 Generate code that uses (does not use) ISEL instruction.
4718 @cindex @code{target("mfcrf")} function attribute, PowerPC
4719 Generate code that uses (does not use) the move from condition
4720 register field instruction implemented on the POWER4 processor and
4721 other processors that support the PowerPC V2.01 architecture.
4725 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4726 Generate code that uses (does not use) the FP move to/from general
4727 purpose register instructions implemented on the POWER6X processor and
4728 other processors that support the extended PowerPC V2.05 architecture.
4732 @cindex @code{target("mulhw")} function attribute, PowerPC
4733 Generate code that uses (does not use) the half-word multiply and
4734 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4735 These instructions are generated by default when targeting those
4740 @cindex @code{target("multiple")} function attribute, PowerPC
4741 Generate code that uses (does not use) the load multiple word
4742 instructions and the store multiple word instructions.
4746 @cindex @code{target("update")} function attribute, PowerPC
4747 Generate code that uses (does not use) the load or store instructions
4748 that update the base register to the address of the calculated memory
4753 @cindex @code{target("popcntb")} function attribute, PowerPC
4754 Generate code that uses (does not use) the popcount and double-precision
4755 FP reciprocal estimate instruction implemented on the POWER5
4756 processor and other processors that support the PowerPC V2.02
4761 @cindex @code{target("popcntd")} function attribute, PowerPC
4762 Generate code that uses (does not use) the popcount instruction
4763 implemented on the POWER7 processor and other processors that support
4764 the PowerPC V2.06 architecture.
4766 @item powerpc-gfxopt
4767 @itemx no-powerpc-gfxopt
4768 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4769 Generate code that uses (does not use) the optional PowerPC
4770 architecture instructions in the Graphics group, including
4771 floating-point select.
4774 @itemx no-powerpc-gpopt
4775 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4776 Generate code that uses (does not use) the optional PowerPC
4777 architecture instructions in the General Purpose group, including
4778 floating-point square root.
4780 @item recip-precision
4781 @itemx no-recip-precision
4782 @cindex @code{target("recip-precision")} function attribute, PowerPC
4783 Assume (do not assume) that the reciprocal estimate instructions
4784 provide higher-precision estimates than is mandated by the PowerPC
4789 @cindex @code{target("string")} function attribute, PowerPC
4790 Generate code that uses (does not use) the load string instructions
4791 and the store string word instructions to save multiple registers and
4792 do small block moves.
4796 @cindex @code{target("vsx")} function attribute, PowerPC
4797 Generate code that uses (does not use) vector/scalar (VSX)
4798 instructions, and also enable the use of built-in functions that allow
4799 more direct access to the VSX instruction set. In 32-bit code, you
4800 cannot enable VSX or AltiVec instructions unless
4801 @option{-mabi=altivec} is used on the command line.
4805 @cindex @code{target("friz")} function attribute, PowerPC
4806 Generate (do not generate) the @code{friz} instruction when the
4807 @option{-funsafe-math-optimizations} option is used to optimize
4808 rounding a floating-point value to 64-bit integer and back to floating
4809 point. The @code{friz} instruction does not return the same value if
4810 the floating-point number is too large to fit in an integer.
4812 @item avoid-indexed-addresses
4813 @itemx no-avoid-indexed-addresses
4814 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4815 Generate code that tries to avoid (not avoid) the use of indexed load
4816 or store instructions.
4820 @cindex @code{target("paired")} function attribute, PowerPC
4821 Generate code that uses (does not use) the generation of PAIRED simd
4826 @cindex @code{target("longcall")} function attribute, PowerPC
4827 Generate code that assumes (does not assume) that all calls are far
4828 away so that a longer more expensive calling sequence is required.
4831 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4832 Specify the architecture to generate code for when compiling the
4833 function. If you select the @code{target("cpu=power7")} attribute when
4834 generating 32-bit code, VSX and AltiVec instructions are not generated
4835 unless you use the @option{-mabi=altivec} option on the command line.
4837 @item tune=@var{TUNE}
4838 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4839 Specify the architecture to tune for when compiling the function. If
4840 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4841 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4842 compilation tunes for the @var{CPU} architecture, and not the
4843 default tuning specified on the command line.
4846 On the PowerPC, the inliner does not inline a
4847 function that has different target options than the caller, unless the
4848 callee has a subset of the target options of the caller.
4851 @node RL78 Function Attributes
4852 @subsection RL78 Function Attributes
4854 These function attributes are supported by the RL78 back end:
4858 @itemx brk_interrupt
4859 @cindex @code{interrupt} function attribute, RL78
4860 @cindex @code{brk_interrupt} function attribute, RL78
4861 These attributes indicate
4862 that the specified function is an interrupt handler. The compiler generates
4863 function entry and exit sequences suitable for use in an interrupt handler
4864 when this attribute is present.
4866 Use @code{brk_interrupt} instead of @code{interrupt} for
4867 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4868 that must end with @code{RETB} instead of @code{RETI}).
4871 @cindex @code{naked} function attribute, RL78
4872 This attribute allows the compiler to construct the
4873 requisite function declaration, while allowing the body of the
4874 function to be assembly code. The specified function will not have
4875 prologue/epilogue sequences generated by the compiler. Only basic
4876 @code{asm} statements can safely be included in naked functions
4877 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4878 basic @code{asm} and C code may appear to work, they cannot be
4879 depended upon to work reliably and are not supported.
4882 @node RX Function Attributes
4883 @subsection RX Function Attributes
4885 These function attributes are supported by the RX back end:
4888 @item fast_interrupt
4889 @cindex @code{fast_interrupt} function attribute, RX
4890 Use this attribute on the RX port to indicate that the specified
4891 function is a fast interrupt handler. This is just like the
4892 @code{interrupt} attribute, except that @code{freit} is used to return
4893 instead of @code{reit}.
4896 @cindex @code{interrupt} function attribute, RX
4897 Use this attribute to indicate
4898 that the specified function is an interrupt handler. The compiler generates
4899 function entry and exit sequences suitable for use in an interrupt handler
4900 when this attribute is present.
4902 On RX targets, you may specify one or more vector numbers as arguments
4903 to the attribute, as well as naming an alternate table name.
4904 Parameters are handled sequentially, so one handler can be assigned to
4905 multiple entries in multiple tables. One may also pass the magic
4906 string @code{"$default"} which causes the function to be used for any
4907 unfilled slots in the current table.
4909 This example shows a simple assignment of a function to one vector in
4910 the default table (note that preprocessor macros may be used for
4911 chip-specific symbolic vector names):
4913 void __attribute__ ((interrupt (5))) txd1_handler ();
4916 This example assigns a function to two slots in the default table
4917 (using preprocessor macros defined elsewhere) and makes it the default
4918 for the @code{dct} table:
4920 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4925 @cindex @code{naked} function attribute, RX
4926 This attribute allows the compiler to construct the
4927 requisite function declaration, while allowing the body of the
4928 function to be assembly code. The specified function will not have
4929 prologue/epilogue sequences generated by the compiler. Only basic
4930 @code{asm} statements can safely be included in naked functions
4931 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4932 basic @code{asm} and C code may appear to work, they cannot be
4933 depended upon to work reliably and are not supported.
4936 @cindex @code{vector} function attribute, RX
4937 This RX attribute is similar to the @code{interrupt} attribute, including its
4938 parameters, but does not make the function an interrupt-handler type
4939 function (i.e. it retains the normal C function calling ABI). See the
4940 @code{interrupt} attribute for a description of its arguments.
4943 @node S/390 Function Attributes
4944 @subsection S/390 Function Attributes
4946 These function attributes are supported on the S/390:
4949 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4950 @cindex @code{hotpatch} function attribute, S/390
4952 On S/390 System z targets, you can use this function attribute to
4953 make GCC generate a ``hot-patching'' function prologue. If the
4954 @option{-mhotpatch=} command-line option is used at the same time,
4955 the @code{hotpatch} attribute takes precedence. The first of the
4956 two arguments specifies the number of halfwords to be added before
4957 the function label. A second argument can be used to specify the
4958 number of halfwords to be added after the function label. For
4959 both arguments the maximum allowed value is 1000000.
4961 If both arguments are zero, hotpatching is disabled.
4963 @item target (@var{options})
4964 @cindex @code{target} function attribute
4965 As discussed in @ref{Common Function Attributes}, this attribute
4966 allows specification of target-specific compilation options.
4968 On S/390, the following options are supported:
4976 @item warn-framesize=
4988 @itemx no-packed-stack
4990 @itemx no-small-exec
4993 @item warn-dynamicstack
4994 @itemx no-warn-dynamicstack
4997 The options work exactly like the S/390 specific command line
4998 options (without the prefix @option{-m}) except that they do not
4999 change any feature macros. For example,
5002 @code{target("no-vx")}
5005 does not undefine the @code{__VEC__} macro.
5008 @node SH Function Attributes
5009 @subsection SH Function Attributes
5011 These function attributes are supported on the SH family of processors:
5014 @item function_vector
5015 @cindex @code{function_vector} function attribute, SH
5016 @cindex calling functions through the function vector on SH2A
5017 On SH2A targets, this attribute declares a function to be called using the
5018 TBR relative addressing mode. The argument to this attribute is the entry
5019 number of the same function in a vector table containing all the TBR
5020 relative addressable functions. For correct operation the TBR must be setup
5021 accordingly to point to the start of the vector table before any functions with
5022 this attribute are invoked. Usually a good place to do the initialization is
5023 the startup routine. The TBR relative vector table can have at max 256 function
5024 entries. The jumps to these functions are generated using a SH2A specific,
5025 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5026 from GNU binutils version 2.7 or later for this attribute to work correctly.
5028 In an application, for a function being called once, this attribute
5029 saves at least 8 bytes of code; and if other successive calls are being
5030 made to the same function, it saves 2 bytes of code per each of these
5033 @item interrupt_handler
5034 @cindex @code{interrupt_handler} function attribute, SH
5035 Use this attribute to
5036 indicate that the specified function is an interrupt handler. The compiler
5037 generates function entry and exit sequences suitable for use in an
5038 interrupt handler when this attribute is present.
5040 @item nosave_low_regs
5041 @cindex @code{nosave_low_regs} function attribute, SH
5042 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5043 function should not save and restore registers R0..R7. This can be used on SH3*
5044 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5048 @cindex @code{renesas} function attribute, SH
5049 On SH targets this attribute specifies that the function or struct follows the
5053 @cindex @code{resbank} function attribute, SH
5054 On the SH2A target, this attribute enables the high-speed register
5055 saving and restoration using a register bank for @code{interrupt_handler}
5056 routines. Saving to the bank is performed automatically after the CPU
5057 accepts an interrupt that uses a register bank.
5059 The nineteen 32-bit registers comprising general register R0 to R14,
5060 control register GBR, and system registers MACH, MACL, and PR and the
5061 vector table address offset are saved into a register bank. Register
5062 banks are stacked in first-in last-out (FILO) sequence. Restoration
5063 from the bank is executed by issuing a RESBANK instruction.
5066 @cindex @code{sp_switch} function attribute, SH
5067 Use this attribute on the SH to indicate an @code{interrupt_handler}
5068 function should switch to an alternate stack. It expects a string
5069 argument that names a global variable holding the address of the
5074 void f () __attribute__ ((interrupt_handler,
5075 sp_switch ("alt_stack")));
5079 @cindex @code{trap_exit} function attribute, SH
5080 Use this attribute on the SH for an @code{interrupt_handler} to return using
5081 @code{trapa} instead of @code{rte}. This attribute expects an integer
5082 argument specifying the trap number to be used.
5085 @cindex @code{trapa_handler} function attribute, SH
5086 On SH targets this function attribute is similar to @code{interrupt_handler}
5087 but it does not save and restore all registers.
5090 @node SPU Function Attributes
5091 @subsection SPU Function Attributes
5093 These function attributes are supported by the SPU back end:
5097 @cindex @code{naked} function attribute, SPU
5098 This attribute allows the compiler to construct the
5099 requisite function declaration, while allowing the body of the
5100 function to be assembly code. The specified function will not have
5101 prologue/epilogue sequences generated by the compiler. Only basic
5102 @code{asm} statements can safely be included in naked functions
5103 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5104 basic @code{asm} and C code may appear to work, they cannot be
5105 depended upon to work reliably and are not supported.
5108 @node Symbian OS Function Attributes
5109 @subsection Symbian OS Function Attributes
5111 @xref{Microsoft Windows Function Attributes}, for discussion of the
5112 @code{dllexport} and @code{dllimport} attributes.
5114 @node V850 Function Attributes
5115 @subsection V850 Function Attributes
5117 The V850 back end supports these function attributes:
5121 @itemx interrupt_handler
5122 @cindex @code{interrupt} function attribute, V850
5123 @cindex @code{interrupt_handler} function attribute, V850
5124 Use these attributes to indicate
5125 that the specified function is an interrupt handler. The compiler generates
5126 function entry and exit sequences suitable for use in an interrupt handler
5127 when either attribute is present.
5130 @node Visium Function Attributes
5131 @subsection Visium Function Attributes
5133 These function attributes are supported by the Visium back end:
5137 @cindex @code{interrupt} function attribute, Visium
5138 Use this attribute to indicate
5139 that the specified function is an interrupt handler. The compiler generates
5140 function entry and exit sequences suitable for use in an interrupt handler
5141 when this attribute is present.
5144 @node x86 Function Attributes
5145 @subsection x86 Function Attributes
5147 These function attributes are supported by the x86 back end:
5151 @cindex @code{cdecl} function attribute, x86-32
5152 @cindex functions that pop the argument stack on x86-32
5154 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5155 assume that the calling function pops off the stack space used to
5156 pass arguments. This is
5157 useful to override the effects of the @option{-mrtd} switch.
5160 @cindex @code{fastcall} function attribute, x86-32
5161 @cindex functions that pop the argument stack on x86-32
5162 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5163 pass the first argument (if of integral type) in the register ECX and
5164 the second argument (if of integral type) in the register EDX@. Subsequent
5165 and other typed arguments are passed on the stack. The called function
5166 pops the arguments off the stack. If the number of arguments is variable all
5167 arguments are pushed on the stack.
5170 @cindex @code{thiscall} function attribute, x86-32
5171 @cindex functions that pop the argument stack on x86-32
5172 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5173 pass the first argument (if of integral type) in the register ECX.
5174 Subsequent and other typed arguments are passed on the stack. The called
5175 function pops the arguments off the stack.
5176 If the number of arguments is variable all arguments are pushed on the
5178 The @code{thiscall} attribute is intended for C++ non-static member functions.
5179 As a GCC extension, this calling convention can be used for C functions
5180 and for static member methods.
5184 @cindex @code{ms_abi} function attribute, x86
5185 @cindex @code{sysv_abi} function attribute, x86
5187 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5188 to indicate which calling convention should be used for a function. The
5189 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5190 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5191 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5192 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5194 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5195 requires the @option{-maccumulate-outgoing-args} option.
5197 @item callee_pop_aggregate_return (@var{number})
5198 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5200 On x86-32 targets, you can use this attribute to control how
5201 aggregates are returned in memory. If the caller is responsible for
5202 popping the hidden pointer together with the rest of the arguments, specify
5203 @var{number} equal to zero. If callee is responsible for popping the
5204 hidden pointer, specify @var{number} equal to one.
5206 The default x86-32 ABI assumes that the callee pops the
5207 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5208 the compiler assumes that the
5209 caller pops the stack for hidden pointer.
5211 @item ms_hook_prologue
5212 @cindex @code{ms_hook_prologue} function attribute, x86
5214 On 32-bit and 64-bit x86 targets, you can use
5215 this function attribute to make GCC generate the ``hot-patching'' function
5216 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5219 @item regparm (@var{number})
5220 @cindex @code{regparm} function attribute, x86
5221 @cindex functions that are passed arguments in registers on x86-32
5222 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5223 pass arguments number one to @var{number} if they are of integral type
5224 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5225 take a variable number of arguments continue to be passed all of their
5226 arguments on the stack.
5228 Beware that on some ELF systems this attribute is unsuitable for
5229 global functions in shared libraries with lazy binding (which is the
5230 default). Lazy binding sends the first call via resolving code in
5231 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5232 per the standard calling conventions. Solaris 8 is affected by this.
5233 Systems with the GNU C Library version 2.1 or higher
5234 and FreeBSD are believed to be
5235 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5236 disabled with the linker or the loader if desired, to avoid the
5240 @cindex @code{sseregparm} function attribute, x86
5241 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5242 causes the compiler to pass up to 3 floating-point arguments in
5243 SSE registers instead of on the stack. Functions that take a
5244 variable number of arguments continue to pass all of their
5245 floating-point arguments on the stack.
5247 @item force_align_arg_pointer
5248 @cindex @code{force_align_arg_pointer} function attribute, x86
5249 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5250 applied to individual function definitions, generating an alternate
5251 prologue and epilogue that realigns the run-time stack if necessary.
5252 This supports mixing legacy codes that run with a 4-byte aligned stack
5253 with modern codes that keep a 16-byte stack for SSE compatibility.
5256 @cindex @code{stdcall} function attribute, x86-32
5257 @cindex functions that pop the argument stack on x86-32
5258 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5259 assume that the called function pops off the stack space used to
5260 pass arguments, unless it takes a variable number of arguments.
5262 @item target (@var{options})
5263 @cindex @code{target} function attribute
5264 As discussed in @ref{Common Function Attributes}, this attribute
5265 allows specification of target-specific compilation options.
5267 On the x86, the following options are allowed:
5271 @cindex @code{target("abm")} function attribute, x86
5272 Enable/disable the generation of the advanced bit instructions.
5276 @cindex @code{target("aes")} function attribute, x86
5277 Enable/disable the generation of the AES instructions.
5280 @cindex @code{target("default")} function attribute, x86
5281 @xref{Function Multiversioning}, where it is used to specify the
5282 default function version.
5286 @cindex @code{target("mmx")} function attribute, x86
5287 Enable/disable the generation of the MMX instructions.
5291 @cindex @code{target("pclmul")} function attribute, x86
5292 Enable/disable the generation of the PCLMUL instructions.
5296 @cindex @code{target("popcnt")} function attribute, x86
5297 Enable/disable the generation of the POPCNT instruction.
5301 @cindex @code{target("sse")} function attribute, x86
5302 Enable/disable the generation of the SSE instructions.
5306 @cindex @code{target("sse2")} function attribute, x86
5307 Enable/disable the generation of the SSE2 instructions.
5311 @cindex @code{target("sse3")} function attribute, x86
5312 Enable/disable the generation of the SSE3 instructions.
5316 @cindex @code{target("sse4")} function attribute, x86
5317 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5322 @cindex @code{target("sse4.1")} function attribute, x86
5323 Enable/disable the generation of the sse4.1 instructions.
5327 @cindex @code{target("sse4.2")} function attribute, x86
5328 Enable/disable the generation of the sse4.2 instructions.
5332 @cindex @code{target("sse4a")} function attribute, x86
5333 Enable/disable the generation of the SSE4A instructions.
5337 @cindex @code{target("fma4")} function attribute, x86
5338 Enable/disable the generation of the FMA4 instructions.
5342 @cindex @code{target("xop")} function attribute, x86
5343 Enable/disable the generation of the XOP instructions.
5347 @cindex @code{target("lwp")} function attribute, x86
5348 Enable/disable the generation of the LWP instructions.
5352 @cindex @code{target("ssse3")} function attribute, x86
5353 Enable/disable the generation of the SSSE3 instructions.
5357 @cindex @code{target("cld")} function attribute, x86
5358 Enable/disable the generation of the CLD before string moves.
5360 @item fancy-math-387
5361 @itemx no-fancy-math-387
5362 @cindex @code{target("fancy-math-387")} function attribute, x86
5363 Enable/disable the generation of the @code{sin}, @code{cos}, and
5364 @code{sqrt} instructions on the 387 floating-point unit.
5367 @itemx no-fused-madd
5368 @cindex @code{target("fused-madd")} function attribute, x86
5369 Enable/disable the generation of the fused multiply/add instructions.
5373 @cindex @code{target("ieee-fp")} function attribute, x86
5374 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5376 @item inline-all-stringops
5377 @itemx no-inline-all-stringops
5378 @cindex @code{target("inline-all-stringops")} function attribute, x86
5379 Enable/disable inlining of string operations.
5381 @item inline-stringops-dynamically
5382 @itemx no-inline-stringops-dynamically
5383 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5384 Enable/disable the generation of the inline code to do small string
5385 operations and calling the library routines for large operations.
5387 @item align-stringops
5388 @itemx no-align-stringops
5389 @cindex @code{target("align-stringops")} function attribute, x86
5390 Do/do not align destination of inlined string operations.
5394 @cindex @code{target("recip")} function attribute, x86
5395 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5396 instructions followed an additional Newton-Raphson step instead of
5397 doing a floating-point division.
5399 @item arch=@var{ARCH}
5400 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5401 Specify the architecture to generate code for in compiling the function.
5403 @item tune=@var{TUNE}
5404 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5405 Specify the architecture to tune for in compiling the function.
5407 @item fpmath=@var{FPMATH}
5408 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5409 Specify which floating-point unit to use. You must specify the
5410 @code{target("fpmath=sse,387")} option as
5411 @code{target("fpmath=sse+387")} because the comma would separate
5415 On the x86, the inliner does not inline a
5416 function that has different target options than the caller, unless the
5417 callee has a subset of the target options of the caller. For example
5418 a function declared with @code{target("sse3")} can inline a function
5419 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5422 @node Xstormy16 Function Attributes
5423 @subsection Xstormy16 Function Attributes
5425 These function attributes are supported by the Xstormy16 back end:
5429 @cindex @code{interrupt} function attribute, Xstormy16
5430 Use this attribute to indicate
5431 that the specified function is an interrupt handler. The compiler generates
5432 function entry and exit sequences suitable for use in an interrupt handler
5433 when this attribute is present.
5436 @node Variable Attributes
5437 @section Specifying Attributes of Variables
5438 @cindex attribute of variables
5439 @cindex variable attributes
5441 The keyword @code{__attribute__} allows you to specify special
5442 attributes of variables or structure fields. This keyword is followed
5443 by an attribute specification inside double parentheses. Some
5444 attributes are currently defined generically for variables.
5445 Other attributes are defined for variables on particular target
5446 systems. Other attributes are available for functions
5447 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5448 enumerators (@pxref{Enumerator Attributes}), and for types
5449 (@pxref{Type Attributes}).
5450 Other front ends might define more attributes
5451 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5453 @xref{Attribute Syntax}, for details of the exact syntax for using
5457 * Common Variable Attributes::
5458 * AVR Variable Attributes::
5459 * Blackfin Variable Attributes::
5460 * H8/300 Variable Attributes::
5461 * IA-64 Variable Attributes::
5462 * M32R/D Variable Attributes::
5463 * MeP Variable Attributes::
5464 * Microsoft Windows Variable Attributes::
5465 * MSP430 Variable Attributes::
5466 * PowerPC Variable Attributes::
5467 * RL78 Variable Attributes::
5468 * SPU Variable Attributes::
5469 * V850 Variable Attributes::
5470 * x86 Variable Attributes::
5471 * Xstormy16 Variable Attributes::
5474 @node Common Variable Attributes
5475 @subsection Common Variable Attributes
5477 The following attributes are supported on most targets.
5480 @cindex @code{aligned} variable attribute
5481 @item aligned (@var{alignment})
5482 This attribute specifies a minimum alignment for the variable or
5483 structure field, measured in bytes. For example, the declaration:
5486 int x __attribute__ ((aligned (16))) = 0;
5490 causes the compiler to allocate the global variable @code{x} on a
5491 16-byte boundary. On a 68040, this could be used in conjunction with
5492 an @code{asm} expression to access the @code{move16} instruction which
5493 requires 16-byte aligned operands.
5495 You can also specify the alignment of structure fields. For example, to
5496 create a double-word aligned @code{int} pair, you could write:
5499 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5503 This is an alternative to creating a union with a @code{double} member,
5504 which forces the union to be double-word aligned.
5506 As in the preceding examples, you can explicitly specify the alignment
5507 (in bytes) that you wish the compiler to use for a given variable or
5508 structure field. Alternatively, you can leave out the alignment factor
5509 and just ask the compiler to align a variable or field to the
5510 default alignment for the target architecture you are compiling for.
5511 The default alignment is sufficient for all scalar types, but may not be
5512 enough for all vector types on a target that supports vector operations.
5513 The default alignment is fixed for a particular target ABI.
5515 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5516 which is the largest alignment ever used for any data type on the
5517 target machine you are compiling for. For example, you could write:
5520 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5523 The compiler automatically sets the alignment for the declared
5524 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5525 often make copy operations more efficient, because the compiler can
5526 use whatever instructions copy the biggest chunks of memory when
5527 performing copies to or from the variables or fields that you have
5528 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5529 may change depending on command-line options.
5531 When used on a struct, or struct member, the @code{aligned} attribute can
5532 only increase the alignment; in order to decrease it, the @code{packed}
5533 attribute must be specified as well. When used as part of a typedef, the
5534 @code{aligned} attribute can both increase and decrease alignment, and
5535 specifying the @code{packed} attribute generates a warning.
5537 Note that the effectiveness of @code{aligned} attributes may be limited
5538 by inherent limitations in your linker. On many systems, the linker is
5539 only able to arrange for variables to be aligned up to a certain maximum
5540 alignment. (For some linkers, the maximum supported alignment may
5541 be very very small.) If your linker is only able to align variables
5542 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5543 in an @code{__attribute__} still only provides you with 8-byte
5544 alignment. See your linker documentation for further information.
5546 The @code{aligned} attribute can also be used for functions
5547 (@pxref{Common Function Attributes}.)
5549 @item cleanup (@var{cleanup_function})
5550 @cindex @code{cleanup} variable attribute
5551 The @code{cleanup} attribute runs a function when the variable goes
5552 out of scope. This attribute can only be applied to auto function
5553 scope variables; it may not be applied to parameters or variables
5554 with static storage duration. The function must take one parameter,
5555 a pointer to a type compatible with the variable. The return value
5556 of the function (if any) is ignored.
5558 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5559 is run during the stack unwinding that happens during the
5560 processing of the exception. Note that the @code{cleanup} attribute
5561 does not allow the exception to be caught, only to perform an action.
5562 It is undefined what happens if @var{cleanup_function} does not
5567 @cindex @code{common} variable attribute
5568 @cindex @code{nocommon} variable attribute
5571 The @code{common} attribute requests GCC to place a variable in
5572 ``common'' storage. The @code{nocommon} attribute requests the
5573 opposite---to allocate space for it directly.
5575 These attributes override the default chosen by the
5576 @option{-fno-common} and @option{-fcommon} flags respectively.
5579 @itemx deprecated (@var{msg})
5580 @cindex @code{deprecated} variable attribute
5581 The @code{deprecated} attribute results in a warning if the variable
5582 is used anywhere in the source file. This is useful when identifying
5583 variables that are expected to be removed in a future version of a
5584 program. The warning also includes the location of the declaration
5585 of the deprecated variable, to enable users to easily find further
5586 information about why the variable is deprecated, or what they should
5587 do instead. Note that the warning only occurs for uses:
5590 extern int old_var __attribute__ ((deprecated));
5592 int new_fn () @{ return old_var; @}
5596 results in a warning on line 3 but not line 2. The optional @var{msg}
5597 argument, which must be a string, is printed in the warning if
5600 The @code{deprecated} attribute can also be used for functions and
5601 types (@pxref{Common Function Attributes},
5602 @pxref{Common Type Attributes}).
5604 @item mode (@var{mode})
5605 @cindex @code{mode} variable attribute
5606 This attribute specifies the data type for the declaration---whichever
5607 type corresponds to the mode @var{mode}. This in effect lets you
5608 request an integer or floating-point type according to its width.
5610 You may also specify a mode of @code{byte} or @code{__byte__} to
5611 indicate the mode corresponding to a one-byte integer, @code{word} or
5612 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5613 or @code{__pointer__} for the mode used to represent pointers.
5616 @cindex @code{packed} variable attribute
5617 The @code{packed} attribute specifies that a variable or structure field
5618 should have the smallest possible alignment---one byte for a variable,
5619 and one bit for a field, unless you specify a larger value with the
5620 @code{aligned} attribute.
5622 Here is a structure in which the field @code{x} is packed, so that it
5623 immediately follows @code{a}:
5629 int x[2] __attribute__ ((packed));
5633 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5634 @code{packed} attribute on bit-fields of type @code{char}. This has
5635 been fixed in GCC 4.4 but the change can lead to differences in the
5636 structure layout. See the documentation of
5637 @option{-Wpacked-bitfield-compat} for more information.
5639 @item section ("@var{section-name}")
5640 @cindex @code{section} variable attribute
5641 Normally, the compiler places the objects it generates in sections like
5642 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5643 or you need certain particular variables to appear in special sections,
5644 for example to map to special hardware. The @code{section}
5645 attribute specifies that a variable (or function) lives in a particular
5646 section. For example, this small program uses several specific section names:
5649 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5650 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5651 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5652 int init_data __attribute__ ((section ("INITDATA")));
5656 /* @r{Initialize stack pointer} */
5657 init_sp (stack + sizeof (stack));
5659 /* @r{Initialize initialized data} */
5660 memcpy (&init_data, &data, &edata - &data);
5662 /* @r{Turn on the serial ports} */
5669 Use the @code{section} attribute with
5670 @emph{global} variables and not @emph{local} variables,
5671 as shown in the example.
5673 You may use the @code{section} attribute with initialized or
5674 uninitialized global variables but the linker requires
5675 each object be defined once, with the exception that uninitialized
5676 variables tentatively go in the @code{common} (or @code{bss}) section
5677 and can be multiply ``defined''. Using the @code{section} attribute
5678 changes what section the variable goes into and may cause the
5679 linker to issue an error if an uninitialized variable has multiple
5680 definitions. You can force a variable to be initialized with the
5681 @option{-fno-common} flag or the @code{nocommon} attribute.
5683 Some file formats do not support arbitrary sections so the @code{section}
5684 attribute is not available on all platforms.
5685 If you need to map the entire contents of a module to a particular
5686 section, consider using the facilities of the linker instead.
5688 @item tls_model ("@var{tls_model}")
5689 @cindex @code{tls_model} variable attribute
5690 The @code{tls_model} attribute sets thread-local storage model
5691 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5692 overriding @option{-ftls-model=} command-line switch on a per-variable
5694 The @var{tls_model} argument should be one of @code{global-dynamic},
5695 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5697 Not all targets support this attribute.
5700 @cindex @code{unused} variable attribute
5701 This attribute, attached to a variable, means that the variable is meant
5702 to be possibly unused. GCC does not produce a warning for this
5706 @cindex @code{used} variable attribute
5707 This attribute, attached to a variable with static storage, means that
5708 the variable must be emitted even if it appears that the variable is not
5711 When applied to a static data member of a C++ class template, the
5712 attribute also means that the member is instantiated if the
5713 class itself is instantiated.
5715 @item vector_size (@var{bytes})
5716 @cindex @code{vector_size} variable attribute
5717 This attribute specifies the vector size for the variable, measured in
5718 bytes. For example, the declaration:
5721 int foo __attribute__ ((vector_size (16)));
5725 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5726 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5727 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5729 This attribute is only applicable to integral and float scalars,
5730 although arrays, pointers, and function return values are allowed in
5731 conjunction with this construct.
5733 Aggregates with this attribute are invalid, even if they are of the same
5734 size as a corresponding scalar. For example, the declaration:
5737 struct S @{ int a; @};
5738 struct S __attribute__ ((vector_size (16))) foo;
5742 is invalid even if the size of the structure is the same as the size of
5745 @item visibility ("@var{visibility_type}")
5746 @cindex @code{visibility} variable attribute
5747 This attribute affects the linkage of the declaration to which it is attached.
5748 The @code{visibility} attribute is described in
5749 @ref{Common Function Attributes}.
5752 @cindex @code{weak} variable attribute
5753 The @code{weak} attribute is described in
5754 @ref{Common Function Attributes}.
5758 @node AVR Variable Attributes
5759 @subsection AVR Variable Attributes
5763 @cindex @code{progmem} variable attribute, AVR
5764 The @code{progmem} attribute is used on the AVR to place read-only
5765 data in the non-volatile program memory (flash). The @code{progmem}
5766 attribute accomplishes this by putting respective variables into a
5767 section whose name starts with @code{.progmem}.
5769 This attribute works similar to the @code{section} attribute
5770 but adds additional checking. Notice that just like the
5771 @code{section} attribute, @code{progmem} affects the location
5772 of the data but not how this data is accessed.
5774 In order to read data located with the @code{progmem} attribute
5775 (inline) assembler must be used.
5777 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5778 #include <avr/pgmspace.h>
5780 /* Locate var in flash memory */
5781 const int var[2] PROGMEM = @{ 1, 2 @};
5783 int read_var (int i)
5785 /* Access var[] by accessor macro from avr/pgmspace.h */
5786 return (int) pgm_read_word (& var[i]);
5790 AVR is a Harvard architecture processor and data and read-only data
5791 normally resides in the data memory (RAM).
5793 See also the @ref{AVR Named Address Spaces} section for
5794 an alternate way to locate and access data in flash memory.
5797 @itemx io (@var{addr})
5798 @cindex @code{io} variable attribute, AVR
5799 Variables with the @code{io} attribute are used to address
5800 memory-mapped peripherals in the io address range.
5801 If an address is specified, the variable
5802 is assigned that address, and the value is interpreted as an
5803 address in the data address space.
5807 volatile int porta __attribute__((io (0x22)));
5810 The address specified in the address in the data address range.
5812 Otherwise, the variable it is not assigned an address, but the
5813 compiler will still use in/out instructions where applicable,
5814 assuming some other module assigns an address in the io address range.
5818 extern volatile int porta __attribute__((io));
5822 @itemx io_low (@var{addr})
5823 @cindex @code{io_low} variable attribute, AVR
5824 This is like the @code{io} attribute, but additionally it informs the
5825 compiler that the object lies in the lower half of the I/O area,
5826 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5830 @itemx address (@var{addr})
5831 @cindex @code{address} variable attribute, AVR
5832 Variables with the @code{address} attribute are used to address
5833 memory-mapped peripherals that may lie outside the io address range.
5836 volatile int porta __attribute__((address (0x600)));
5841 @node Blackfin Variable Attributes
5842 @subsection Blackfin Variable Attributes
5844 Three attributes are currently defined for the Blackfin.
5850 @cindex @code{l1_data} variable attribute, Blackfin
5851 @cindex @code{l1_data_A} variable attribute, Blackfin
5852 @cindex @code{l1_data_B} variable attribute, Blackfin
5853 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5854 Variables with @code{l1_data} attribute are put into the specific section
5855 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5856 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5857 attribute are put into the specific section named @code{.l1.data.B}.
5860 @cindex @code{l2} variable attribute, Blackfin
5861 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5862 Variables with @code{l2} attribute are put into the specific section
5863 named @code{.l2.data}.
5866 @node H8/300 Variable Attributes
5867 @subsection H8/300 Variable Attributes
5869 These variable attributes are available for H8/300 targets:
5873 @cindex @code{eightbit_data} variable attribute, H8/300
5874 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5875 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5876 variable should be placed into the eight-bit data section.
5877 The compiler generates more efficient code for certain operations
5878 on data in the eight-bit data area. Note the eight-bit data area is limited to
5881 You must use GAS and GLD from GNU binutils version 2.7 or later for
5882 this attribute to work correctly.
5885 @cindex @code{tiny_data} variable attribute, H8/300
5886 @cindex tiny data section on the H8/300H and H8S
5887 Use this attribute on the H8/300H and H8S to indicate that the specified
5888 variable should be placed into the tiny data section.
5889 The compiler generates more efficient code for loads and stores
5890 on data in the tiny data section. Note the tiny data area is limited to
5891 slightly under 32KB of data.
5895 @node IA-64 Variable Attributes
5896 @subsection IA-64 Variable Attributes
5898 The IA-64 back end supports the following variable attribute:
5901 @item model (@var{model-name})
5902 @cindex @code{model} variable attribute, IA-64
5904 On IA-64, use this attribute to set the addressability of an object.
5905 At present, the only supported identifier for @var{model-name} is
5906 @code{small}, indicating addressability via ``small'' (22-bit)
5907 addresses (so that their addresses can be loaded with the @code{addl}
5908 instruction). Caveat: such addressing is by definition not position
5909 independent and hence this attribute must not be used for objects
5910 defined by shared libraries.
5914 @node M32R/D Variable Attributes
5915 @subsection M32R/D Variable Attributes
5917 One attribute is currently defined for the M32R/D@.
5920 @item model (@var{model-name})
5921 @cindex @code{model-name} variable attribute, M32R/D
5922 @cindex variable addressability on the M32R/D
5923 Use this attribute on the M32R/D to set the addressability of an object.
5924 The identifier @var{model-name} is one of @code{small}, @code{medium},
5925 or @code{large}, representing each of the code models.
5927 Small model objects live in the lower 16MB of memory (so that their
5928 addresses can be loaded with the @code{ld24} instruction).
5930 Medium and large model objects may live anywhere in the 32-bit address space
5931 (the compiler generates @code{seth/add3} instructions to load their
5935 @node MeP Variable Attributes
5936 @subsection MeP Variable Attributes
5938 The MeP target has a number of addressing modes and busses. The
5939 @code{near} space spans the standard memory space's first 16 megabytes
5940 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5941 The @code{based} space is a 128-byte region in the memory space that
5942 is addressed relative to the @code{$tp} register. The @code{tiny}
5943 space is a 65536-byte region relative to the @code{$gp} register. In
5944 addition to these memory regions, the MeP target has a separate 16-bit
5945 control bus which is specified with @code{cb} attributes.
5950 @cindex @code{based} variable attribute, MeP
5951 Any variable with the @code{based} attribute is assigned to the
5952 @code{.based} section, and is accessed with relative to the
5953 @code{$tp} register.
5956 @cindex @code{tiny} variable attribute, MeP
5957 Likewise, the @code{tiny} attribute assigned variables to the
5958 @code{.tiny} section, relative to the @code{$gp} register.
5961 @cindex @code{near} variable attribute, MeP
5962 Variables with the @code{near} attribute are assumed to have addresses
5963 that fit in a 24-bit addressing mode. This is the default for large
5964 variables (@code{-mtiny=4} is the default) but this attribute can
5965 override @code{-mtiny=} for small variables, or override @code{-ml}.
5968 @cindex @code{far} variable attribute, MeP
5969 Variables with the @code{far} attribute are addressed using a full
5970 32-bit address. Since this covers the entire memory space, this
5971 allows modules to make no assumptions about where variables might be
5975 @cindex @code{io} variable attribute, MeP
5976 @itemx io (@var{addr})
5977 Variables with the @code{io} attribute are used to address
5978 memory-mapped peripherals. If an address is specified, the variable
5979 is assigned that address, else it is not assigned an address (it is
5980 assumed some other module assigns an address). Example:
5983 int timer_count __attribute__((io(0x123)));
5987 @itemx cb (@var{addr})
5988 @cindex @code{cb} variable attribute, MeP
5989 Variables with the @code{cb} attribute are used to access the control
5990 bus, using special instructions. @code{addr} indicates the control bus
5994 int cpu_clock __attribute__((cb(0x123)));
5999 @node Microsoft Windows Variable Attributes
6000 @subsection Microsoft Windows Variable Attributes
6002 You can use these attributes on Microsoft Windows targets.
6003 @ref{x86 Variable Attributes} for additional Windows compatibility
6004 attributes available on all x86 targets.
6009 @cindex @code{dllimport} variable attribute
6010 @cindex @code{dllexport} variable attribute
6011 The @code{dllimport} and @code{dllexport} attributes are described in
6012 @ref{Microsoft Windows Function Attributes}.
6015 @cindex @code{selectany} variable attribute
6016 The @code{selectany} attribute causes an initialized global variable to
6017 have link-once semantics. When multiple definitions of the variable are
6018 encountered by the linker, the first is selected and the remainder are
6019 discarded. Following usage by the Microsoft compiler, the linker is told
6020 @emph{not} to warn about size or content differences of the multiple
6023 Although the primary usage of this attribute is for POD types, the
6024 attribute can also be applied to global C++ objects that are initialized
6025 by a constructor. In this case, the static initialization and destruction
6026 code for the object is emitted in each translation defining the object,
6027 but the calls to the constructor and destructor are protected by a
6028 link-once guard variable.
6030 The @code{selectany} attribute is only available on Microsoft Windows
6031 targets. You can use @code{__declspec (selectany)} as a synonym for
6032 @code{__attribute__ ((selectany))} for compatibility with other
6036 @cindex @code{shared} variable attribute
6037 On Microsoft Windows, in addition to putting variable definitions in a named
6038 section, the section can also be shared among all running copies of an
6039 executable or DLL@. For example, this small program defines shared data
6040 by putting it in a named section @code{shared} and marking the section
6044 int foo __attribute__((section ("shared"), shared)) = 0;
6049 /* @r{Read and write foo. All running
6050 copies see the same value.} */
6056 You may only use the @code{shared} attribute along with @code{section}
6057 attribute with a fully-initialized global definition because of the way
6058 linkers work. See @code{section} attribute for more information.
6060 The @code{shared} attribute is only available on Microsoft Windows@.
6064 @node MSP430 Variable Attributes
6065 @subsection MSP430 Variable Attributes
6069 @cindex @code{noinit} variable attribute, MSP430
6070 Any data with the @code{noinit} attribute will not be initialised by
6071 the C runtime startup code, or the program loader. Not initialising
6072 data in this way can reduce program startup times.
6075 @cindex @code{persistent} variable attribute, MSP430
6076 Any variable with the @code{persistent} attribute will not be
6077 initialised by the C runtime startup code. Instead its value will be
6078 set once, when the application is loaded, and then never initialised
6079 again, even if the processor is reset or the program restarts.
6080 Persistent data is intended to be placed into FLASH RAM, where its
6081 value will be retained across resets. The linker script being used to
6082 create the application should ensure that persistent data is correctly
6088 @cindex @code{lower} variable attribute, MSP430
6089 @cindex @code{upper} variable attribute, MSP430
6090 @cindex @code{either} variable attribute, MSP430
6091 These attributes are the same as the MSP430 function attributes of the
6092 same name (@pxref{MSP430 Function Attributes}).
6093 These attributes can be applied to both functions and variables.
6096 @node PowerPC Variable Attributes
6097 @subsection PowerPC Variable Attributes
6099 Three attributes currently are defined for PowerPC configurations:
6100 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6102 @cindex @code{ms_struct} variable attribute, PowerPC
6103 @cindex @code{gcc_struct} variable attribute, PowerPC
6104 For full documentation of the struct attributes please see the
6105 documentation in @ref{x86 Variable Attributes}.
6107 @cindex @code{altivec} variable attribute, PowerPC
6108 For documentation of @code{altivec} attribute please see the
6109 documentation in @ref{PowerPC Type Attributes}.
6111 @node RL78 Variable Attributes
6112 @subsection RL78 Variable Attributes
6114 @cindex @code{saddr} variable attribute, RL78
6115 The RL78 back end supports the @code{saddr} variable attribute. This
6116 specifies placement of the corresponding variable in the SADDR area,
6117 which can be accessed more efficiently than the default memory region.
6119 @node SPU Variable Attributes
6120 @subsection SPU Variable Attributes
6122 @cindex @code{spu_vector} variable attribute, SPU
6123 The SPU supports the @code{spu_vector} attribute for variables. For
6124 documentation of this attribute please see the documentation in
6125 @ref{SPU Type Attributes}.
6127 @node V850 Variable Attributes
6128 @subsection V850 Variable Attributes
6130 These variable attributes are supported by the V850 back end:
6135 @cindex @code{sda} variable attribute, V850
6136 Use this attribute to explicitly place a variable in the small data area,
6137 which can hold up to 64 kilobytes.
6140 @cindex @code{tda} variable attribute, V850
6141 Use this attribute to explicitly place a variable in the tiny data area,
6142 which can hold up to 256 bytes in total.
6145 @cindex @code{zda} variable attribute, V850
6146 Use this attribute to explicitly place a variable in the first 32 kilobytes
6150 @node x86 Variable Attributes
6151 @subsection x86 Variable Attributes
6153 Two attributes are currently defined for x86 configurations:
6154 @code{ms_struct} and @code{gcc_struct}.
6159 @cindex @code{ms_struct} variable attribute, x86
6160 @cindex @code{gcc_struct} variable attribute, x86
6162 If @code{packed} is used on a structure, or if bit-fields are used,
6163 it may be that the Microsoft ABI lays out the structure differently
6164 than the way GCC normally does. Particularly when moving packed
6165 data between functions compiled with GCC and the native Microsoft compiler
6166 (either via function call or as data in a file), it may be necessary to access
6169 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6170 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6171 command-line options, respectively;
6172 see @ref{x86 Options}, for details of how structure layout is affected.
6173 @xref{x86 Type Attributes}, for information about the corresponding
6174 attributes on types.
6178 @node Xstormy16 Variable Attributes
6179 @subsection Xstormy16 Variable Attributes
6181 One attribute is currently defined for xstormy16 configurations:
6186 @cindex @code{below100} variable attribute, Xstormy16
6188 If a variable has the @code{below100} attribute (@code{BELOW100} is
6189 allowed also), GCC places the variable in the first 0x100 bytes of
6190 memory and use special opcodes to access it. Such variables are
6191 placed in either the @code{.bss_below100} section or the
6192 @code{.data_below100} section.
6196 @node Type Attributes
6197 @section Specifying Attributes of Types
6198 @cindex attribute of types
6199 @cindex type attributes
6201 The keyword @code{__attribute__} allows you to specify special
6202 attributes of types. Some type attributes apply only to @code{struct}
6203 and @code{union} types, while others can apply to any type defined
6204 via a @code{typedef} declaration. Other attributes are defined for
6205 functions (@pxref{Function Attributes}), labels (@pxref{Label
6206 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6207 variables (@pxref{Variable Attributes}).
6209 The @code{__attribute__} keyword is followed by an attribute specification
6210 inside double parentheses.
6212 You may specify type attributes in an enum, struct or union type
6213 declaration or definition by placing them immediately after the
6214 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6215 syntax is to place them just past the closing curly brace of the
6218 You can also include type attributes in a @code{typedef} declaration.
6219 @xref{Attribute Syntax}, for details of the exact syntax for using
6223 * Common Type Attributes::
6224 * ARM Type Attributes::
6225 * MeP Type Attributes::
6226 * PowerPC Type Attributes::
6227 * SPU Type Attributes::
6228 * x86 Type Attributes::
6231 @node Common Type Attributes
6232 @subsection Common Type Attributes
6234 The following type attributes are supported on most targets.
6237 @cindex @code{aligned} type attribute
6238 @item aligned (@var{alignment})
6239 This attribute specifies a minimum alignment (in bytes) for variables
6240 of the specified type. For example, the declarations:
6243 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6244 typedef int more_aligned_int __attribute__ ((aligned (8)));
6248 force the compiler to ensure (as far as it can) that each variable whose
6249 type is @code{struct S} or @code{more_aligned_int} is allocated and
6250 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6251 variables of type @code{struct S} aligned to 8-byte boundaries allows
6252 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6253 store) instructions when copying one variable of type @code{struct S} to
6254 another, thus improving run-time efficiency.
6256 Note that the alignment of any given @code{struct} or @code{union} type
6257 is required by the ISO C standard to be at least a perfect multiple of
6258 the lowest common multiple of the alignments of all of the members of
6259 the @code{struct} or @code{union} in question. This means that you @emph{can}
6260 effectively adjust the alignment of a @code{struct} or @code{union}
6261 type by attaching an @code{aligned} attribute to any one of the members
6262 of such a type, but the notation illustrated in the example above is a
6263 more obvious, intuitive, and readable way to request the compiler to
6264 adjust the alignment of an entire @code{struct} or @code{union} type.
6266 As in the preceding example, you can explicitly specify the alignment
6267 (in bytes) that you wish the compiler to use for a given @code{struct}
6268 or @code{union} type. Alternatively, you can leave out the alignment factor
6269 and just ask the compiler to align a type to the maximum
6270 useful alignment for the target machine you are compiling for. For
6271 example, you could write:
6274 struct S @{ short f[3]; @} __attribute__ ((aligned));
6277 Whenever you leave out the alignment factor in an @code{aligned}
6278 attribute specification, the compiler automatically sets the alignment
6279 for the type to the largest alignment that is ever used for any data
6280 type on the target machine you are compiling for. Doing this can often
6281 make copy operations more efficient, because the compiler can use
6282 whatever instructions copy the biggest chunks of memory when performing
6283 copies to or from the variables that have types that you have aligned
6286 In the example above, if the size of each @code{short} is 2 bytes, then
6287 the size of the entire @code{struct S} type is 6 bytes. The smallest
6288 power of two that is greater than or equal to that is 8, so the
6289 compiler sets the alignment for the entire @code{struct S} type to 8
6292 Note that although you can ask the compiler to select a time-efficient
6293 alignment for a given type and then declare only individual stand-alone
6294 objects of that type, the compiler's ability to select a time-efficient
6295 alignment is primarily useful only when you plan to create arrays of
6296 variables having the relevant (efficiently aligned) type. If you
6297 declare or use arrays of variables of an efficiently-aligned type, then
6298 it is likely that your program also does pointer arithmetic (or
6299 subscripting, which amounts to the same thing) on pointers to the
6300 relevant type, and the code that the compiler generates for these
6301 pointer arithmetic operations is often more efficient for
6302 efficiently-aligned types than for other types.
6304 Note that the effectiveness of @code{aligned} attributes may be limited
6305 by inherent limitations in your linker. On many systems, the linker is
6306 only able to arrange for variables to be aligned up to a certain maximum
6307 alignment. (For some linkers, the maximum supported alignment may
6308 be very very small.) If your linker is only able to align variables
6309 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6310 in an @code{__attribute__} still only provides you with 8-byte
6311 alignment. See your linker documentation for further information.
6313 The @code{aligned} attribute can only increase alignment. Alignment
6314 can be decreased by specifying the @code{packed} attribute. See below.
6316 @item bnd_variable_size
6317 @cindex @code{bnd_variable_size} type attribute
6318 @cindex Pointer Bounds Checker attributes
6319 When applied to a structure field, this attribute tells Pointer
6320 Bounds Checker that the size of this field should not be computed
6321 using static type information. It may be used to mark variably-sized
6322 static array fields placed at the end of a structure.
6330 S *p = (S *)malloc (sizeof(S) + 100);
6331 p->data[10] = 0; //Bounds violation
6335 By using an attribute for the field we may avoid unwanted bound
6342 char data[1] __attribute__((bnd_variable_size));
6344 S *p = (S *)malloc (sizeof(S) + 100);
6345 p->data[10] = 0; //OK
6349 @itemx deprecated (@var{msg})
6350 @cindex @code{deprecated} type attribute
6351 The @code{deprecated} attribute results in a warning if the type
6352 is used anywhere in the source file. This is useful when identifying
6353 types that are expected to be removed in a future version of a program.
6354 If possible, the warning also includes the location of the declaration
6355 of the deprecated type, to enable users to easily find further
6356 information about why the type is deprecated, or what they should do
6357 instead. Note that the warnings only occur for uses and then only
6358 if the type is being applied to an identifier that itself is not being
6359 declared as deprecated.
6362 typedef int T1 __attribute__ ((deprecated));
6366 typedef T1 T3 __attribute__ ((deprecated));
6367 T3 z __attribute__ ((deprecated));
6371 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6372 warning is issued for line 4 because T2 is not explicitly
6373 deprecated. Line 5 has no warning because T3 is explicitly
6374 deprecated. Similarly for line 6. The optional @var{msg}
6375 argument, which must be a string, is printed in the warning if
6378 The @code{deprecated} attribute can also be used for functions and
6379 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6381 @item designated_init
6382 @cindex @code{designated_init} type attribute
6383 This attribute may only be applied to structure types. It indicates
6384 that any initialization of an object of this type must use designated
6385 initializers rather than positional initializers. The intent of this
6386 attribute is to allow the programmer to indicate that a structure's
6387 layout may change, and that therefore relying on positional
6388 initialization will result in future breakage.
6390 GCC emits warnings based on this attribute by default; use
6391 @option{-Wno-designated-init} to suppress them.
6394 @cindex @code{may_alias} type attribute
6395 Accesses through pointers to types with this attribute are not subject
6396 to type-based alias analysis, but are instead assumed to be able to alias
6397 any other type of objects.
6398 In the context of section 6.5 paragraph 7 of the C99 standard,
6399 an lvalue expression
6400 dereferencing such a pointer is treated like having a character type.
6401 See @option{-fstrict-aliasing} for more information on aliasing issues.
6402 This extension exists to support some vector APIs, in which pointers to
6403 one vector type are permitted to alias pointers to a different vector type.
6405 Note that an object of a type with this attribute does not have any
6411 typedef short __attribute__((__may_alias__)) short_a;
6417 short_a *b = (short_a *) &a;
6421 if (a == 0x12345678)
6429 If you replaced @code{short_a} with @code{short} in the variable
6430 declaration, the above program would abort when compiled with
6431 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6435 @cindex @code{packed} type attribute
6436 This attribute, attached to @code{struct} or @code{union} type
6437 definition, specifies that each member (other than zero-width bit-fields)
6438 of the structure or union is placed to minimize the memory required. When
6439 attached to an @code{enum} definition, it indicates that the smallest
6440 integral type should be used.
6442 @opindex fshort-enums
6443 Specifying the @code{packed} attribute for @code{struct} and @code{union}
6444 types is equivalent to specifying the @code{packed} attribute on each
6445 of the structure or union members. Specifying the @option{-fshort-enums}
6446 flag on the command line is equivalent to specifying the @code{packed}
6447 attribute on all @code{enum} definitions.
6449 In the following example @code{struct my_packed_struct}'s members are
6450 packed closely together, but the internal layout of its @code{s} member
6451 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6455 struct my_unpacked_struct
6461 struct __attribute__ ((__packed__)) my_packed_struct
6465 struct my_unpacked_struct s;
6469 You may only specify the @code{packed} attribute attribute on the definition
6470 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
6471 that does not also define the enumerated type, structure or union.
6473 @item scalar_storage_order ("@var{endianness}")
6474 @cindex @code{scalar_storage_order} type attribute
6475 When attached to a @code{union} or a @code{struct}, this attribute sets
6476 the storage order, aka endianness, of the scalar fields of the type, as
6477 well as the array fields whose component is scalar. The supported
6478 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
6479 has no effects on fields which are themselves a @code{union}, a @code{struct}
6480 or an array whose component is a @code{union} or a @code{struct}, and it is
6481 possible for these fields to have a different scalar storage order than the
6484 This attribute is supported only for targets that use a uniform default
6485 scalar storage order (fortunately, most of them), i.e. targets that store
6486 the scalars either all in big-endian or all in little-endian.
6488 Additional restrictions are enforced for types with the reverse scalar
6489 storage order with regard to the scalar storage order of the target:
6492 @item Taking the address of a scalar field of a @code{union} or a
6493 @code{struct} with reverse scalar storage order is not permitted and yields
6495 @item Taking the address of an array field, whose component is scalar, of
6496 a @code{union} or a @code{struct} with reverse scalar storage order is
6497 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
6499 @item Taking the address of a @code{union} or a @code{struct} with reverse
6500 scalar storage order is permitted.
6503 These restrictions exist because the storage order attribute is lost when
6504 the address of a scalar or the address of an array with scalar component is
6505 taken, so storing indirectly through this address generally does not work.
6506 The second case is nevertheless allowed to be able to perform a block copy
6507 from or to the array.
6509 Moreover, the use of type punning or aliasing to toggle the storage order
6510 is not supported; that is to say, a given scalar object cannot be accessed
6511 through distinct types that assign a different storage order to it.
6513 @item transparent_union
6514 @cindex @code{transparent_union} type attribute
6516 This attribute, attached to a @code{union} type definition, indicates
6517 that any function parameter having that union type causes calls to that
6518 function to be treated in a special way.
6520 First, the argument corresponding to a transparent union type can be of
6521 any type in the union; no cast is required. Also, if the union contains
6522 a pointer type, the corresponding argument can be a null pointer
6523 constant or a void pointer expression; and if the union contains a void
6524 pointer type, the corresponding argument can be any pointer expression.
6525 If the union member type is a pointer, qualifiers like @code{const} on
6526 the referenced type must be respected, just as with normal pointer
6529 Second, the argument is passed to the function using the calling
6530 conventions of the first member of the transparent union, not the calling
6531 conventions of the union itself. All members of the union must have the
6532 same machine representation; this is necessary for this argument passing
6535 Transparent unions are designed for library functions that have multiple
6536 interfaces for compatibility reasons. For example, suppose the
6537 @code{wait} function must accept either a value of type @code{int *} to
6538 comply with POSIX, or a value of type @code{union wait *} to comply with
6539 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6540 @code{wait} would accept both kinds of arguments, but it would also
6541 accept any other pointer type and this would make argument type checking
6542 less useful. Instead, @code{<sys/wait.h>} might define the interface
6546 typedef union __attribute__ ((__transparent_union__))
6550 @} wait_status_ptr_t;
6552 pid_t wait (wait_status_ptr_t);
6556 This interface allows either @code{int *} or @code{union wait *}
6557 arguments to be passed, using the @code{int *} calling convention.
6558 The program can call @code{wait} with arguments of either type:
6561 int w1 () @{ int w; return wait (&w); @}
6562 int w2 () @{ union wait w; return wait (&w); @}
6566 With this interface, @code{wait}'s implementation might look like this:
6569 pid_t wait (wait_status_ptr_t p)
6571 return waitpid (-1, p.__ip, 0);
6576 @cindex @code{unused} type attribute
6577 When attached to a type (including a @code{union} or a @code{struct}),
6578 this attribute means that variables of that type are meant to appear
6579 possibly unused. GCC does not produce a warning for any variables of
6580 that type, even if the variable appears to do nothing. This is often
6581 the case with lock or thread classes, which are usually defined and then
6582 not referenced, but contain constructors and destructors that have
6583 nontrivial bookkeeping functions.
6586 @cindex @code{visibility} type attribute
6587 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6588 applied to class, struct, union and enum types. Unlike other type
6589 attributes, the attribute must appear between the initial keyword and
6590 the name of the type; it cannot appear after the body of the type.
6592 Note that the type visibility is applied to vague linkage entities
6593 associated with the class (vtable, typeinfo node, etc.). In
6594 particular, if a class is thrown as an exception in one shared object
6595 and caught in another, the class must have default visibility.
6596 Otherwise the two shared objects are unable to use the same
6597 typeinfo node and exception handling will break.
6601 To specify multiple attributes, separate them by commas within the
6602 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6605 @node ARM Type Attributes
6606 @subsection ARM Type Attributes
6608 @cindex @code{notshared} type attribute, ARM
6609 On those ARM targets that support @code{dllimport} (such as Symbian
6610 OS), you can use the @code{notshared} attribute to indicate that the
6611 virtual table and other similar data for a class should not be
6612 exported from a DLL@. For example:
6615 class __declspec(notshared) C @{
6617 __declspec(dllimport) C();
6621 __declspec(dllexport)
6626 In this code, @code{C::C} is exported from the current DLL, but the
6627 virtual table for @code{C} is not exported. (You can use
6628 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6629 most Symbian OS code uses @code{__declspec}.)
6631 @node MeP Type Attributes
6632 @subsection MeP Type Attributes
6634 @cindex @code{based} type attribute, MeP
6635 @cindex @code{tiny} type attribute, MeP
6636 @cindex @code{near} type attribute, MeP
6637 @cindex @code{far} type attribute, MeP
6638 Many of the MeP variable attributes may be applied to types as well.
6639 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6640 @code{far} attributes may be applied to either. The @code{io} and
6641 @code{cb} attributes may not be applied to types.
6643 @node PowerPC Type Attributes
6644 @subsection PowerPC Type Attributes
6646 Three attributes currently are defined for PowerPC configurations:
6647 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6649 @cindex @code{ms_struct} type attribute, PowerPC
6650 @cindex @code{gcc_struct} type attribute, PowerPC
6651 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6652 attributes please see the documentation in @ref{x86 Type Attributes}.
6654 @cindex @code{altivec} type attribute, PowerPC
6655 The @code{altivec} attribute allows one to declare AltiVec vector data
6656 types supported by the AltiVec Programming Interface Manual. The
6657 attribute requires an argument to specify one of three vector types:
6658 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6659 and @code{bool__} (always followed by unsigned).
6662 __attribute__((altivec(vector__)))
6663 __attribute__((altivec(pixel__))) unsigned short
6664 __attribute__((altivec(bool__))) unsigned
6667 These attributes mainly are intended to support the @code{__vector},
6668 @code{__pixel}, and @code{__bool} AltiVec keywords.
6670 @node SPU Type Attributes
6671 @subsection SPU Type Attributes
6673 @cindex @code{spu_vector} type attribute, SPU
6674 The SPU supports the @code{spu_vector} attribute for types. This attribute
6675 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6676 Language Extensions Specification. It is intended to support the
6677 @code{__vector} keyword.
6679 @node x86 Type Attributes
6680 @subsection x86 Type Attributes
6682 Two attributes are currently defined for x86 configurations:
6683 @code{ms_struct} and @code{gcc_struct}.
6689 @cindex @code{ms_struct} type attribute, x86
6690 @cindex @code{gcc_struct} type attribute, x86
6692 If @code{packed} is used on a structure, or if bit-fields are used
6693 it may be that the Microsoft ABI packs them differently
6694 than GCC normally packs them. Particularly when moving packed
6695 data between functions compiled with GCC and the native Microsoft compiler
6696 (either via function call or as data in a file), it may be necessary to access
6699 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6700 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6701 command-line options, respectively;
6702 see @ref{x86 Options}, for details of how structure layout is affected.
6703 @xref{x86 Variable Attributes}, for information about the corresponding
6704 attributes on variables.
6708 @node Label Attributes
6709 @section Label Attributes
6710 @cindex Label Attributes
6712 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6713 details of the exact syntax for using attributes. Other attributes are
6714 available for functions (@pxref{Function Attributes}), variables
6715 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6716 and for types (@pxref{Type Attributes}).
6718 This example uses the @code{cold} label attribute to indicate the
6719 @code{ErrorHandling} branch is unlikely to be taken and that the
6720 @code{ErrorHandling} label is unused:
6724 asm goto ("some asm" : : : : NoError);
6726 /* This branch (the fall-through from the asm) is less commonly used */
6728 __attribute__((cold, unused)); /* Semi-colon is required here */
6733 printf("no error\n");
6739 @cindex @code{unused} label attribute
6740 This feature is intended for program-generated code that may contain
6741 unused labels, but which is compiled with @option{-Wall}. It is
6742 not normally appropriate to use in it human-written code, though it
6743 could be useful in cases where the code that jumps to the label is
6744 contained within an @code{#ifdef} conditional.
6747 @cindex @code{hot} label attribute
6748 The @code{hot} attribute on a label is used to inform the compiler that
6749 the path following the label is more likely than paths that are not so
6750 annotated. This attribute is used in cases where @code{__builtin_expect}
6751 cannot be used, for instance with computed goto or @code{asm goto}.
6754 @cindex @code{cold} label attribute
6755 The @code{cold} attribute on labels is used to inform the compiler that
6756 the path following the label is unlikely to be executed. This attribute
6757 is used in cases where @code{__builtin_expect} cannot be used, for instance
6758 with computed goto or @code{asm goto}.
6762 @node Enumerator Attributes
6763 @section Enumerator Attributes
6764 @cindex Enumerator Attributes
6766 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6767 details of the exact syntax for using attributes. Other attributes are
6768 available for functions (@pxref{Function Attributes}), variables
6769 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6770 and for types (@pxref{Type Attributes}).
6772 This example uses the @code{deprecated} enumerator attribute to indicate the
6773 @code{oldval} enumerator is deprecated:
6777 oldval __attribute__((deprecated)),
6790 @cindex @code{deprecated} enumerator attribute
6791 The @code{deprecated} attribute results in a warning if the enumerator
6792 is used anywhere in the source file. This is useful when identifying
6793 enumerators that are expected to be removed in a future version of a
6794 program. The warning also includes the location of the declaration
6795 of the deprecated enumerator, to enable users to easily find further
6796 information about why the enumerator is deprecated, or what they should
6797 do instead. Note that the warnings only occurs for uses.
6801 @node Attribute Syntax
6802 @section Attribute Syntax
6803 @cindex attribute syntax
6805 This section describes the syntax with which @code{__attribute__} may be
6806 used, and the constructs to which attribute specifiers bind, for the C
6807 language. Some details may vary for C++ and Objective-C@. Because of
6808 infelicities in the grammar for attributes, some forms described here
6809 may not be successfully parsed in all cases.
6811 There are some problems with the semantics of attributes in C++. For
6812 example, there are no manglings for attributes, although they may affect
6813 code generation, so problems may arise when attributed types are used in
6814 conjunction with templates or overloading. Similarly, @code{typeid}
6815 does not distinguish between types with different attributes. Support
6816 for attributes in C++ may be restricted in future to attributes on
6817 declarations only, but not on nested declarators.
6819 @xref{Function Attributes}, for details of the semantics of attributes
6820 applying to functions. @xref{Variable Attributes}, for details of the
6821 semantics of attributes applying to variables. @xref{Type Attributes},
6822 for details of the semantics of attributes applying to structure, union
6823 and enumerated types.
6824 @xref{Label Attributes}, for details of the semantics of attributes
6826 @xref{Enumerator Attributes}, for details of the semantics of attributes
6827 applying to enumerators.
6829 An @dfn{attribute specifier} is of the form
6830 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6831 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6832 each attribute is one of the following:
6836 Empty. Empty attributes are ignored.
6840 (which may be an identifier such as @code{unused}, or a reserved
6841 word such as @code{const}).
6844 An attribute name followed by a parenthesized list of
6845 parameters for the attribute.
6846 These parameters take one of the following forms:
6850 An identifier. For example, @code{mode} attributes use this form.
6853 An identifier followed by a comma and a non-empty comma-separated list
6854 of expressions. For example, @code{format} attributes use this form.
6857 A possibly empty comma-separated list of expressions. For example,
6858 @code{format_arg} attributes use this form with the list being a single
6859 integer constant expression, and @code{alias} attributes use this form
6860 with the list being a single string constant.
6864 An @dfn{attribute specifier list} is a sequence of one or more attribute
6865 specifiers, not separated by any other tokens.
6867 You may optionally specify attribute names with @samp{__}
6868 preceding and following the name.
6869 This allows you to use them in header files without
6870 being concerned about a possible macro of the same name. For example,
6871 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6874 @subsubheading Label Attributes
6876 In GNU C, an attribute specifier list may appear after the colon following a
6877 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6878 attributes on labels if the attribute specifier is immediately
6879 followed by a semicolon (i.e., the label applies to an empty
6880 statement). If the semicolon is missing, C++ label attributes are
6881 ambiguous, as it is permissible for a declaration, which could begin
6882 with an attribute list, to be labelled in C++. Declarations cannot be
6883 labelled in C90 or C99, so the ambiguity does not arise there.
6885 @subsubheading Enumerator Attributes
6887 In GNU C, an attribute specifier list may appear as part of an enumerator.
6888 The attribute goes after the enumeration constant, before @code{=}, if
6889 present. The optional attribute in the enumerator appertains to the
6890 enumeration constant. It is not possible to place the attribute after
6891 the constant expression, if present.
6893 @subsubheading Type Attributes
6895 An attribute specifier list may appear as part of a @code{struct},
6896 @code{union} or @code{enum} specifier. It may go either immediately
6897 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6898 the closing brace. The former syntax is preferred.
6899 Where attribute specifiers follow the closing brace, they are considered
6900 to relate to the structure, union or enumerated type defined, not to any
6901 enclosing declaration the type specifier appears in, and the type
6902 defined is not complete until after the attribute specifiers.
6903 @c Otherwise, there would be the following problems: a shift/reduce
6904 @c conflict between attributes binding the struct/union/enum and
6905 @c binding to the list of specifiers/qualifiers; and "aligned"
6906 @c attributes could use sizeof for the structure, but the size could be
6907 @c changed later by "packed" attributes.
6910 @subsubheading All other attributes
6912 Otherwise, an attribute specifier appears as part of a declaration,
6913 counting declarations of unnamed parameters and type names, and relates
6914 to that declaration (which may be nested in another declaration, for
6915 example in the case of a parameter declaration), or to a particular declarator
6916 within a declaration. Where an
6917 attribute specifier is applied to a parameter declared as a function or
6918 an array, it should apply to the function or array rather than the
6919 pointer to which the parameter is implicitly converted, but this is not
6920 yet correctly implemented.
6922 Any list of specifiers and qualifiers at the start of a declaration may
6923 contain attribute specifiers, whether or not such a list may in that
6924 context contain storage class specifiers. (Some attributes, however,
6925 are essentially in the nature of storage class specifiers, and only make
6926 sense where storage class specifiers may be used; for example,
6927 @code{section}.) There is one necessary limitation to this syntax: the
6928 first old-style parameter declaration in a function definition cannot
6929 begin with an attribute specifier, because such an attribute applies to
6930 the function instead by syntax described below (which, however, is not
6931 yet implemented in this case). In some other cases, attribute
6932 specifiers are permitted by this grammar but not yet supported by the
6933 compiler. All attribute specifiers in this place relate to the
6934 declaration as a whole. In the obsolescent usage where a type of
6935 @code{int} is implied by the absence of type specifiers, such a list of
6936 specifiers and qualifiers may be an attribute specifier list with no
6937 other specifiers or qualifiers.
6939 At present, the first parameter in a function prototype must have some
6940 type specifier that is not an attribute specifier; this resolves an
6941 ambiguity in the interpretation of @code{void f(int
6942 (__attribute__((foo)) x))}, but is subject to change. At present, if
6943 the parentheses of a function declarator contain only attributes then
6944 those attributes are ignored, rather than yielding an error or warning
6945 or implying a single parameter of type int, but this is subject to
6948 An attribute specifier list may appear immediately before a declarator
6949 (other than the first) in a comma-separated list of declarators in a
6950 declaration of more than one identifier using a single list of
6951 specifiers and qualifiers. Such attribute specifiers apply
6952 only to the identifier before whose declarator they appear. For
6956 __attribute__((noreturn)) void d0 (void),
6957 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6962 the @code{noreturn} attribute applies to all the functions
6963 declared; the @code{format} attribute only applies to @code{d1}.
6965 An attribute specifier list may appear immediately before the comma,
6966 @code{=} or semicolon terminating the declaration of an identifier other
6967 than a function definition. Such attribute specifiers apply
6968 to the declared object or function. Where an
6969 assembler name for an object or function is specified (@pxref{Asm
6970 Labels}), the attribute must follow the @code{asm}
6973 An attribute specifier list may, in future, be permitted to appear after
6974 the declarator in a function definition (before any old-style parameter
6975 declarations or the function body).
6977 Attribute specifiers may be mixed with type qualifiers appearing inside
6978 the @code{[]} of a parameter array declarator, in the C99 construct by
6979 which such qualifiers are applied to the pointer to which the array is
6980 implicitly converted. Such attribute specifiers apply to the pointer,
6981 not to the array, but at present this is not implemented and they are
6984 An attribute specifier list may appear at the start of a nested
6985 declarator. At present, there are some limitations in this usage: the
6986 attributes correctly apply to the declarator, but for most individual
6987 attributes the semantics this implies are not implemented.
6988 When attribute specifiers follow the @code{*} of a pointer
6989 declarator, they may be mixed with any type qualifiers present.
6990 The following describes the formal semantics of this syntax. It makes the
6991 most sense if you are familiar with the formal specification of
6992 declarators in the ISO C standard.
6994 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
6995 D1}, where @code{T} contains declaration specifiers that specify a type
6996 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
6997 contains an identifier @var{ident}. The type specified for @var{ident}
6998 for derived declarators whose type does not include an attribute
6999 specifier is as in the ISO C standard.
7001 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7002 and the declaration @code{T D} specifies the type
7003 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7004 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7005 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7007 If @code{D1} has the form @code{*
7008 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7009 declaration @code{T D} specifies the type
7010 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7011 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7012 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7018 void (__attribute__((noreturn)) ****f) (void);
7022 specifies the type ``pointer to pointer to pointer to pointer to
7023 non-returning function returning @code{void}''. As another example,
7026 char *__attribute__((aligned(8))) *f;
7030 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7031 Note again that this does not work with most attributes; for example,
7032 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7033 is not yet supported.
7035 For compatibility with existing code written for compiler versions that
7036 did not implement attributes on nested declarators, some laxity is
7037 allowed in the placing of attributes. If an attribute that only applies
7038 to types is applied to a declaration, it is treated as applying to
7039 the type of that declaration. If an attribute that only applies to
7040 declarations is applied to the type of a declaration, it is treated
7041 as applying to that declaration; and, for compatibility with code
7042 placing the attributes immediately before the identifier declared, such
7043 an attribute applied to a function return type is treated as
7044 applying to the function type, and such an attribute applied to an array
7045 element type is treated as applying to the array type. If an
7046 attribute that only applies to function types is applied to a
7047 pointer-to-function type, it is treated as applying to the pointer
7048 target type; if such an attribute is applied to a function return type
7049 that is not a pointer-to-function type, it is treated as applying
7050 to the function type.
7052 @node Function Prototypes
7053 @section Prototypes and Old-Style Function Definitions
7054 @cindex function prototype declarations
7055 @cindex old-style function definitions
7056 @cindex promotion of formal parameters
7058 GNU C extends ISO C to allow a function prototype to override a later
7059 old-style non-prototype definition. Consider the following example:
7062 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7069 /* @r{Prototype function declaration.} */
7070 int isroot P((uid_t));
7072 /* @r{Old-style function definition.} */
7074 isroot (x) /* @r{??? lossage here ???} */
7081 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7082 not allow this example, because subword arguments in old-style
7083 non-prototype definitions are promoted. Therefore in this example the
7084 function definition's argument is really an @code{int}, which does not
7085 match the prototype argument type of @code{short}.
7087 This restriction of ISO C makes it hard to write code that is portable
7088 to traditional C compilers, because the programmer does not know
7089 whether the @code{uid_t} type is @code{short}, @code{int}, or
7090 @code{long}. Therefore, in cases like these GNU C allows a prototype
7091 to override a later old-style definition. More precisely, in GNU C, a
7092 function prototype argument type overrides the argument type specified
7093 by a later old-style definition if the former type is the same as the
7094 latter type before promotion. Thus in GNU C the above example is
7095 equivalent to the following:
7108 GNU C++ does not support old-style function definitions, so this
7109 extension is irrelevant.
7112 @section C++ Style Comments
7114 @cindex C++ comments
7115 @cindex comments, C++ style
7117 In GNU C, you may use C++ style comments, which start with @samp{//} and
7118 continue until the end of the line. Many other C implementations allow
7119 such comments, and they are included in the 1999 C standard. However,
7120 C++ style comments are not recognized if you specify an @option{-std}
7121 option specifying a version of ISO C before C99, or @option{-ansi}
7122 (equivalent to @option{-std=c90}).
7125 @section Dollar Signs in Identifier Names
7127 @cindex dollar signs in identifier names
7128 @cindex identifier names, dollar signs in
7130 In GNU C, you may normally use dollar signs in identifier names.
7131 This is because many traditional C implementations allow such identifiers.
7132 However, dollar signs in identifiers are not supported on a few target
7133 machines, typically because the target assembler does not allow them.
7135 @node Character Escapes
7136 @section The Character @key{ESC} in Constants
7138 You can use the sequence @samp{\e} in a string or character constant to
7139 stand for the ASCII character @key{ESC}.
7142 @section Inquiring on Alignment of Types or Variables
7144 @cindex type alignment
7145 @cindex variable alignment
7147 The keyword @code{__alignof__} allows you to inquire about how an object
7148 is aligned, or the minimum alignment usually required by a type. Its
7149 syntax is just like @code{sizeof}.
7151 For example, if the target machine requires a @code{double} value to be
7152 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7153 This is true on many RISC machines. On more traditional machine
7154 designs, @code{__alignof__ (double)} is 4 or even 2.
7156 Some machines never actually require alignment; they allow reference to any
7157 data type even at an odd address. For these machines, @code{__alignof__}
7158 reports the smallest alignment that GCC gives the data type, usually as
7159 mandated by the target ABI.
7161 If the operand of @code{__alignof__} is an lvalue rather than a type,
7162 its value is the required alignment for its type, taking into account
7163 any minimum alignment specified with GCC's @code{__attribute__}
7164 extension (@pxref{Variable Attributes}). For example, after this
7168 struct foo @{ int x; char y; @} foo1;
7172 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7173 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7175 It is an error to ask for the alignment of an incomplete type.
7179 @section An Inline Function is As Fast As a Macro
7180 @cindex inline functions
7181 @cindex integrating function code
7183 @cindex macros, inline alternative
7185 By declaring a function inline, you can direct GCC to make
7186 calls to that function faster. One way GCC can achieve this is to
7187 integrate that function's code into the code for its callers. This
7188 makes execution faster by eliminating the function-call overhead; in
7189 addition, if any of the actual argument values are constant, their
7190 known values may permit simplifications at compile time so that not
7191 all of the inline function's code needs to be included. The effect on
7192 code size is less predictable; object code may be larger or smaller
7193 with function inlining, depending on the particular case. You can
7194 also direct GCC to try to integrate all ``simple enough'' functions
7195 into their callers with the option @option{-finline-functions}.
7197 GCC implements three different semantics of declaring a function
7198 inline. One is available with @option{-std=gnu89} or
7199 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7200 on all inline declarations, another when
7201 @option{-std=c99}, @option{-std=c11},
7202 @option{-std=gnu99} or @option{-std=gnu11}
7203 (without @option{-fgnu89-inline}), and the third
7204 is used when compiling C++.
7206 To declare a function inline, use the @code{inline} keyword in its
7207 declaration, like this:
7217 If you are writing a header file to be included in ISO C90 programs, write
7218 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7220 The three types of inlining behave similarly in two important cases:
7221 when the @code{inline} keyword is used on a @code{static} function,
7222 like the example above, and when a function is first declared without
7223 using the @code{inline} keyword and then is defined with
7224 @code{inline}, like this:
7227 extern int inc (int *a);
7235 In both of these common cases, the program behaves the same as if you
7236 had not used the @code{inline} keyword, except for its speed.
7238 @cindex inline functions, omission of
7239 @opindex fkeep-inline-functions
7240 When a function is both inline and @code{static}, if all calls to the
7241 function are integrated into the caller, and the function's address is
7242 never used, then the function's own assembler code is never referenced.
7243 In this case, GCC does not actually output assembler code for the
7244 function, unless you specify the option @option{-fkeep-inline-functions}.
7245 If there is a nonintegrated call, then the function is compiled to
7246 assembler code as usual. The function must also be compiled as usual if
7247 the program refers to its address, because that can't be inlined.
7250 Note that certain usages in a function definition can make it unsuitable
7251 for inline substitution. Among these usages are: variadic functions,
7252 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7253 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7254 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7255 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7256 function marked @code{inline} could not be substituted, and gives the
7257 reason for the failure.
7259 @cindex automatic @code{inline} for C++ member fns
7260 @cindex @code{inline} automatic for C++ member fns
7261 @cindex member fns, automatically @code{inline}
7262 @cindex C++ member fns, automatically @code{inline}
7263 @opindex fno-default-inline
7264 As required by ISO C++, GCC considers member functions defined within
7265 the body of a class to be marked inline even if they are
7266 not explicitly declared with the @code{inline} keyword. You can
7267 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7268 Options,,Options Controlling C++ Dialect}.
7270 GCC does not inline any functions when not optimizing unless you specify
7271 the @samp{always_inline} attribute for the function, like this:
7274 /* @r{Prototype.} */
7275 inline void foo (const char) __attribute__((always_inline));
7278 The remainder of this section is specific to GNU C90 inlining.
7280 @cindex non-static inline function
7281 When an inline function is not @code{static}, then the compiler must assume
7282 that there may be calls from other source files; since a global symbol can
7283 be defined only once in any program, the function must not be defined in
7284 the other source files, so the calls therein cannot be integrated.
7285 Therefore, a non-@code{static} inline function is always compiled on its
7286 own in the usual fashion.
7288 If you specify both @code{inline} and @code{extern} in the function
7289 definition, then the definition is used only for inlining. In no case
7290 is the function compiled on its own, not even if you refer to its
7291 address explicitly. Such an address becomes an external reference, as
7292 if you had only declared the function, and had not defined it.
7294 This combination of @code{inline} and @code{extern} has almost the
7295 effect of a macro. The way to use it is to put a function definition in
7296 a header file with these keywords, and put another copy of the
7297 definition (lacking @code{inline} and @code{extern}) in a library file.
7298 The definition in the header file causes most calls to the function
7299 to be inlined. If any uses of the function remain, they refer to
7300 the single copy in the library.
7303 @section When is a Volatile Object Accessed?
7304 @cindex accessing volatiles
7305 @cindex volatile read
7306 @cindex volatile write
7307 @cindex volatile access
7309 C has the concept of volatile objects. These are normally accessed by
7310 pointers and used for accessing hardware or inter-thread
7311 communication. The standard encourages compilers to refrain from
7312 optimizations concerning accesses to volatile objects, but leaves it
7313 implementation defined as to what constitutes a volatile access. The
7314 minimum requirement is that at a sequence point all previous accesses
7315 to volatile objects have stabilized and no subsequent accesses have
7316 occurred. Thus an implementation is free to reorder and combine
7317 volatile accesses that occur between sequence points, but cannot do
7318 so for accesses across a sequence point. The use of volatile does
7319 not allow you to violate the restriction on updating objects multiple
7320 times between two sequence points.
7322 Accesses to non-volatile objects are not ordered with respect to
7323 volatile accesses. You cannot use a volatile object as a memory
7324 barrier to order a sequence of writes to non-volatile memory. For
7328 int *ptr = @var{something};
7330 *ptr = @var{something};
7335 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7336 that the write to @var{*ptr} occurs by the time the update
7337 of @var{vobj} happens. If you need this guarantee, you must use
7338 a stronger memory barrier such as:
7341 int *ptr = @var{something};
7343 *ptr = @var{something};
7344 asm volatile ("" : : : "memory");
7348 A scalar volatile object is read when it is accessed in a void context:
7351 volatile int *src = @var{somevalue};
7355 Such expressions are rvalues, and GCC implements this as a
7356 read of the volatile object being pointed to.
7358 Assignments are also expressions and have an rvalue. However when
7359 assigning to a scalar volatile, the volatile object is not reread,
7360 regardless of whether the assignment expression's rvalue is used or
7361 not. If the assignment's rvalue is used, the value is that assigned
7362 to the volatile object. For instance, there is no read of @var{vobj}
7363 in all the following cases:
7368 vobj = @var{something};
7369 obj = vobj = @var{something};
7370 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7371 obj = (@var{something}, vobj = @var{anotherthing});
7374 If you need to read the volatile object after an assignment has
7375 occurred, you must use a separate expression with an intervening
7378 As bit-fields are not individually addressable, volatile bit-fields may
7379 be implicitly read when written to, or when adjacent bit-fields are
7380 accessed. Bit-field operations may be optimized such that adjacent
7381 bit-fields are only partially accessed, if they straddle a storage unit
7382 boundary. For these reasons it is unwise to use volatile bit-fields to
7385 @node Using Assembly Language with C
7386 @section How to Use Inline Assembly Language in C Code
7387 @cindex @code{asm} keyword
7388 @cindex assembly language in C
7389 @cindex inline assembly language
7390 @cindex mixing assembly language and C
7392 The @code{asm} keyword allows you to embed assembler instructions
7393 within C code. GCC provides two forms of inline @code{asm}
7394 statements. A @dfn{basic @code{asm}} statement is one with no
7395 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7396 statement (@pxref{Extended Asm}) includes one or more operands.
7397 The extended form is preferred for mixing C and assembly language
7398 within a function, but to include assembly language at
7399 top level you must use basic @code{asm}.
7401 You can also use the @code{asm} keyword to override the assembler name
7402 for a C symbol, or to place a C variable in a specific register.
7405 * Basic Asm:: Inline assembler without operands.
7406 * Extended Asm:: Inline assembler with operands.
7407 * Constraints:: Constraints for @code{asm} operands
7408 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7409 * Explicit Register Variables:: Defining variables residing in specified
7411 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7415 @subsection Basic Asm --- Assembler Instructions Without Operands
7416 @cindex basic @code{asm}
7417 @cindex assembly language in C, basic
7419 A basic @code{asm} statement has the following syntax:
7422 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7425 The @code{asm} keyword is a GNU extension.
7426 When writing code that can be compiled with @option{-ansi} and the
7427 various @option{-std} options, use @code{__asm__} instead of
7428 @code{asm} (@pxref{Alternate Keywords}).
7430 @subsubheading Qualifiers
7433 The optional @code{volatile} qualifier has no effect.
7434 All basic @code{asm} blocks are implicitly volatile.
7437 @subsubheading Parameters
7440 @item AssemblerInstructions
7441 This is a literal string that specifies the assembler code. The string can
7442 contain any instructions recognized by the assembler, including directives.
7443 GCC does not parse the assembler instructions themselves and
7444 does not know what they mean or even whether they are valid assembler input.
7446 You may place multiple assembler instructions together in a single @code{asm}
7447 string, separated by the characters normally used in assembly code for the
7448 system. A combination that works in most places is a newline to break the
7449 line, plus a tab character (written as @samp{\n\t}).
7450 Some assemblers allow semicolons as a line separator. However,
7451 note that some assembler dialects use semicolons to start a comment.
7454 @subsubheading Remarks
7455 Using extended @code{asm} typically produces smaller, safer, and more
7456 efficient code, and in most cases it is a better solution than basic
7457 @code{asm}. However, there are two situations where only basic @code{asm}
7462 Extended @code{asm} statements have to be inside a C
7463 function, so to write inline assembly language at file scope (``top-level''),
7464 outside of C functions, you must use basic @code{asm}.
7465 You can use this technique to emit assembler directives,
7466 define assembly language macros that can be invoked elsewhere in the file,
7467 or write entire functions in assembly language.
7471 with the @code{naked} attribute also require basic @code{asm}
7472 (@pxref{Function Attributes}).
7475 Safely accessing C data and calling functions from basic @code{asm} is more
7476 complex than it may appear. To access C data, it is better to use extended
7479 Do not expect a sequence of @code{asm} statements to remain perfectly
7480 consecutive after compilation. If certain instructions need to remain
7481 consecutive in the output, put them in a single multi-instruction @code{asm}
7482 statement. Note that GCC's optimizers can move @code{asm} statements
7483 relative to other code, including across jumps.
7485 @code{asm} statements may not perform jumps into other @code{asm} statements.
7486 GCC does not know about these jumps, and therefore cannot take
7487 account of them when deciding how to optimize. Jumps from @code{asm} to C
7488 labels are only supported in extended @code{asm}.
7490 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7491 assembly code when optimizing. This can lead to unexpected duplicate
7492 symbol errors during compilation if your assembly code defines symbols or
7495 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7496 visibility of any symbols it references. This may result in GCC discarding
7497 those symbols as unreferenced.
7499 The compiler copies the assembler instructions in a basic @code{asm}
7500 verbatim to the assembly language output file, without
7501 processing dialects or any of the @samp{%} operators that are available with
7502 extended @code{asm}. This results in minor differences between basic
7503 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7504 registers you might use @samp{%eax} in basic @code{asm} and
7505 @samp{%%eax} in extended @code{asm}.
7507 On targets such as x86 that support multiple assembler dialects,
7508 all basic @code{asm} blocks use the assembler dialect specified by the
7509 @option{-masm} command-line option (@pxref{x86 Options}).
7510 Basic @code{asm} provides no
7511 mechanism to provide different assembler strings for different dialects.
7513 Here is an example of basic @code{asm} for i386:
7516 /* Note that this code will not compile with -masm=intel */
7517 #define DebugBreak() asm("int $3")
7521 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7522 @cindex extended @code{asm}
7523 @cindex assembly language in C, extended
7525 With extended @code{asm} you can read and write C variables from
7526 assembler and perform jumps from assembler code to C labels.
7527 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7528 the operand parameters after the assembler template:
7531 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7532 : @var{OutputOperands}
7533 @r{[} : @var{InputOperands}
7534 @r{[} : @var{Clobbers} @r{]} @r{]})
7536 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7538 : @var{InputOperands}
7543 The @code{asm} keyword is a GNU extension.
7544 When writing code that can be compiled with @option{-ansi} and the
7545 various @option{-std} options, use @code{__asm__} instead of
7546 @code{asm} (@pxref{Alternate Keywords}).
7548 @subsubheading Qualifiers
7552 The typical use of extended @code{asm} statements is to manipulate input
7553 values to produce output values. However, your @code{asm} statements may
7554 also produce side effects. If so, you may need to use the @code{volatile}
7555 qualifier to disable certain optimizations. @xref{Volatile}.
7558 This qualifier informs the compiler that the @code{asm} statement may
7559 perform a jump to one of the labels listed in the @var{GotoLabels}.
7563 @subsubheading Parameters
7565 @item AssemblerTemplate
7566 This is a literal string that is the template for the assembler code. It is a
7567 combination of fixed text and tokens that refer to the input, output,
7568 and goto parameters. @xref{AssemblerTemplate}.
7570 @item OutputOperands
7571 A comma-separated list of the C variables modified by the instructions in the
7572 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7575 A comma-separated list of C expressions read by the instructions in the
7576 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7579 A comma-separated list of registers or other values changed by the
7580 @var{AssemblerTemplate}, beyond those listed as outputs.
7581 An empty list is permitted. @xref{Clobbers}.
7584 When you are using the @code{goto} form of @code{asm}, this section contains
7585 the list of all C labels to which the code in the
7586 @var{AssemblerTemplate} may jump.
7589 @code{asm} statements may not perform jumps into other @code{asm} statements,
7590 only to the listed @var{GotoLabels}.
7591 GCC's optimizers do not know about other jumps; therefore they cannot take
7592 account of them when deciding how to optimize.
7595 The total number of input + output + goto operands is limited to 30.
7597 @subsubheading Remarks
7598 The @code{asm} statement allows you to include assembly instructions directly
7599 within C code. This may help you to maximize performance in time-sensitive
7600 code or to access assembly instructions that are not readily available to C
7603 Note that extended @code{asm} statements must be inside a function. Only
7604 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7605 Functions declared with the @code{naked} attribute also require basic
7606 @code{asm} (@pxref{Function Attributes}).
7608 While the uses of @code{asm} are many and varied, it may help to think of an
7609 @code{asm} statement as a series of low-level instructions that convert input
7610 parameters to output parameters. So a simple (if not particularly useful)
7611 example for i386 using @code{asm} might look like this:
7617 asm ("mov %1, %0\n\t"
7622 printf("%d\n", dst);
7625 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7628 @subsubsection Volatile
7629 @cindex volatile @code{asm}
7630 @cindex @code{asm} volatile
7632 GCC's optimizers sometimes discard @code{asm} statements if they determine
7633 there is no need for the output variables. Also, the optimizers may move
7634 code out of loops if they believe that the code will always return the same
7635 result (i.e. none of its input values change between calls). Using the
7636 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7637 that have no output operands, including @code{asm goto} statements,
7638 are implicitly volatile.
7640 This i386 code demonstrates a case that does not use (or require) the
7641 @code{volatile} qualifier. If it is performing assertion checking, this code
7642 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7643 unreferenced by any code. As a result, the optimizers can discard the
7644 @code{asm} statement, which in turn removes the need for the entire
7645 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7646 isn't needed you allow the optimizers to produce the most efficient code
7650 void DoCheck(uint32_t dwSomeValue)
7654 // Assumes dwSomeValue is not zero.
7664 The next example shows a case where the optimizers can recognize that the input
7665 (@code{dwSomeValue}) never changes during the execution of the function and can
7666 therefore move the @code{asm} outside the loop to produce more efficient code.
7667 Again, using @code{volatile} disables this type of optimization.
7670 void do_print(uint32_t dwSomeValue)
7674 for (uint32_t x=0; x < 5; x++)
7676 // Assumes dwSomeValue is not zero.
7682 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7687 The following example demonstrates a case where you need to use the
7688 @code{volatile} qualifier.
7689 It uses the x86 @code{rdtsc} instruction, which reads
7690 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7691 the optimizers might assume that the @code{asm} block will always return the
7692 same value and therefore optimize away the second call.
7697 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7698 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7699 "or %%rdx, %0" // 'Or' in the lower bits.
7704 printf("msr: %llx\n", msr);
7708 // Reprint the timestamp
7709 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7710 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7711 "or %%rdx, %0" // 'Or' in the lower bits.
7716 printf("msr: %llx\n", msr);
7719 GCC's optimizers do not treat this code like the non-volatile code in the
7720 earlier examples. They do not move it out of loops or omit it on the
7721 assumption that the result from a previous call is still valid.
7723 Note that the compiler can move even volatile @code{asm} instructions relative
7724 to other code, including across jump instructions. For example, on many
7725 targets there is a system register that controls the rounding mode of
7726 floating-point operations. Setting it with a volatile @code{asm}, as in the
7727 following PowerPC example, does not work reliably.
7730 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7734 The compiler may move the addition back before the volatile @code{asm}. To
7735 make it work as expected, add an artificial dependency to the @code{asm} by
7736 referencing a variable in the subsequent code, for example:
7739 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7743 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7744 assembly code when optimizing. This can lead to unexpected duplicate symbol
7745 errors during compilation if your asm code defines symbols or labels.
7747 (@pxref{AssemblerTemplate}) may help resolve this problem.
7749 @anchor{AssemblerTemplate}
7750 @subsubsection Assembler Template
7751 @cindex @code{asm} assembler template
7753 An assembler template is a literal string containing assembler instructions.
7754 The compiler replaces tokens in the template that refer
7755 to inputs, outputs, and goto labels,
7756 and then outputs the resulting string to the assembler. The
7757 string can contain any instructions recognized by the assembler, including
7758 directives. GCC does not parse the assembler instructions
7759 themselves and does not know what they mean or even whether they are valid
7760 assembler input. However, it does count the statements
7761 (@pxref{Size of an asm}).
7763 You may place multiple assembler instructions together in a single @code{asm}
7764 string, separated by the characters normally used in assembly code for the
7765 system. A combination that works in most places is a newline to break the
7766 line, plus a tab character to move to the instruction field (written as
7768 Some assemblers allow semicolons as a line separator. However, note
7769 that some assembler dialects use semicolons to start a comment.
7771 Do not expect a sequence of @code{asm} statements to remain perfectly
7772 consecutive after compilation, even when you are using the @code{volatile}
7773 qualifier. If certain instructions need to remain consecutive in the output,
7774 put them in a single multi-instruction asm statement.
7776 Accessing data from C programs without using input/output operands (such as
7777 by using global symbols directly from the assembler template) may not work as
7778 expected. Similarly, calling functions directly from an assembler template
7779 requires a detailed understanding of the target assembler and ABI.
7781 Since GCC does not parse the assembler template,
7782 it has no visibility of any
7783 symbols it references. This may result in GCC discarding those symbols as
7784 unreferenced unless they are also listed as input, output, or goto operands.
7786 @subsubheading Special format strings
7788 In addition to the tokens described by the input, output, and goto operands,
7789 these tokens have special meanings in the assembler template:
7793 Outputs a single @samp{%} into the assembler code.
7796 Outputs a number that is unique to each instance of the @code{asm}
7797 statement in the entire compilation. This option is useful when creating local
7798 labels and referring to them multiple times in a single template that
7799 generates multiple assembler instructions.
7804 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7805 into the assembler code. When unescaped, these characters have special
7806 meaning to indicate multiple assembler dialects, as described below.
7809 @subsubheading Multiple assembler dialects in @code{asm} templates
7811 On targets such as x86, GCC supports multiple assembler dialects.
7812 The @option{-masm} option controls which dialect GCC uses as its
7813 default for inline assembler. The target-specific documentation for the
7814 @option{-masm} option contains the list of supported dialects, as well as the
7815 default dialect if the option is not specified. This information may be
7816 important to understand, since assembler code that works correctly when
7817 compiled using one dialect will likely fail if compiled using another.
7820 If your code needs to support multiple assembler dialects (for example, if
7821 you are writing public headers that need to support a variety of compilation
7822 options), use constructs of this form:
7825 @{ dialect0 | dialect1 | dialect2... @}
7828 This construct outputs @code{dialect0}
7829 when using dialect #0 to compile the code,
7830 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7831 braces than the number of dialects the compiler supports, the construct
7834 For example, if an x86 compiler supports two dialects
7835 (@samp{att}, @samp{intel}), an
7836 assembler template such as this:
7839 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7843 is equivalent to one of
7846 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7847 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7850 Using that same compiler, this code:
7853 "xchg@{l@}\t@{%%@}ebx, %1"
7857 corresponds to either
7860 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7861 "xchg\tebx, %1" @r{/* intel dialect */}
7864 There is no support for nesting dialect alternatives.
7866 @anchor{OutputOperands}
7867 @subsubsection Output Operands
7868 @cindex @code{asm} output operands
7870 An @code{asm} statement has zero or more output operands indicating the names
7871 of C variables modified by the assembler code.
7873 In this i386 example, @code{old} (referred to in the template string as
7874 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7875 (@code{%2}) is an input:
7880 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7881 "sbb %0,%0" // Use the CF to calculate old.
7882 : "=r" (old), "+rm" (*Base)
7889 Operands are separated by commas. Each operand has this format:
7892 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7896 @item asmSymbolicName
7897 Specifies a symbolic name for the operand.
7898 Reference the name in the assembler template
7899 by enclosing it in square brackets
7900 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7901 that contains the definition. Any valid C variable name is acceptable,
7902 including names already defined in the surrounding code. No two operands
7903 within the same @code{asm} statement can use the same symbolic name.
7905 When not using an @var{asmSymbolicName}, use the (zero-based) position
7907 in the list of operands in the assembler template. For example if there are
7908 three output operands, use @samp{%0} in the template to refer to the first,
7909 @samp{%1} for the second, and @samp{%2} for the third.
7912 A string constant specifying constraints on the placement of the operand;
7913 @xref{Constraints}, for details.
7915 Output constraints must begin with either @samp{=} (a variable overwriting an
7916 existing value) or @samp{+} (when reading and writing). When using
7917 @samp{=}, do not assume the location contains the existing value
7918 on entry to the @code{asm}, except
7919 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7921 After the prefix, there must be one or more additional constraints
7922 (@pxref{Constraints}) that describe where the value resides. Common
7923 constraints include @samp{r} for register and @samp{m} for memory.
7924 When you list more than one possible location (for example, @code{"=rm"}),
7925 the compiler chooses the most efficient one based on the current context.
7926 If you list as many alternates as the @code{asm} statement allows, you permit
7927 the optimizers to produce the best possible code.
7928 If you must use a specific register, but your Machine Constraints do not
7929 provide sufficient control to select the specific register you want,
7930 local register variables may provide a solution (@pxref{Local Register
7934 Specifies a C lvalue expression to hold the output, typically a variable name.
7935 The enclosing parentheses are a required part of the syntax.
7939 When the compiler selects the registers to use to
7940 represent the output operands, it does not use any of the clobbered registers
7943 Output operand expressions must be lvalues. The compiler cannot check whether
7944 the operands have data types that are reasonable for the instruction being
7945 executed. For output expressions that are not directly addressable (for
7946 example a bit-field), the constraint must allow a register. In that case, GCC
7947 uses the register as the output of the @code{asm}, and then stores that
7948 register into the output.
7950 Operands using the @samp{+} constraint modifier count as two operands
7951 (that is, both as input and output) towards the total maximum of 30 operands
7952 per @code{asm} statement.
7954 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7955 operands that must not overlap an input. Otherwise,
7956 GCC may allocate the output operand in the same register as an unrelated
7957 input operand, on the assumption that the assembler code consumes its
7958 inputs before producing outputs. This assumption may be false if the assembler
7959 code actually consists of more than one instruction.
7961 The same problem can occur if one output parameter (@var{a}) allows a register
7962 constraint and another output parameter (@var{b}) allows a memory constraint.
7963 The code generated by GCC to access the memory address in @var{b} can contain
7964 registers which @emph{might} be shared by @var{a}, and GCC considers those
7965 registers to be inputs to the asm. As above, GCC assumes that such input
7966 registers are consumed before any outputs are written. This assumption may
7967 result in incorrect behavior if the asm writes to @var{a} before using
7968 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7969 ensures that modifying @var{a} does not affect the address referenced by
7970 @var{b}. Otherwise, the location of @var{b}
7971 is undefined if @var{a} is modified before using @var{b}.
7973 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7974 instead of simply @samp{%2}). Typically these qualifiers are hardware
7975 dependent. The list of supported modifiers for x86 is found at
7976 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7978 If the C code that follows the @code{asm} makes no use of any of the output
7979 operands, use @code{volatile} for the @code{asm} statement to prevent the
7980 optimizers from discarding the @code{asm} statement as unneeded
7981 (see @ref{Volatile}).
7983 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7984 references the first output operand as @code{%0} (were there a second, it
7985 would be @code{%1}, etc). The number of the first input operand is one greater
7986 than that of the last output operand. In this i386 example, that makes
7987 @code{Mask} referenced as @code{%1}:
7990 uint32_t Mask = 1234;
7999 That code overwrites the variable @code{Index} (@samp{=}),
8000 placing the value in a register (@samp{r}).
8001 Using the generic @samp{r} constraint instead of a constraint for a specific
8002 register allows the compiler to pick the register to use, which can result
8003 in more efficient code. This may not be possible if an assembler instruction
8004 requires a specific register.
8006 The following i386 example uses the @var{asmSymbolicName} syntax.
8008 same result as the code above, but some may consider it more readable or more
8009 maintainable since reordering index numbers is not necessary when adding or
8010 removing operands. The names @code{aIndex} and @code{aMask}
8011 are only used in this example to emphasize which
8012 names get used where.
8013 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8016 uint32_t Mask = 1234;
8019 asm ("bsfl %[aMask], %[aIndex]"
8020 : [aIndex] "=r" (Index)
8021 : [aMask] "r" (Mask)
8025 Here are some more examples of output operands.
8032 asm ("mov %[e], %[d]"
8037 Here, @code{d} may either be in a register or in memory. Since the compiler
8038 might already have the current value of the @code{uint32_t} location
8039 pointed to by @code{e}
8040 in a register, you can enable it to choose the best location
8041 for @code{d} by specifying both constraints.
8043 @anchor{FlagOutputOperands}
8044 @subsubsection Flag Output Operands
8045 @cindex @code{asm} flag output operands
8047 Some targets have a special register that holds the ``flags'' for the
8048 result of an operation or comparison. Normally, the contents of that
8049 register are either unmodifed by the asm, or the asm is considered to
8050 clobber the contents.
8052 On some targets, a special form of output operand exists by which
8053 conditions in the flags register may be outputs of the asm. The set of
8054 conditions supported are target specific, but the general rule is that
8055 the output variable must be a scalar integer, and the value is boolean.
8056 When supported, the target defines the preprocessor symbol
8057 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8059 Because of the special nature of the flag output operands, the constraint
8060 may not include alternatives.
8062 Most often, the target has only one flags register, and thus is an implied
8063 operand of many instructions. In this case, the operand should not be
8064 referenced within the assembler template via @code{%0} etc, as there's
8065 no corresponding text in the assembly language.
8069 The flag output constraints for the x86 family are of the form
8070 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8071 conditions defined in the ISA manual for @code{j@var{cc}} or
8076 ``above'' or unsigned greater than
8078 ``above or equal'' or unsigned greater than or equal
8080 ``below'' or unsigned less than
8082 ``below or equal'' or unsigned less than or equal
8087 ``equal'' or zero flag set
8091 signed greater than or equal
8095 signed less than or equal
8116 ``not'' @var{flag}, or inverted versions of those above
8121 @anchor{InputOperands}
8122 @subsubsection Input Operands
8123 @cindex @code{asm} input operands
8124 @cindex @code{asm} expressions
8126 Input operands make values from C variables and expressions available to the
8129 Operands are separated by commas. Each operand has this format:
8132 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8136 @item asmSymbolicName
8137 Specifies a symbolic name for the operand.
8138 Reference the name in the assembler template
8139 by enclosing it in square brackets
8140 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8141 that contains the definition. Any valid C variable name is acceptable,
8142 including names already defined in the surrounding code. No two operands
8143 within the same @code{asm} statement can use the same symbolic name.
8145 When not using an @var{asmSymbolicName}, use the (zero-based) position
8147 in the list of operands in the assembler template. For example if there are
8148 two output operands and three inputs,
8149 use @samp{%2} in the template to refer to the first input operand,
8150 @samp{%3} for the second, and @samp{%4} for the third.
8153 A string constant specifying constraints on the placement of the operand;
8154 @xref{Constraints}, for details.
8156 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8157 When you list more than one possible location (for example, @samp{"irm"}),
8158 the compiler chooses the most efficient one based on the current context.
8159 If you must use a specific register, but your Machine Constraints do not
8160 provide sufficient control to select the specific register you want,
8161 local register variables may provide a solution (@pxref{Local Register
8164 Input constraints can also be digits (for example, @code{"0"}). This indicates
8165 that the specified input must be in the same place as the output constraint
8166 at the (zero-based) index in the output constraint list.
8167 When using @var{asmSymbolicName} syntax for the output operands,
8168 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8171 This is the C variable or expression being passed to the @code{asm} statement
8172 as input. The enclosing parentheses are a required part of the syntax.
8176 When the compiler selects the registers to use to represent the input
8177 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8179 If there are no output operands but there are input operands, place two
8180 consecutive colons where the output operands would go:
8183 __asm__ ("some instructions"
8185 : "r" (Offset / 8));
8188 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8189 (except for inputs tied to outputs). The compiler assumes that on exit from
8190 the @code{asm} statement these operands contain the same values as they
8191 had before executing the statement.
8192 It is @emph{not} possible to use clobbers
8193 to inform the compiler that the values in these inputs are changing. One
8194 common work-around is to tie the changing input variable to an output variable
8195 that never gets used. Note, however, that if the code that follows the
8196 @code{asm} statement makes no use of any of the output operands, the GCC
8197 optimizers may discard the @code{asm} statement as unneeded
8198 (see @ref{Volatile}).
8200 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8201 instead of simply @samp{%2}). Typically these qualifiers are hardware
8202 dependent. The list of supported modifiers for x86 is found at
8203 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8205 In this example using the fictitious @code{combine} instruction, the
8206 constraint @code{"0"} for input operand 1 says that it must occupy the same
8207 location as output operand 0. Only input operands may use numbers in
8208 constraints, and they must each refer to an output operand. Only a number (or
8209 the symbolic assembler name) in the constraint can guarantee that one operand
8210 is in the same place as another. The mere fact that @code{foo} is the value of
8211 both operands is not enough to guarantee that they are in the same place in
8212 the generated assembler code.
8215 asm ("combine %2, %0"
8217 : "0" (foo), "g" (bar));
8220 Here is an example using symbolic names.
8223 asm ("cmoveq %1, %2, %[result]"
8224 : [result] "=r"(result)
8225 : "r" (test), "r" (new), "[result]" (old));
8229 @subsubsection Clobbers
8230 @cindex @code{asm} clobbers
8232 While the compiler is aware of changes to entries listed in the output
8233 operands, the inline @code{asm} code may modify more than just the outputs. For
8234 example, calculations may require additional registers, or the processor may
8235 overwrite a register as a side effect of a particular assembler instruction.
8236 In order to inform the compiler of these changes, list them in the clobber
8237 list. Clobber list items are either register names or the special clobbers
8238 (listed below). Each clobber list item is a string constant
8239 enclosed in double quotes and separated by commas.
8241 Clobber descriptions may not in any way overlap with an input or output
8242 operand. For example, you may not have an operand describing a register class
8243 with one member when listing that register in the clobber list. Variables
8244 declared to live in specific registers (@pxref{Explicit Register
8245 Variables}) and used
8246 as @code{asm} input or output operands must have no part mentioned in the
8247 clobber description. In particular, there is no way to specify that input
8248 operands get modified without also specifying them as output operands.
8250 When the compiler selects which registers to use to represent input and output
8251 operands, it does not use any of the clobbered registers. As a result,
8252 clobbered registers are available for any use in the assembler code.
8254 Here is a realistic example for the VAX showing the use of clobbered
8258 asm volatile ("movc3 %0, %1, %2"
8260 : "g" (from), "g" (to), "g" (count)
8261 : "r0", "r1", "r2", "r3", "r4", "r5");
8264 Also, there are two special clobber arguments:
8268 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8269 register. On some machines, GCC represents the condition codes as a specific
8270 hardware register; @code{"cc"} serves to name this register.
8271 On other machines, condition code handling is different,
8272 and specifying @code{"cc"} has no effect. But
8273 it is valid no matter what the target.
8276 The @code{"memory"} clobber tells the compiler that the assembly code
8278 reads or writes to items other than those listed in the input and output
8279 operands (for example, accessing the memory pointed to by one of the input
8280 parameters). To ensure memory contains correct values, GCC may need to flush
8281 specific register values to memory before executing the @code{asm}. Further,
8282 the compiler does not assume that any values read from memory before an
8283 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8285 Using the @code{"memory"} clobber effectively forms a read/write
8286 memory barrier for the compiler.
8288 Note that this clobber does not prevent the @emph{processor} from doing
8289 speculative reads past the @code{asm} statement. To prevent that, you need
8290 processor-specific fence instructions.
8292 Flushing registers to memory has performance implications and may be an issue
8293 for time-sensitive code. You can use a trick to avoid this if the size of
8294 the memory being accessed is known at compile time. For example, if accessing
8295 ten bytes of a string, use a memory input like:
8297 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8302 @subsubsection Goto Labels
8303 @cindex @code{asm} goto labels
8305 @code{asm goto} allows assembly code to jump to one or more C labels. The
8306 @var{GotoLabels} section in an @code{asm goto} statement contains
8308 list of all C labels to which the assembler code may jump. GCC assumes that
8309 @code{asm} execution falls through to the next statement (if this is not the
8310 case, consider using the @code{__builtin_unreachable} intrinsic after the
8311 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8312 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8315 An @code{asm goto} statement cannot have outputs.
8316 This is due to an internal restriction of
8317 the compiler: control transfer instructions cannot have outputs.
8318 If the assembler code does modify anything, use the @code{"memory"} clobber
8320 optimizers to flush all register values to memory and reload them if
8321 necessary after the @code{asm} statement.
8323 Also note that an @code{asm goto} statement is always implicitly
8324 considered volatile.
8326 To reference a label in the assembler template,
8327 prefix it with @samp{%l} (lowercase @samp{L}) followed
8328 by its (zero-based) position in @var{GotoLabels} plus the number of input
8329 operands. For example, if the @code{asm} has three inputs and references two
8330 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8332 Alternately, you can reference labels using the actual C label name enclosed
8333 in brackets. For example, to reference a label named @code{carry}, you can
8334 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8335 section when using this approach.
8337 Here is an example of @code{asm goto} for i386:
8344 : "r" (p1), "r" (p2)
8354 The following example shows an @code{asm goto} that uses a memory clobber.
8360 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8371 @anchor{x86Operandmodifiers}
8372 @subsubsection x86 Operand Modifiers
8374 References to input, output, and goto operands in the assembler template
8375 of extended @code{asm} statements can use
8376 modifiers to affect the way the operands are formatted in
8377 the code output to the assembler. For example, the
8378 following code uses the @samp{h} and @samp{b} modifiers for x86:
8382 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8386 These modifiers generate this assembler code:
8392 The rest of this discussion uses the following code for illustrative purposes.
8401 asm volatile goto ("some assembler instructions here"
8403 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8404 : /* No clobbers. */
8409 With no modifiers, this is what the output from the operands would be for the
8410 @samp{att} and @samp{intel} dialects of assembler:
8412 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8413 @headitem Operand @tab masm=att @tab masm=intel
8422 @tab @code{OFFSET FLAT:.L2}
8425 The table below shows the list of supported modifiers and their effects.
8427 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8428 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8430 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8435 @tab Print the QImode name of the register.
8440 @tab Print the QImode name for a ``high'' register.
8445 @tab Print the HImode name of the register.
8450 @tab Print the SImode name of the register.
8455 @tab Print the DImode name of the register.
8460 @tab Print the label name with no punctuation.
8465 @tab Require a constant operand and print the constant expression with no punctuation.
8471 @anchor{x86floatingpointasmoperands}
8472 @subsubsection x86 Floating-Point @code{asm} Operands
8474 On x86 targets, there are several rules on the usage of stack-like registers
8475 in the operands of an @code{asm}. These rules apply only to the operands
8476 that are stack-like registers:
8480 Given a set of input registers that die in an @code{asm}, it is
8481 necessary to know which are implicitly popped by the @code{asm}, and
8482 which must be explicitly popped by GCC@.
8484 An input register that is implicitly popped by the @code{asm} must be
8485 explicitly clobbered, unless it is constrained to match an
8489 For any input register that is implicitly popped by an @code{asm}, it is
8490 necessary to know how to adjust the stack to compensate for the pop.
8491 If any non-popped input is closer to the top of the reg-stack than
8492 the implicitly popped register, it would not be possible to know what the
8493 stack looked like---it's not clear how the rest of the stack ``slides
8496 All implicitly popped input registers must be closer to the top of
8497 the reg-stack than any input that is not implicitly popped.
8499 It is possible that if an input dies in an @code{asm}, the compiler might
8500 use the input register for an output reload. Consider this example:
8503 asm ("foo" : "=t" (a) : "f" (b));
8507 This code says that input @code{b} is not popped by the @code{asm}, and that
8508 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8509 deeper after the @code{asm} than it was before. But, it is possible that
8510 reload may think that it can use the same register for both the input and
8513 To prevent this from happening,
8514 if any input operand uses the @samp{f} constraint, all output register
8515 constraints must use the @samp{&} early-clobber modifier.
8517 The example above is correctly written as:
8520 asm ("foo" : "=&t" (a) : "f" (b));
8524 Some operands need to be in particular places on the stack. All
8525 output operands fall in this category---GCC has no other way to
8526 know which registers the outputs appear in unless you indicate
8527 this in the constraints.
8529 Output operands must specifically indicate which register an output
8530 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8531 constraints must select a class with a single register.
8534 Output operands may not be ``inserted'' between existing stack registers.
8535 Since no 387 opcode uses a read/write operand, all output operands
8536 are dead before the @code{asm}, and are pushed by the @code{asm}.
8537 It makes no sense to push anywhere but the top of the reg-stack.
8539 Output operands must start at the top of the reg-stack: output
8540 operands may not ``skip'' a register.
8543 Some @code{asm} statements may need extra stack space for internal
8544 calculations. This can be guaranteed by clobbering stack registers
8545 unrelated to the inputs and outputs.
8550 takes one input, which is internally popped, and produces two outputs.
8553 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8557 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8558 and replaces them with one output. The @code{st(1)} clobber is necessary
8559 for the compiler to know that @code{fyl2xp1} pops both inputs.
8562 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8570 @subsection Controlling Names Used in Assembler Code
8571 @cindex assembler names for identifiers
8572 @cindex names used in assembler code
8573 @cindex identifiers, names in assembler code
8575 You can specify the name to be used in the assembler code for a C
8576 function or variable by writing the @code{asm} (or @code{__asm__})
8577 keyword after the declarator.
8578 It is up to you to make sure that the assembler names you choose do not
8579 conflict with any other assembler symbols, or reference registers.
8581 @subsubheading Assembler names for data:
8583 This sample shows how to specify the assembler name for data:
8586 int foo asm ("myfoo") = 2;
8590 This specifies that the name to be used for the variable @code{foo} in
8591 the assembler code should be @samp{myfoo} rather than the usual
8594 On systems where an underscore is normally prepended to the name of a C
8595 variable, this feature allows you to define names for the
8596 linker that do not start with an underscore.
8598 GCC does not support using this feature with a non-static local variable
8599 since such variables do not have assembler names. If you are
8600 trying to put the variable in a particular register, see
8601 @ref{Explicit Register Variables}.
8603 @subsubheading Assembler names for functions:
8605 To specify the assembler name for functions, write a declaration for the
8606 function before its definition and put @code{asm} there, like this:
8609 int func (int x, int y) asm ("MYFUNC");
8611 int func (int x, int y)
8617 This specifies that the name to be used for the function @code{func} in
8618 the assembler code should be @code{MYFUNC}.
8620 @node Explicit Register Variables
8621 @subsection Variables in Specified Registers
8622 @anchor{Explicit Reg Vars}
8623 @cindex explicit register variables
8624 @cindex variables in specified registers
8625 @cindex specified registers
8627 GNU C allows you to associate specific hardware registers with C
8628 variables. In almost all cases, allowing the compiler to assign
8629 registers produces the best code. However under certain unusual
8630 circumstances, more precise control over the variable storage is
8633 Both global and local variables can be associated with a register. The
8634 consequences of performing this association are very different between
8635 the two, as explained in the sections below.
8638 * Global Register Variables:: Variables declared at global scope.
8639 * Local Register Variables:: Variables declared within a function.
8642 @node Global Register Variables
8643 @subsubsection Defining Global Register Variables
8644 @anchor{Global Reg Vars}
8645 @cindex global register variables
8646 @cindex registers, global variables in
8647 @cindex registers, global allocation
8649 You can define a global register variable and associate it with a specified
8653 register int *foo asm ("r12");
8657 Here @code{r12} is the name of the register that should be used. Note that
8658 this is the same syntax used for defining local register variables, but for
8659 a global variable the declaration appears outside a function. The
8660 @code{register} keyword is required, and cannot be combined with
8661 @code{static}. The register name must be a valid register name for the
8664 Registers are a scarce resource on most systems and allowing the
8665 compiler to manage their usage usually results in the best code. However,
8666 under special circumstances it can make sense to reserve some globally.
8667 For example this may be useful in programs such as programming language
8668 interpreters that have a couple of global variables that are accessed
8671 After defining a global register variable, for the current compilation
8675 @item The register is reserved entirely for this use, and will not be
8676 allocated for any other purpose.
8677 @item The register is not saved and restored by any functions.
8678 @item Stores into this register are never deleted even if they appear to be
8679 dead, but references may be deleted, moved or simplified.
8682 Note that these points @emph{only} apply to code that is compiled with the
8683 definition. The behavior of code that is merely linked in (for example
8684 code from libraries) is not affected.
8686 If you want to recompile source files that do not actually use your global
8687 register variable so they do not use the specified register for any other
8688 purpose, you need not actually add the global register declaration to
8689 their source code. It suffices to specify the compiler option
8690 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8693 @subsubheading Declaring the variable
8695 Global register variables can not have initial values, because an
8696 executable file has no means to supply initial contents for a register.
8698 When selecting a register, choose one that is normally saved and
8699 restored by function calls on your machine. This ensures that code
8700 which is unaware of this reservation (such as library routines) will
8701 restore it before returning.
8703 On machines with register windows, be sure to choose a global
8704 register that is not affected magically by the function call mechanism.
8706 @subsubheading Using the variable
8708 @cindex @code{qsort}, and global register variables
8709 When calling routines that are not aware of the reservation, be
8710 cautious if those routines call back into code which uses them. As an
8711 example, if you call the system library version of @code{qsort}, it may
8712 clobber your registers during execution, but (if you have selected
8713 appropriate registers) it will restore them before returning. However
8714 it will @emph{not} restore them before calling @code{qsort}'s comparison
8715 function. As a result, global values will not reliably be available to
8716 the comparison function unless the @code{qsort} function itself is rebuilt.
8718 Similarly, it is not safe to access the global register variables from signal
8719 handlers or from more than one thread of control. Unless you recompile
8720 them specially for the task at hand, the system library routines may
8721 temporarily use the register for other things.
8723 @cindex register variable after @code{longjmp}
8724 @cindex global register after @code{longjmp}
8725 @cindex value after @code{longjmp}
8728 On most machines, @code{longjmp} restores to each global register
8729 variable the value it had at the time of the @code{setjmp}. On some
8730 machines, however, @code{longjmp} does not change the value of global
8731 register variables. To be portable, the function that called @code{setjmp}
8732 should make other arrangements to save the values of the global register
8733 variables, and to restore them in a @code{longjmp}. This way, the same
8734 thing happens regardless of what @code{longjmp} does.
8736 Eventually there may be a way of asking the compiler to choose a register
8737 automatically, but first we need to figure out how it should choose and
8738 how to enable you to guide the choice. No solution is evident.
8740 @node Local Register Variables
8741 @subsubsection Specifying Registers for Local Variables
8742 @anchor{Local Reg Vars}
8743 @cindex local variables, specifying registers
8744 @cindex specifying registers for local variables
8745 @cindex registers for local variables
8747 You can define a local register variable and associate it with a specified
8751 register int *foo asm ("r12");
8755 Here @code{r12} is the name of the register that should be used. Note
8756 that this is the same syntax used for defining global register variables,
8757 but for a local variable the declaration appears within a function. The
8758 @code{register} keyword is required, and cannot be combined with
8759 @code{static}. The register name must be a valid register name for the
8762 As with global register variables, it is recommended that you choose
8763 a register that is normally saved and restored by function calls on your
8764 machine, so that calls to library routines will not clobber it.
8766 The only supported use for this feature is to specify registers
8767 for input and output operands when calling Extended @code{asm}
8768 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8769 particular machine don't provide sufficient control to select the desired
8770 register. To force an operand into a register, create a local variable
8771 and specify the register name after the variable's declaration. Then use
8772 the local variable for the @code{asm} operand and specify any constraint
8773 letter that matches the register:
8776 register int *p1 asm ("r0") = @dots{};
8777 register int *p2 asm ("r1") = @dots{};
8778 register int *result asm ("r0");
8779 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8782 @emph{Warning:} In the above example, be aware that a register (for example
8783 @code{r0}) can be call-clobbered by subsequent code, including function
8784 calls and library calls for arithmetic operators on other variables (for
8785 example the initialization of @code{p2}). In this case, use temporary
8786 variables for expressions between the register assignments:
8790 register int *p1 asm ("r0") = @dots{};
8791 register int *p2 asm ("r1") = t1;
8792 register int *result asm ("r0");
8793 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8796 Defining a register variable does not reserve the register. Other than
8797 when invoking the Extended @code{asm}, the contents of the specified
8798 register are not guaranteed. For this reason, the following uses
8799 are explicitly @emph{not} supported. If they appear to work, it is only
8800 happenstance, and may stop working as intended due to (seemingly)
8801 unrelated changes in surrounding code, or even minor changes in the
8802 optimization of a future version of gcc:
8805 @item Passing parameters to or from Basic @code{asm}
8806 @item Passing parameters to or from Extended @code{asm} without using input
8808 @item Passing parameters to or from routines written in assembler (or
8809 other languages) using non-standard calling conventions.
8812 Some developers use Local Register Variables in an attempt to improve
8813 gcc's allocation of registers, especially in large functions. In this
8814 case the register name is essentially a hint to the register allocator.
8815 While in some instances this can generate better code, improvements are
8816 subject to the whims of the allocator/optimizers. Since there are no
8817 guarantees that your improvements won't be lost, this usage of Local
8818 Register Variables is discouraged.
8820 On the MIPS platform, there is related use for local register variables
8821 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8822 Defining coprocessor specifics for MIPS targets, gccint,
8823 GNU Compiler Collection (GCC) Internals}).
8825 @node Size of an asm
8826 @subsection Size of an @code{asm}
8828 Some targets require that GCC track the size of each instruction used
8829 in order to generate correct code. Because the final length of the
8830 code produced by an @code{asm} statement is only known by the
8831 assembler, GCC must make an estimate as to how big it will be. It
8832 does this by counting the number of instructions in the pattern of the
8833 @code{asm} and multiplying that by the length of the longest
8834 instruction supported by that processor. (When working out the number
8835 of instructions, it assumes that any occurrence of a newline or of
8836 whatever statement separator character is supported by the assembler --
8837 typically @samp{;} --- indicates the end of an instruction.)
8839 Normally, GCC's estimate is adequate to ensure that correct
8840 code is generated, but it is possible to confuse the compiler if you use
8841 pseudo instructions or assembler macros that expand into multiple real
8842 instructions, or if you use assembler directives that expand to more
8843 space in the object file than is needed for a single instruction.
8844 If this happens then the assembler may produce a diagnostic saying that
8845 a label is unreachable.
8847 @node Alternate Keywords
8848 @section Alternate Keywords
8849 @cindex alternate keywords
8850 @cindex keywords, alternate
8852 @option{-ansi} and the various @option{-std} options disable certain
8853 keywords. This causes trouble when you want to use GNU C extensions, or
8854 a general-purpose header file that should be usable by all programs,
8855 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8856 @code{inline} are not available in programs compiled with
8857 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8858 program compiled with @option{-std=c99} or @option{-std=c11}). The
8860 @code{restrict} is only available when @option{-std=gnu99} (which will
8861 eventually be the default) or @option{-std=c99} (or the equivalent
8862 @option{-std=iso9899:1999}), or an option for a later standard
8865 The way to solve these problems is to put @samp{__} at the beginning and
8866 end of each problematical keyword. For example, use @code{__asm__}
8867 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8869 Other C compilers won't accept these alternative keywords; if you want to
8870 compile with another compiler, you can define the alternate keywords as
8871 macros to replace them with the customary keywords. It looks like this:
8879 @findex __extension__
8881 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8883 prevent such warnings within one expression by writing
8884 @code{__extension__} before the expression. @code{__extension__} has no
8885 effect aside from this.
8887 @node Incomplete Enums
8888 @section Incomplete @code{enum} Types
8890 You can define an @code{enum} tag without specifying its possible values.
8891 This results in an incomplete type, much like what you get if you write
8892 @code{struct foo} without describing the elements. A later declaration
8893 that does specify the possible values completes the type.
8895 You can't allocate variables or storage using the type while it is
8896 incomplete. However, you can work with pointers to that type.
8898 This extension may not be very useful, but it makes the handling of
8899 @code{enum} more consistent with the way @code{struct} and @code{union}
8902 This extension is not supported by GNU C++.
8904 @node Function Names
8905 @section Function Names as Strings
8906 @cindex @code{__func__} identifier
8907 @cindex @code{__FUNCTION__} identifier
8908 @cindex @code{__PRETTY_FUNCTION__} identifier
8910 GCC provides three magic variables that hold the name of the current
8911 function, as a string. The first of these is @code{__func__}, which
8912 is part of the C99 standard:
8914 The identifier @code{__func__} is implicitly declared by the translator
8915 as if, immediately following the opening brace of each function
8916 definition, the declaration
8919 static const char __func__[] = "function-name";
8923 appeared, where function-name is the name of the lexically-enclosing
8924 function. This name is the unadorned name of the function.
8926 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8927 backward compatibility with old versions of GCC.
8929 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8930 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8931 the type signature of the function as well as its bare name. For
8932 example, this program:
8936 extern int printf (char *, ...);
8943 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8944 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8962 __PRETTY_FUNCTION__ = void a::sub(int)
8965 These identifiers are variables, not preprocessor macros, and may not
8966 be used to initialize @code{char} arrays or be concatenated with other string
8969 @node Return Address
8970 @section Getting the Return or Frame Address of a Function
8972 These functions may be used to get information about the callers of a
8975 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8976 This function returns the return address of the current function, or of
8977 one of its callers. The @var{level} argument is number of frames to
8978 scan up the call stack. A value of @code{0} yields the return address
8979 of the current function, a value of @code{1} yields the return address
8980 of the caller of the current function, and so forth. When inlining
8981 the expected behavior is that the function returns the address of
8982 the function that is returned to. To work around this behavior use
8983 the @code{noinline} function attribute.
8985 The @var{level} argument must be a constant integer.
8987 On some machines it may be impossible to determine the return address of
8988 any function other than the current one; in such cases, or when the top
8989 of the stack has been reached, this function returns @code{0} or a
8990 random value. In addition, @code{__builtin_frame_address} may be used
8991 to determine if the top of the stack has been reached.
8993 Additional post-processing of the returned value may be needed, see
8994 @code{__builtin_extract_return_addr}.
8996 Calling this function with a nonzero argument can have unpredictable
8997 effects, including crashing the calling program. As a result, calls
8998 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8999 option is in effect. Such calls should only be made in debugging
9003 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9004 The address as returned by @code{__builtin_return_address} may have to be fed
9005 through this function to get the actual encoded address. For example, on the
9006 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9007 platforms an offset has to be added for the true next instruction to be
9010 If no fixup is needed, this function simply passes through @var{addr}.
9013 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9014 This function does the reverse of @code{__builtin_extract_return_addr}.
9017 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9018 This function is similar to @code{__builtin_return_address}, but it
9019 returns the address of the function frame rather than the return address
9020 of the function. Calling @code{__builtin_frame_address} with a value of
9021 @code{0} yields the frame address of the current function, a value of
9022 @code{1} yields the frame address of the caller of the current function,
9025 The frame is the area on the stack that holds local variables and saved
9026 registers. The frame address is normally the address of the first word
9027 pushed on to the stack by the function. However, the exact definition
9028 depends upon the processor and the calling convention. If the processor
9029 has a dedicated frame pointer register, and the function has a frame,
9030 then @code{__builtin_frame_address} returns the value of the frame
9033 On some machines it may be impossible to determine the frame address of
9034 any function other than the current one; in such cases, or when the top
9035 of the stack has been reached, this function returns @code{0} if
9036 the first frame pointer is properly initialized by the startup code.
9038 Calling this function with a nonzero argument can have unpredictable
9039 effects, including crashing the calling program. As a result, calls
9040 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9041 option is in effect. Such calls should only be made in debugging
9045 @node Vector Extensions
9046 @section Using Vector Instructions through Built-in Functions
9048 On some targets, the instruction set contains SIMD vector instructions which
9049 operate on multiple values contained in one large register at the same time.
9050 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9053 The first step in using these extensions is to provide the necessary data
9054 types. This should be done using an appropriate @code{typedef}:
9057 typedef int v4si __attribute__ ((vector_size (16)));
9061 The @code{int} type specifies the base type, while the attribute specifies
9062 the vector size for the variable, measured in bytes. For example, the
9063 declaration above causes the compiler to set the mode for the @code{v4si}
9064 type to be 16 bytes wide and divided into @code{int} sized units. For
9065 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9066 corresponding mode of @code{foo} is @acronym{V4SI}.
9068 The @code{vector_size} attribute is only applicable to integral and
9069 float scalars, although arrays, pointers, and function return values
9070 are allowed in conjunction with this construct. Only sizes that are
9071 a power of two are currently allowed.
9073 All the basic integer types can be used as base types, both as signed
9074 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9075 @code{long long}. In addition, @code{float} and @code{double} can be
9076 used to build floating-point vector types.
9078 Specifying a combination that is not valid for the current architecture
9079 causes GCC to synthesize the instructions using a narrower mode.
9080 For example, if you specify a variable of type @code{V4SI} and your
9081 architecture does not allow for this specific SIMD type, GCC
9082 produces code that uses 4 @code{SIs}.
9084 The types defined in this manner can be used with a subset of normal C
9085 operations. Currently, GCC allows using the following operators
9086 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9088 The operations behave like C++ @code{valarrays}. Addition is defined as
9089 the addition of the corresponding elements of the operands. For
9090 example, in the code below, each of the 4 elements in @var{a} is
9091 added to the corresponding 4 elements in @var{b} and the resulting
9092 vector is stored in @var{c}.
9095 typedef int v4si __attribute__ ((vector_size (16)));
9102 Subtraction, multiplication, division, and the logical operations
9103 operate in a similar manner. Likewise, the result of using the unary
9104 minus or complement operators on a vector type is a vector whose
9105 elements are the negative or complemented values of the corresponding
9106 elements in the operand.
9108 It is possible to use shifting operators @code{<<}, @code{>>} on
9109 integer-type vectors. The operation is defined as following: @code{@{a0,
9110 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9111 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9114 For convenience, it is allowed to use a binary vector operation
9115 where one operand is a scalar. In that case the compiler transforms
9116 the scalar operand into a vector where each element is the scalar from
9117 the operation. The transformation happens only if the scalar could be
9118 safely converted to the vector-element type.
9119 Consider the following code.
9122 typedef int v4si __attribute__ ((vector_size (16)));
9127 a = b + 1; /* a = b + @{1,1,1,1@}; */
9128 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9130 a = l + a; /* Error, cannot convert long to int. */
9133 Vectors can be subscripted as if the vector were an array with
9134 the same number of elements and base type. Out of bound accesses
9135 invoke undefined behavior at run time. Warnings for out of bound
9136 accesses for vector subscription can be enabled with
9137 @option{-Warray-bounds}.
9139 Vector comparison is supported with standard comparison
9140 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9141 vector expressions of integer-type or real-type. Comparison between
9142 integer-type vectors and real-type vectors are not supported. The
9143 result of the comparison is a vector of the same width and number of
9144 elements as the comparison operands with a signed integral element
9147 Vectors are compared element-wise producing 0 when comparison is false
9148 and -1 (constant of the appropriate type where all bits are set)
9149 otherwise. Consider the following example.
9152 typedef int v4si __attribute__ ((vector_size (16)));
9154 v4si a = @{1,2,3,4@};
9155 v4si b = @{3,2,1,4@};
9158 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9159 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9162 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9163 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9164 integer vector with the same number of elements of the same size as @code{b}
9165 and @code{c}, computes all three arguments and creates a vector
9166 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9167 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9168 As in the case of binary operations, this syntax is also accepted when
9169 one of @code{b} or @code{c} is a scalar that is then transformed into a
9170 vector. If both @code{b} and @code{c} are scalars and the type of
9171 @code{true?b:c} has the same size as the element type of @code{a}, then
9172 @code{b} and @code{c} are converted to a vector type whose elements have
9173 this type and with the same number of elements as @code{a}.
9175 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9176 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9177 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9178 For mixed operations between a scalar @code{s} and a vector @code{v},
9179 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9180 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9182 Vector shuffling is available using functions
9183 @code{__builtin_shuffle (vec, mask)} and
9184 @code{__builtin_shuffle (vec0, vec1, mask)}.
9185 Both functions construct a permutation of elements from one or two
9186 vectors and return a vector of the same type as the input vector(s).
9187 The @var{mask} is an integral vector with the same width (@var{W})
9188 and element count (@var{N}) as the output vector.
9190 The elements of the input vectors are numbered in memory ordering of
9191 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9192 elements of @var{mask} are considered modulo @var{N} in the single-operand
9193 case and modulo @math{2*@var{N}} in the two-operand case.
9195 Consider the following example,
9198 typedef int v4si __attribute__ ((vector_size (16)));
9200 v4si a = @{1,2,3,4@};
9201 v4si b = @{5,6,7,8@};
9202 v4si mask1 = @{0,1,1,3@};
9203 v4si mask2 = @{0,4,2,5@};
9206 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9207 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9210 Note that @code{__builtin_shuffle} is intentionally semantically
9211 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9213 You can declare variables and use them in function calls and returns, as
9214 well as in assignments and some casts. You can specify a vector type as
9215 a return type for a function. Vector types can also be used as function
9216 arguments. It is possible to cast from one vector type to another,
9217 provided they are of the same size (in fact, you can also cast vectors
9218 to and from other datatypes of the same size).
9220 You cannot operate between vectors of different lengths or different
9221 signedness without a cast.
9224 @section Support for @code{offsetof}
9225 @findex __builtin_offsetof
9227 GCC implements for both C and C++ a syntactic extension to implement
9228 the @code{offsetof} macro.
9232 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9234 offsetof_member_designator:
9236 | offsetof_member_designator "." @code{identifier}
9237 | offsetof_member_designator "[" @code{expr} "]"
9240 This extension is sufficient such that
9243 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9247 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9248 may be dependent. In either case, @var{member} may consist of a single
9249 identifier, or a sequence of member accesses and array references.
9251 @node __sync Builtins
9252 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9254 The following built-in functions
9255 are intended to be compatible with those described
9256 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9257 section 7.4. As such, they depart from normal GCC practice by not using
9258 the @samp{__builtin_} prefix and also by being overloaded so that they
9259 work on multiple types.
9261 The definition given in the Intel documentation allows only for the use of
9262 the types @code{int}, @code{long}, @code{long long} or their unsigned
9263 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
9264 size other than the C type @code{_Bool} or the C++ type @code{bool}.
9265 Operations on pointer arguments are performed as if the operands were
9266 of the @code{uintptr_t} type. That is, they are not scaled by the size
9267 of the type to which the pointer points.
9269 These functions are implemented in terms of the @samp{__atomic}
9270 builtins (@pxref{__atomic Builtins}). They should not be used for new
9271 code which should use the @samp{__atomic} builtins instead.
9273 Not all operations are supported by all target processors. If a particular
9274 operation cannot be implemented on the target processor, a warning is
9275 generated and a call to an external function is generated. The external
9276 function carries the same name as the built-in version,
9277 with an additional suffix
9278 @samp{_@var{n}} where @var{n} is the size of the data type.
9280 @c ??? Should we have a mechanism to suppress this warning? This is almost
9281 @c useful for implementing the operation under the control of an external
9284 In most cases, these built-in functions are considered a @dfn{full barrier}.
9286 no memory operand is moved across the operation, either forward or
9287 backward. Further, instructions are issued as necessary to prevent the
9288 processor from speculating loads across the operation and from queuing stores
9289 after the operation.
9291 All of the routines are described in the Intel documentation to take
9292 ``an optional list of variables protected by the memory barrier''. It's
9293 not clear what is meant by that; it could mean that @emph{only} the
9294 listed variables are protected, or it could mean a list of additional
9295 variables to be protected. The list is ignored by GCC which treats it as
9296 empty. GCC interprets an empty list as meaning that all globally
9297 accessible variables should be protected.
9300 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9301 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9302 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9303 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9304 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9305 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9306 @findex __sync_fetch_and_add
9307 @findex __sync_fetch_and_sub
9308 @findex __sync_fetch_and_or
9309 @findex __sync_fetch_and_and
9310 @findex __sync_fetch_and_xor
9311 @findex __sync_fetch_and_nand
9312 These built-in functions perform the operation suggested by the name, and
9313 returns the value that had previously been in memory. That is, operations
9314 on integer operands have the following semantics. Operations on pointer
9315 arguments are performed as if the operands were of the @code{uintptr_t}
9316 type. That is, they are not scaled by the size of the type to which
9320 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9321 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9324 The object pointed to by the first argument must be of integer or pointer
9325 type. It must not be a Boolean type.
9327 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9328 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9330 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9331 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9332 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9333 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9334 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9335 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9336 @findex __sync_add_and_fetch
9337 @findex __sync_sub_and_fetch
9338 @findex __sync_or_and_fetch
9339 @findex __sync_and_and_fetch
9340 @findex __sync_xor_and_fetch
9341 @findex __sync_nand_and_fetch
9342 These built-in functions perform the operation suggested by the name, and
9343 return the new value. That is, operations on integer operands have
9344 the following semantics. Operations on pointer operands are performed as
9345 if the operand's type were @code{uintptr_t}.
9348 @{ *ptr @var{op}= value; return *ptr; @}
9349 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9352 The same constraints on arguments apply as for the corresponding
9353 @code{__sync_op_and_fetch} built-in functions.
9355 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9356 as @code{*ptr = ~(*ptr & value)} instead of
9357 @code{*ptr = ~*ptr & value}.
9359 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9360 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9361 @findex __sync_bool_compare_and_swap
9362 @findex __sync_val_compare_and_swap
9363 These built-in functions perform an atomic compare and swap.
9364 That is, if the current
9365 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9368 The ``bool'' version returns true if the comparison is successful and
9369 @var{newval} is written. The ``val'' version returns the contents
9370 of @code{*@var{ptr}} before the operation.
9372 @item __sync_synchronize (...)
9373 @findex __sync_synchronize
9374 This built-in function issues a full memory barrier.
9376 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9377 @findex __sync_lock_test_and_set
9378 This built-in function, as described by Intel, is not a traditional test-and-set
9379 operation, but rather an atomic exchange operation. It writes @var{value}
9380 into @code{*@var{ptr}}, and returns the previous contents of
9383 Many targets have only minimal support for such locks, and do not support
9384 a full exchange operation. In this case, a target may support reduced
9385 functionality here by which the @emph{only} valid value to store is the
9386 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9387 is implementation defined.
9389 This built-in function is not a full barrier,
9390 but rather an @dfn{acquire barrier}.
9391 This means that references after the operation cannot move to (or be
9392 speculated to) before the operation, but previous memory stores may not
9393 be globally visible yet, and previous memory loads may not yet be
9396 @item void __sync_lock_release (@var{type} *ptr, ...)
9397 @findex __sync_lock_release
9398 This built-in function releases the lock acquired by
9399 @code{__sync_lock_test_and_set}.
9400 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9402 This built-in function is not a full barrier,
9403 but rather a @dfn{release barrier}.
9404 This means that all previous memory stores are globally visible, and all
9405 previous memory loads have been satisfied, but following memory reads
9406 are not prevented from being speculated to before the barrier.
9409 @node __atomic Builtins
9410 @section Built-in Functions for Memory Model Aware Atomic Operations
9412 The following built-in functions approximately match the requirements
9413 for the C++11 memory model. They are all
9414 identified by being prefixed with @samp{__atomic} and most are
9415 overloaded so that they work with multiple types.
9417 These functions are intended to replace the legacy @samp{__sync}
9418 builtins. The main difference is that the memory order that is requested
9419 is a parameter to the functions. New code should always use the
9420 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9422 Note that the @samp{__atomic} builtins assume that programs will
9423 conform to the C++11 memory model. In particular, they assume
9424 that programs are free of data races. See the C++11 standard for
9425 detailed requirements.
9427 The @samp{__atomic} builtins can be used with any integral scalar or
9428 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9429 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9430 supported by the architecture.
9432 The four non-arithmetic functions (load, store, exchange, and
9433 compare_exchange) all have a generic version as well. This generic
9434 version works on any data type. It uses the lock-free built-in function
9435 if the specific data type size makes that possible; otherwise, an
9436 external call is left to be resolved at run time. This external call is
9437 the same format with the addition of a @samp{size_t} parameter inserted
9438 as the first parameter indicating the size of the object being pointed to.
9439 All objects must be the same size.
9441 There are 6 different memory orders that can be specified. These map
9442 to the C++11 memory orders with the same names, see the C++11 standard
9443 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9444 on atomic synchronization} for detailed definitions. Individual
9445 targets may also support additional memory orders for use on specific
9446 architectures. Refer to the target documentation for details of
9449 An atomic operation can both constrain code motion and
9450 be mapped to hardware instructions for synchronization between threads
9451 (e.g., a fence). To which extent this happens is controlled by the
9452 memory orders, which are listed here in approximately ascending order of
9453 strength. The description of each memory order is only meant to roughly
9454 illustrate the effects and is not a specification; see the C++11
9455 memory model for precise semantics.
9458 @item __ATOMIC_RELAXED
9459 Implies no inter-thread ordering constraints.
9460 @item __ATOMIC_CONSUME
9461 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9462 memory order because of a deficiency in C++11's semantics for
9463 @code{memory_order_consume}.
9464 @item __ATOMIC_ACQUIRE
9465 Creates an inter-thread happens-before constraint from the release (or
9466 stronger) semantic store to this acquire load. Can prevent hoisting
9467 of code to before the operation.
9468 @item __ATOMIC_RELEASE
9469 Creates an inter-thread happens-before constraint to acquire (or stronger)
9470 semantic loads that read from this release store. Can prevent sinking
9471 of code to after the operation.
9472 @item __ATOMIC_ACQ_REL
9473 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9474 @code{__ATOMIC_RELEASE}.
9475 @item __ATOMIC_SEQ_CST
9476 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9479 Note that in the C++11 memory model, @emph{fences} (e.g.,
9480 @samp{__atomic_thread_fence}) take effect in combination with other
9481 atomic operations on specific memory locations (e.g., atomic loads);
9482 operations on specific memory locations do not necessarily affect other
9483 operations in the same way.
9485 Target architectures are encouraged to provide their own patterns for
9486 each of the atomic built-in functions. If no target is provided, the original
9487 non-memory model set of @samp{__sync} atomic built-in functions are
9488 used, along with any required synchronization fences surrounding it in
9489 order to achieve the proper behavior. Execution in this case is subject
9490 to the same restrictions as those built-in functions.
9492 If there is no pattern or mechanism to provide a lock-free instruction
9493 sequence, a call is made to an external routine with the same parameters
9494 to be resolved at run time.
9496 When implementing patterns for these built-in functions, the memory order
9497 parameter can be ignored as long as the pattern implements the most
9498 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9499 orders execute correctly with this memory order but they may not execute as
9500 efficiently as they could with a more appropriate implementation of the
9501 relaxed requirements.
9503 Note that the C++11 standard allows for the memory order parameter to be
9504 determined at run time rather than at compile time. These built-in
9505 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9506 than invoke a runtime library call or inline a switch statement. This is
9507 standard compliant, safe, and the simplest approach for now.
9509 The memory order parameter is a signed int, but only the lower 16 bits are
9510 reserved for the memory order. The remainder of the signed int is reserved
9511 for target use and should be 0. Use of the predefined atomic values
9512 ensures proper usage.
9514 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9515 This built-in function implements an atomic load operation. It returns the
9516 contents of @code{*@var{ptr}}.
9518 The valid memory order variants are
9519 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9520 and @code{__ATOMIC_CONSUME}.
9524 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9525 This is the generic version of an atomic load. It returns the
9526 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9530 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9531 This built-in function implements an atomic store operation. It writes
9532 @code{@var{val}} into @code{*@var{ptr}}.
9534 The valid memory order variants are
9535 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9539 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9540 This is the generic version of an atomic store. It stores the value
9541 of @code{*@var{val}} into @code{*@var{ptr}}.
9545 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9546 This built-in function implements an atomic exchange operation. It writes
9547 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9550 The valid memory order variants are
9551 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9552 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9556 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9557 This is the generic version of an atomic exchange. It stores the
9558 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9559 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9563 @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)
9564 This built-in function implements an atomic compare and exchange operation.
9565 This compares the contents of @code{*@var{ptr}} with the contents of
9566 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9567 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9568 equal, the operation is a @emph{read} and the current contents of
9569 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
9570 for weak compare_exchange, which may fail spuriously, and false for
9571 the strong variation, which never fails spuriously. Many targets
9572 only offer the strong variation and ignore the parameter. When in doubt, use
9573 the strong variation.
9575 If @var{desired} is written into @code{*@var{ptr}} then true is returned
9576 and memory is affected according to the
9577 memory order specified by @var{success_memorder}. There are no
9578 restrictions on what memory order can be used here.
9580 Otherwise, false is returned and memory is affected according
9581 to @var{failure_memorder}. This memory order cannot be
9582 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9583 stronger order than that specified by @var{success_memorder}.
9587 @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)
9588 This built-in function implements the generic version of
9589 @code{__atomic_compare_exchange}. The function is virtually identical to
9590 @code{__atomic_compare_exchange_n}, except the desired value is also a
9595 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9596 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9597 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9598 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9599 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9600 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9601 These built-in functions perform the operation suggested by the name, and
9602 return the result of the operation. Operations on pointer arguments are
9603 performed as if the operands were of the @code{uintptr_t} type. That is,
9604 they are not scaled by the size of the type to which the pointer points.
9607 @{ *ptr @var{op}= val; return *ptr; @}
9610 The object pointed to by the first argument must be of integer or pointer
9611 type. It must not be a Boolean type. All memory orders are valid.
9615 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9616 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9617 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9618 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9619 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9620 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9621 These built-in functions perform the operation suggested by the name, and
9622 return the value that had previously been in @code{*@var{ptr}}. Operations
9623 on pointer arguments are performed as if the operands were of
9624 the @code{uintptr_t} type. That is, they are not scaled by the size of
9625 the type to which the pointer points.
9628 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9631 The same constraints on arguments apply as for the corresponding
9632 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
9636 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9638 This built-in function performs an atomic test-and-set operation on
9639 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9640 defined nonzero ``set'' value and the return value is @code{true} if and only
9641 if the previous contents were ``set''.
9642 It should be only used for operands of type @code{bool} or @code{char}. For
9643 other types only part of the value may be set.
9645 All memory orders are valid.
9649 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9651 This built-in function performs an atomic clear operation on
9652 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9653 It should be only used for operands of type @code{bool} or @code{char} and
9654 in conjunction with @code{__atomic_test_and_set}.
9655 For other types it may only clear partially. If the type is not @code{bool}
9656 prefer using @code{__atomic_store}.
9658 The valid memory order variants are
9659 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9660 @code{__ATOMIC_RELEASE}.
9664 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9666 This built-in function acts as a synchronization fence between threads
9667 based on the specified memory order.
9669 All memory orders are valid.
9673 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9675 This built-in function acts as a synchronization fence between a thread
9676 and signal handlers based in the same thread.
9678 All memory orders are valid.
9682 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9684 This built-in function returns true if objects of @var{size} bytes always
9685 generate lock-free atomic instructions for the target architecture.
9686 @var{size} must resolve to a compile-time constant and the result also
9687 resolves to a compile-time constant.
9689 @var{ptr} is an optional pointer to the object that may be used to determine
9690 alignment. A value of 0 indicates typical alignment should be used. The
9691 compiler may also ignore this parameter.
9694 if (__atomic_always_lock_free (sizeof (long long), 0))
9699 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9701 This built-in function returns true if objects of @var{size} bytes always
9702 generate lock-free atomic instructions for the target architecture. If
9703 the built-in function is not known to be lock-free, a call is made to a
9704 runtime routine named @code{__atomic_is_lock_free}.
9706 @var{ptr} is an optional pointer to the object that may be used to determine
9707 alignment. A value of 0 indicates typical alignment should be used. The
9708 compiler may also ignore this parameter.
9711 @node Integer Overflow Builtins
9712 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9714 The following built-in functions allow performing simple arithmetic operations
9715 together with checking whether the operations overflowed.
9717 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9718 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9719 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9720 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9721 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9722 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9723 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9725 These built-in functions promote the first two operands into infinite precision signed
9726 type and perform addition on those promoted operands. The result is then
9727 cast to the type the third pointer argument points to and stored there.
9728 If the stored result is equal to the infinite precision result, the built-in
9729 functions return false, otherwise they return true. As the addition is
9730 performed in infinite signed precision, these built-in functions have fully defined
9731 behavior for all argument values.
9733 The first built-in function allows arbitrary integral types for operands and
9734 the result type must be pointer to some integer type, the rest of the built-in
9735 functions have explicit integer types.
9737 The compiler will attempt to use hardware instructions to implement
9738 these built-in functions where possible, like conditional jump on overflow
9739 after addition, conditional jump on carry etc.
9743 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9744 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9745 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9746 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9747 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9748 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9749 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9751 These built-in functions are similar to the add overflow checking built-in
9752 functions above, except they perform subtraction, subtract the second argument
9753 from the first one, instead of addition.
9757 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9758 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9759 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9760 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9761 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9762 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9763 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9765 These built-in functions are similar to the add overflow checking built-in
9766 functions above, except they perform multiplication, instead of addition.
9770 @node x86 specific memory model extensions for transactional memory
9771 @section x86-Specific Memory Model Extensions for Transactional Memory
9773 The x86 architecture supports additional memory ordering flags
9774 to mark lock critical sections for hardware lock elision.
9775 These must be specified in addition to an existing memory order to
9779 @item __ATOMIC_HLE_ACQUIRE
9780 Start lock elision on a lock variable.
9781 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9782 @item __ATOMIC_HLE_RELEASE
9783 End lock elision on a lock variable.
9784 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9787 When a lock acquire fails, it is required for good performance to abort
9788 the transaction quickly. This can be done with a @code{_mm_pause}.
9791 #include <immintrin.h> // For _mm_pause
9795 /* Acquire lock with lock elision */
9796 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9797 _mm_pause(); /* Abort failed transaction */
9799 /* Free lock with lock elision */
9800 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9803 @node Object Size Checking
9804 @section Object Size Checking Built-in Functions
9805 @findex __builtin_object_size
9806 @findex __builtin___memcpy_chk
9807 @findex __builtin___mempcpy_chk
9808 @findex __builtin___memmove_chk
9809 @findex __builtin___memset_chk
9810 @findex __builtin___strcpy_chk
9811 @findex __builtin___stpcpy_chk
9812 @findex __builtin___strncpy_chk
9813 @findex __builtin___strcat_chk
9814 @findex __builtin___strncat_chk
9815 @findex __builtin___sprintf_chk
9816 @findex __builtin___snprintf_chk
9817 @findex __builtin___vsprintf_chk
9818 @findex __builtin___vsnprintf_chk
9819 @findex __builtin___printf_chk
9820 @findex __builtin___vprintf_chk
9821 @findex __builtin___fprintf_chk
9822 @findex __builtin___vfprintf_chk
9824 GCC implements a limited buffer overflow protection mechanism
9825 that can prevent some buffer overflow attacks.
9827 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9828 is a built-in construct that returns a constant number of bytes from
9829 @var{ptr} to the end of the object @var{ptr} pointer points to
9830 (if known at compile time). @code{__builtin_object_size} never evaluates
9831 its arguments for side-effects. If there are any side-effects in them, it
9832 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9833 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9834 point to and all of them are known at compile time, the returned number
9835 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9836 0 and minimum if nonzero. If it is not possible to determine which objects
9837 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9838 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9839 for @var{type} 2 or 3.
9841 @var{type} is an integer constant from 0 to 3. If the least significant
9842 bit is clear, objects are whole variables, if it is set, a closest
9843 surrounding subobject is considered the object a pointer points to.
9844 The second bit determines if maximum or minimum of remaining bytes
9848 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9849 char *p = &var.buf1[1], *q = &var.b;
9851 /* Here the object p points to is var. */
9852 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9853 /* The subobject p points to is var.buf1. */
9854 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9855 /* The object q points to is var. */
9856 assert (__builtin_object_size (q, 0)
9857 == (char *) (&var + 1) - (char *) &var.b);
9858 /* The subobject q points to is var.b. */
9859 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9863 There are built-in functions added for many common string operation
9864 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9865 built-in is provided. This built-in has an additional last argument,
9866 which is the number of bytes remaining in object the @var{dest}
9867 argument points to or @code{(size_t) -1} if the size is not known.
9869 The built-in functions are optimized into the normal string functions
9870 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9871 it is known at compile time that the destination object will not
9872 be overflown. If the compiler can determine at compile time the
9873 object will be always overflown, it issues a warning.
9875 The intended use can be e.g.@:
9879 #define bos0(dest) __builtin_object_size (dest, 0)
9880 #define memcpy(dest, src, n) \
9881 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9885 /* It is unknown what object p points to, so this is optimized
9886 into plain memcpy - no checking is possible. */
9887 memcpy (p, "abcde", n);
9888 /* Destination is known and length too. It is known at compile
9889 time there will be no overflow. */
9890 memcpy (&buf[5], "abcde", 5);
9891 /* Destination is known, but the length is not known at compile time.
9892 This will result in __memcpy_chk call that can check for overflow
9894 memcpy (&buf[5], "abcde", n);
9895 /* Destination is known and it is known at compile time there will
9896 be overflow. There will be a warning and __memcpy_chk call that
9897 will abort the program at run time. */
9898 memcpy (&buf[6], "abcde", 5);
9901 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9902 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9903 @code{strcat} and @code{strncat}.
9905 There are also checking built-in functions for formatted output functions.
9907 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9908 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9909 const char *fmt, ...);
9910 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9912 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9913 const char *fmt, va_list ap);
9916 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9917 etc.@: functions and can contain implementation specific flags on what
9918 additional security measures the checking function might take, such as
9919 handling @code{%n} differently.
9921 The @var{os} argument is the object size @var{s} points to, like in the
9922 other built-in functions. There is a small difference in the behavior
9923 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9924 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9925 the checking function is called with @var{os} argument set to
9928 In addition to this, there are checking built-in functions
9929 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9930 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9931 These have just one additional argument, @var{flag}, right before
9932 format string @var{fmt}. If the compiler is able to optimize them to
9933 @code{fputc} etc.@: functions, it does, otherwise the checking function
9934 is called and the @var{flag} argument passed to it.
9936 @node Pointer Bounds Checker builtins
9937 @section Pointer Bounds Checker Built-in Functions
9938 @cindex Pointer Bounds Checker builtins
9939 @findex __builtin___bnd_set_ptr_bounds
9940 @findex __builtin___bnd_narrow_ptr_bounds
9941 @findex __builtin___bnd_copy_ptr_bounds
9942 @findex __builtin___bnd_init_ptr_bounds
9943 @findex __builtin___bnd_null_ptr_bounds
9944 @findex __builtin___bnd_store_ptr_bounds
9945 @findex __builtin___bnd_chk_ptr_lbounds
9946 @findex __builtin___bnd_chk_ptr_ubounds
9947 @findex __builtin___bnd_chk_ptr_bounds
9948 @findex __builtin___bnd_get_ptr_lbound
9949 @findex __builtin___bnd_get_ptr_ubound
9951 GCC provides a set of built-in functions to control Pointer Bounds Checker
9952 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9953 even if you compile with Pointer Bounds Checker off
9954 (@option{-fno-check-pointer-bounds}).
9955 The behavior may differ in such case as documented below.
9957 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9959 This built-in function returns a new pointer with the value of @var{q}, and
9960 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9961 Bounds Checker off, the built-in function just returns the first argument.
9964 extern void *__wrap_malloc (size_t n)
9966 void *p = (void *)__real_malloc (n);
9967 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9968 return __builtin___bnd_set_ptr_bounds (p, n);
9974 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9976 This built-in function returns a new pointer with the value of @var{p}
9977 and associates it with the narrowed bounds formed by the intersection
9978 of bounds associated with @var{q} and the bounds
9979 [@var{p}, @var{p} + @var{size} - 1].
9980 With Pointer Bounds Checker off, the built-in function just returns the first
9984 void init_objects (object *objs, size_t size)
9987 /* Initialize objects one-by-one passing pointers with bounds of
9988 an object, not the full array of objects. */
9989 for (i = 0; i < size; i++)
9990 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
9997 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
9999 This built-in function returns a new pointer with the value of @var{q},
10000 and associates it with the bounds already associated with pointer @var{r}.
10001 With Pointer Bounds Checker off, the built-in function just returns the first
10005 /* Here is a way to get pointer to object's field but
10006 still with the full object's bounds. */
10007 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10013 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10015 This built-in function returns a new pointer with the value of @var{q}, and
10016 associates it with INIT (allowing full memory access) bounds. With Pointer
10017 Bounds Checker off, the built-in function just returns the first argument.
10021 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10023 This built-in function returns a new pointer with the value of @var{q}, and
10024 associates it with NULL (allowing no memory access) bounds. With Pointer
10025 Bounds Checker off, the built-in function just returns the first argument.
10029 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10031 This built-in function stores the bounds associated with pointer @var{ptr_val}
10032 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10033 bounds from legacy code without touching the associated pointer's memory when
10034 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10035 function call is ignored.
10039 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10041 This built-in function checks if the pointer @var{q} is within the lower
10042 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10043 function call is ignored.
10046 extern void *__wrap_memset (void *dst, int c, size_t len)
10050 __builtin___bnd_chk_ptr_lbounds (dst);
10051 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10052 __real_memset (dst, c, len);
10060 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10062 This built-in function checks if the pointer @var{q} is within the upper
10063 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10064 function call is ignored.
10068 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10070 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10071 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10072 off, the built-in function call is ignored.
10075 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10079 __bnd_chk_ptr_bounds (dst, n);
10080 __bnd_chk_ptr_bounds (src, n);
10081 __real_memcpy (dst, src, n);
10089 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10091 This built-in function returns the lower bound associated
10092 with the pointer @var{q}, as a pointer value.
10093 This is useful for debugging using @code{printf}.
10094 With Pointer Bounds Checker off, the built-in function returns 0.
10097 void *lb = __builtin___bnd_get_ptr_lbound (q);
10098 void *ub = __builtin___bnd_get_ptr_ubound (q);
10099 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10104 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10106 This built-in function returns the upper bound (which is a pointer) associated
10107 with the pointer @var{q}. With Pointer Bounds Checker off,
10108 the built-in function returns -1.
10112 @node Cilk Plus Builtins
10113 @section Cilk Plus C/C++ Language Extension Built-in Functions
10115 GCC provides support for the following built-in reduction functions if Cilk Plus
10116 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10119 @item @code{__sec_implicit_index}
10120 @item @code{__sec_reduce}
10121 @item @code{__sec_reduce_add}
10122 @item @code{__sec_reduce_all_nonzero}
10123 @item @code{__sec_reduce_all_zero}
10124 @item @code{__sec_reduce_any_nonzero}
10125 @item @code{__sec_reduce_any_zero}
10126 @item @code{__sec_reduce_max}
10127 @item @code{__sec_reduce_min}
10128 @item @code{__sec_reduce_max_ind}
10129 @item @code{__sec_reduce_min_ind}
10130 @item @code{__sec_reduce_mul}
10131 @item @code{__sec_reduce_mutating}
10134 Further details and examples about these built-in functions are described
10135 in the Cilk Plus language manual which can be found at
10136 @uref{http://www.cilkplus.org}.
10138 @node Other Builtins
10139 @section Other Built-in Functions Provided by GCC
10140 @cindex built-in functions
10141 @findex __builtin_alloca
10142 @findex __builtin_alloca_with_align
10143 @findex __builtin_call_with_static_chain
10144 @findex __builtin_fpclassify
10145 @findex __builtin_isfinite
10146 @findex __builtin_isnormal
10147 @findex __builtin_isgreater
10148 @findex __builtin_isgreaterequal
10149 @findex __builtin_isinf_sign
10150 @findex __builtin_isless
10151 @findex __builtin_islessequal
10152 @findex __builtin_islessgreater
10153 @findex __builtin_isunordered
10154 @findex __builtin_powi
10155 @findex __builtin_powif
10156 @findex __builtin_powil
10317 @findex fprintf_unlocked
10319 @findex fputs_unlocked
10427 @findex nexttowardf
10428 @findex nexttowardl
10436 @findex printf_unlocked
10466 @findex signbitd128
10467 @findex significand
10468 @findex significandf
10469 @findex significandl
10497 @findex strncasecmp
10540 GCC provides a large number of built-in functions other than the ones
10541 mentioned above. Some of these are for internal use in the processing
10542 of exceptions or variable-length argument lists and are not
10543 documented here because they may change from time to time; we do not
10544 recommend general use of these functions.
10546 The remaining functions are provided for optimization purposes.
10548 With the exception of built-ins that have library equivalents such as
10549 the standard C library functions discussed below, or that expand to
10550 library calls, GCC built-in functions are always expanded inline and
10551 thus do not have corresponding entry points and their address cannot
10552 be obtained. Attempting to use them in an expression other than
10553 a function call results in a compile-time error.
10555 @opindex fno-builtin
10556 GCC includes built-in versions of many of the functions in the standard
10557 C library. These functions come in two forms: one whose names start with
10558 the @code{__builtin_} prefix, and the other without. Both forms have the
10559 same type (including prototype), the same address (when their address is
10560 taken), and the same meaning as the C library functions even if you specify
10561 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10562 functions are only optimized in certain cases; if they are not optimized in
10563 a particular case, a call to the library function is emitted.
10567 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10568 @option{-std=c99} or @option{-std=c11}), the functions
10569 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10570 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10571 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10572 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10573 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10574 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10575 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10576 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10577 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10578 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10579 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10580 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10581 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10582 @code{significandl}, @code{significand}, @code{sincosf},
10583 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10584 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10585 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10586 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10588 may be handled as built-in functions.
10589 All these functions have corresponding versions
10590 prefixed with @code{__builtin_}, which may be used even in strict C90
10593 The ISO C99 functions
10594 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10595 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10596 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10597 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10598 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10599 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10600 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10601 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10602 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10603 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10604 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10605 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10606 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10607 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10608 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10609 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10610 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10611 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10612 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10613 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10614 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10615 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10616 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10617 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10618 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10619 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10620 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10621 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10622 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10623 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10624 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10625 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10626 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10627 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10628 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10629 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10630 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10631 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10632 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10633 are handled as built-in functions
10634 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10636 There are also built-in versions of the ISO C99 functions
10637 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10638 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10639 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10640 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10641 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10642 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10643 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10644 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10645 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10646 that are recognized in any mode since ISO C90 reserves these names for
10647 the purpose to which ISO C99 puts them. All these functions have
10648 corresponding versions prefixed with @code{__builtin_}.
10650 There are also GNU extension functions @code{clog10}, @code{clog10f} and
10651 @code{clog10l} which names are reserved by ISO C99 for future use.
10652 All these functions have versions prefixed with @code{__builtin_}.
10654 The ISO C94 functions
10655 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10656 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10657 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10659 are handled as built-in functions
10660 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10662 The ISO C90 functions
10663 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10664 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10665 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10666 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10667 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10668 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10669 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10670 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10671 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10672 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10673 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10674 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10675 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10676 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10677 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10678 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10679 are all recognized as built-in functions unless
10680 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10681 is specified for an individual function). All of these functions have
10682 corresponding versions prefixed with @code{__builtin_}.
10684 GCC provides built-in versions of the ISO C99 floating-point comparison
10685 macros that avoid raising exceptions for unordered operands. They have
10686 the same names as the standard macros ( @code{isgreater},
10687 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10688 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10689 prefixed. We intend for a library implementor to be able to simply
10690 @code{#define} each standard macro to its built-in equivalent.
10691 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10692 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10693 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10694 built-in functions appear both with and without the @code{__builtin_} prefix.
10696 @deftypefn {Built-in Function} void *__builtin_alloca (size_t size)
10697 The @code{__builtin_alloca} function must be called at block scope.
10698 The function allocates an object @var{size} bytes large on the stack
10699 of the calling function. The object is aligned on the default stack
10700 alignment boundary for the target determined by the
10701 @code{__BIGGEST_ALIGNMENT__} macro. The @code{__builtin_alloca}
10702 function returns a pointer to the first byte of the allocated object.
10703 The lifetime of the allocated object ends just before the calling
10704 function returns to its caller. This is so even when
10705 @code{__builtin_alloca} is called within a nested block.
10707 For example, the following function allocates eight objects of @code{n}
10708 bytes each on the stack, storing a pointer to each in consecutive elements
10709 of the array @code{a}. It then passes the array to function @code{g}
10710 which can safely use the storage pointed to by each of the array elements.
10713 void f (unsigned n)
10716 for (int i = 0; i != 8; ++i)
10717 a [i] = __builtin_alloca (n);
10719 g (a, n); // @r{safe}
10723 Since the @code{__builtin_alloca} function doesn't validate its argument
10724 it is the responsibility of its caller to make sure the argument doesn't
10725 cause it to exceed the stack size limit.
10726 The @code{__builtin_alloca} function is provided to make it possible to
10727 allocate on the stack arrays of bytes with an upper bound that may be
10728 computed at run time. Since C99 Variable Length Arrays offer
10729 similar functionality under a portable, more convenient, and safer
10730 interface they are recommended instead, in both C99 and C++ programs
10731 where GCC provides them as an extension.
10732 @xref{Variable Length}, for details.
10736 @deftypefn {Built-in Function} void *__builtin_alloca_with_align (size_t size, size_t alignment)
10737 The @code{__builtin_alloca_with_align} function must be called at block
10738 scope. The function allocates an object @var{size} bytes large on
10739 the stack of the calling function. The allocated object is aligned on
10740 the boundary specified by the argument @var{alignment} whose unit is given
10741 in bits (not bytes). The @var{size} argument must be positive and not
10742 exceed the stack size limit. The @var{alignment} argument must be a constant
10743 integer expression that evaluates to a power of 2 greater than or equal to
10744 @code{CHAR_BIT} and less than some unspecified maximum. Invocations
10745 with other values are rejected with an error indicating the valid bounds.
10746 The function returns a pointer to the first byte of the allocated object.
10747 The lifetime of the allocated object ends at the end of the block in which
10748 the function was called. The allocated storage is released no later than
10749 just before the calling function returns to its caller, but may be released
10750 at the end of the block in which the function was called.
10752 For example, in the following function the call to @code{g} is unsafe
10753 because when @code{overalign} is non-zero, the space allocated by
10754 @code{__builtin_alloca_with_align} may have been released at the end
10755 of the @code{if} statement in which it was called.
10758 void f (unsigned n, bool overalign)
10762 p = __builtin_alloca_with_align (n, 64 /* bits */);
10764 p = __builtin_alloc (n);
10766 g (p, n); // @r{unsafe}
10770 Since the @code{__builtin_alloca_with_align} function doesn't validate its
10771 @var{size} argument it is the responsibility of its caller to make sure
10772 the argument doesn't cause it to exceed the stack size limit.
10773 The @code{__builtin_alloca_with_align} function is provided to make
10774 it possible to allocate on the stack overaligned arrays of bytes with
10775 an upper bound that may be computed at run time. Since C99
10776 Variable Length Arrays offer the same functionality under
10777 a portable, more convenient, and safer interface they are recommended
10778 instead, in both C99 and C++ programs where GCC provides them as
10779 an extension. @xref{Variable Length}, for details.
10783 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10785 You can use the built-in function @code{__builtin_types_compatible_p} to
10786 determine whether two types are the same.
10788 This built-in function returns 1 if the unqualified versions of the
10789 types @var{type1} and @var{type2} (which are types, not expressions) are
10790 compatible, 0 otherwise. The result of this built-in function can be
10791 used in integer constant expressions.
10793 This built-in function ignores top level qualifiers (e.g., @code{const},
10794 @code{volatile}). For example, @code{int} is equivalent to @code{const
10797 The type @code{int[]} and @code{int[5]} are compatible. On the other
10798 hand, @code{int} and @code{char *} are not compatible, even if the size
10799 of their types, on the particular architecture are the same. Also, the
10800 amount of pointer indirection is taken into account when determining
10801 similarity. Consequently, @code{short *} is not similar to
10802 @code{short **}. Furthermore, two types that are typedefed are
10803 considered compatible if their underlying types are compatible.
10805 An @code{enum} type is not considered to be compatible with another
10806 @code{enum} type even if both are compatible with the same integer
10807 type; this is what the C standard specifies.
10808 For example, @code{enum @{foo, bar@}} is not similar to
10809 @code{enum @{hot, dog@}}.
10811 You typically use this function in code whose execution varies
10812 depending on the arguments' types. For example:
10817 typeof (x) tmp = (x); \
10818 if (__builtin_types_compatible_p (typeof (x), long double)) \
10819 tmp = foo_long_double (tmp); \
10820 else if (__builtin_types_compatible_p (typeof (x), double)) \
10821 tmp = foo_double (tmp); \
10822 else if (__builtin_types_compatible_p (typeof (x), float)) \
10823 tmp = foo_float (tmp); \
10830 @emph{Note:} This construct is only available for C@.
10834 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10836 The @var{call_exp} expression must be a function call, and the
10837 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10838 is passed to the function call in the target's static chain location.
10839 The result of builtin is the result of the function call.
10841 @emph{Note:} This builtin is only available for C@.
10842 This builtin can be used to call Go closures from C.
10846 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10848 You can use the built-in function @code{__builtin_choose_expr} to
10849 evaluate code depending on the value of a constant expression. This
10850 built-in function returns @var{exp1} if @var{const_exp}, which is an
10851 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10853 This built-in function is analogous to the @samp{? :} operator in C,
10854 except that the expression returned has its type unaltered by promotion
10855 rules. Also, the built-in function does not evaluate the expression
10856 that is not chosen. For example, if @var{const_exp} evaluates to true,
10857 @var{exp2} is not evaluated even if it has side-effects.
10859 This built-in function can return an lvalue if the chosen argument is an
10862 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10863 type. Similarly, if @var{exp2} is returned, its return type is the same
10870 __builtin_choose_expr ( \
10871 __builtin_types_compatible_p (typeof (x), double), \
10873 __builtin_choose_expr ( \
10874 __builtin_types_compatible_p (typeof (x), float), \
10876 /* @r{The void expression results in a compile-time error} \
10877 @r{when assigning the result to something.} */ \
10881 @emph{Note:} This construct is only available for C@. Furthermore, the
10882 unused expression (@var{exp1} or @var{exp2} depending on the value of
10883 @var{const_exp}) may still generate syntax errors. This may change in
10888 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10890 The built-in function @code{__builtin_complex} is provided for use in
10891 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10892 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10893 real binary floating-point type, and the result has the corresponding
10894 complex type with real and imaginary parts @var{real} and @var{imag}.
10895 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10896 infinities, NaNs and negative zeros are involved.
10900 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10901 You can use the built-in function @code{__builtin_constant_p} to
10902 determine if a value is known to be constant at compile time and hence
10903 that GCC can perform constant-folding on expressions involving that
10904 value. The argument of the function is the value to test. The function
10905 returns the integer 1 if the argument is known to be a compile-time
10906 constant and 0 if it is not known to be a compile-time constant. A
10907 return of 0 does not indicate that the value is @emph{not} a constant,
10908 but merely that GCC cannot prove it is a constant with the specified
10909 value of the @option{-O} option.
10911 You typically use this function in an embedded application where
10912 memory is a critical resource. If you have some complex calculation,
10913 you may want it to be folded if it involves constants, but need to call
10914 a function if it does not. For example:
10917 #define Scale_Value(X) \
10918 (__builtin_constant_p (X) \
10919 ? ((X) * SCALE + OFFSET) : Scale (X))
10922 You may use this built-in function in either a macro or an inline
10923 function. However, if you use it in an inlined function and pass an
10924 argument of the function as the argument to the built-in, GCC
10925 never returns 1 when you call the inline function with a string constant
10926 or compound literal (@pxref{Compound Literals}) and does not return 1
10927 when you pass a constant numeric value to the inline function unless you
10928 specify the @option{-O} option.
10930 You may also use @code{__builtin_constant_p} in initializers for static
10931 data. For instance, you can write
10934 static const int table[] = @{
10935 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10941 This is an acceptable initializer even if @var{EXPRESSION} is not a
10942 constant expression, including the case where
10943 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10944 folded to a constant but @var{EXPRESSION} contains operands that are
10945 not otherwise permitted in a static initializer (for example,
10946 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10947 built-in in this case, because it has no opportunity to perform
10951 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10952 @opindex fprofile-arcs
10953 You may use @code{__builtin_expect} to provide the compiler with
10954 branch prediction information. In general, you should prefer to
10955 use actual profile feedback for this (@option{-fprofile-arcs}), as
10956 programmers are notoriously bad at predicting how their programs
10957 actually perform. However, there are applications in which this
10958 data is hard to collect.
10960 The return value is the value of @var{exp}, which should be an integral
10961 expression. The semantics of the built-in are that it is expected that
10962 @var{exp} == @var{c}. For example:
10965 if (__builtin_expect (x, 0))
10970 indicates that we do not expect to call @code{foo}, since
10971 we expect @code{x} to be zero. Since you are limited to integral
10972 expressions for @var{exp}, you should use constructions such as
10975 if (__builtin_expect (ptr != NULL, 1))
10980 when testing pointer or floating-point values.
10983 @deftypefn {Built-in Function} void __builtin_trap (void)
10984 This function causes the program to exit abnormally. GCC implements
10985 this function by using a target-dependent mechanism (such as
10986 intentionally executing an illegal instruction) or by calling
10987 @code{abort}. The mechanism used may vary from release to release so
10988 you should not rely on any particular implementation.
10991 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10992 If control flow reaches the point of the @code{__builtin_unreachable},
10993 the program is undefined. It is useful in situations where the
10994 compiler cannot deduce the unreachability of the code.
10996 One such case is immediately following an @code{asm} statement that
10997 either never terminates, or one that transfers control elsewhere
10998 and never returns. In this example, without the
10999 @code{__builtin_unreachable}, GCC issues a warning that control
11000 reaches the end of a non-void function. It also generates code
11001 to return after the @code{asm}.
11004 int f (int c, int v)
11012 asm("jmp error_handler");
11013 __builtin_unreachable ();
11019 Because the @code{asm} statement unconditionally transfers control out
11020 of the function, control never reaches the end of the function
11021 body. The @code{__builtin_unreachable} is in fact unreachable and
11022 communicates this fact to the compiler.
11024 Another use for @code{__builtin_unreachable} is following a call a
11025 function that never returns but that is not declared
11026 @code{__attribute__((noreturn))}, as in this example:
11029 void function_that_never_returns (void);
11039 function_that_never_returns ();
11040 __builtin_unreachable ();
11047 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
11048 This function returns its first argument, and allows the compiler
11049 to assume that the returned pointer is at least @var{align} bytes
11050 aligned. This built-in can have either two or three arguments,
11051 if it has three, the third argument should have integer type, and
11052 if it is nonzero means misalignment offset. For example:
11055 void *x = __builtin_assume_aligned (arg, 16);
11059 means that the compiler can assume @code{x}, set to @code{arg}, is at least
11060 16-byte aligned, while:
11063 void *x = __builtin_assume_aligned (arg, 32, 8);
11067 means that the compiler can assume for @code{x}, set to @code{arg}, that
11068 @code{(char *) x - 8} is 32-byte aligned.
11071 @deftypefn {Built-in Function} int __builtin_LINE ()
11072 This function is the equivalent to the preprocessor @code{__LINE__}
11073 macro and returns the line number of the invocation of the built-in.
11074 In a C++ default argument for a function @var{F}, it gets the line number of
11075 the call to @var{F}.
11078 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
11079 This function is the equivalent to the preprocessor @code{__FUNCTION__}
11080 macro and returns the function name the invocation of the built-in is in.
11083 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
11084 This function is the equivalent to the preprocessor @code{__FILE__}
11085 macro and returns the file name the invocation of the built-in is in.
11086 In a C++ default argument for a function @var{F}, it gets the file name of
11087 the call to @var{F}.
11090 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
11091 This function is used to flush the processor's instruction cache for
11092 the region of memory between @var{begin} inclusive and @var{end}
11093 exclusive. Some targets require that the instruction cache be
11094 flushed, after modifying memory containing code, in order to obtain
11095 deterministic behavior.
11097 If the target does not require instruction cache flushes,
11098 @code{__builtin___clear_cache} has no effect. Otherwise either
11099 instructions are emitted in-line to clear the instruction cache or a
11100 call to the @code{__clear_cache} function in libgcc is made.
11103 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
11104 This function is used to minimize cache-miss latency by moving data into
11105 a cache before it is accessed.
11106 You can insert calls to @code{__builtin_prefetch} into code for which
11107 you know addresses of data in memory that is likely to be accessed soon.
11108 If the target supports them, data prefetch instructions are generated.
11109 If the prefetch is done early enough before the access then the data will
11110 be in the cache by the time it is accessed.
11112 The value of @var{addr} is the address of the memory to prefetch.
11113 There are two optional arguments, @var{rw} and @var{locality}.
11114 The value of @var{rw} is a compile-time constant one or zero; one
11115 means that the prefetch is preparing for a write to the memory address
11116 and zero, the default, means that the prefetch is preparing for a read.
11117 The value @var{locality} must be a compile-time constant integer between
11118 zero and three. A value of zero means that the data has no temporal
11119 locality, so it need not be left in the cache after the access. A value
11120 of three means that the data has a high degree of temporal locality and
11121 should be left in all levels of cache possible. Values of one and two
11122 mean, respectively, a low or moderate degree of temporal locality. The
11126 for (i = 0; i < n; i++)
11128 a[i] = a[i] + b[i];
11129 __builtin_prefetch (&a[i+j], 1, 1);
11130 __builtin_prefetch (&b[i+j], 0, 1);
11135 Data prefetch does not generate faults if @var{addr} is invalid, but
11136 the address expression itself must be valid. For example, a prefetch
11137 of @code{p->next} does not fault if @code{p->next} is not a valid
11138 address, but evaluation faults if @code{p} is not a valid address.
11140 If the target does not support data prefetch, the address expression
11141 is evaluated if it includes side effects but no other code is generated
11142 and GCC does not issue a warning.
11145 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11146 Returns a positive infinity, if supported by the floating-point format,
11147 else @code{DBL_MAX}. This function is suitable for implementing the
11148 ISO C macro @code{HUGE_VAL}.
11151 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11152 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11155 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11156 Similar to @code{__builtin_huge_val}, except the return
11157 type is @code{long double}.
11160 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11161 This built-in implements the C99 fpclassify functionality. The first
11162 five int arguments should be the target library's notion of the
11163 possible FP classes and are used for return values. They must be
11164 constant values and they must appear in this order: @code{FP_NAN},
11165 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11166 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11167 to classify. GCC treats the last argument as type-generic, which
11168 means it does not do default promotion from float to double.
11171 @deftypefn {Built-in Function} double __builtin_inf (void)
11172 Similar to @code{__builtin_huge_val}, except a warning is generated
11173 if the target floating-point format does not support infinities.
11176 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11177 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11180 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11181 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11184 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11185 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11188 @deftypefn {Built-in Function} float __builtin_inff (void)
11189 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11190 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11193 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11194 Similar to @code{__builtin_inf}, except the return
11195 type is @code{long double}.
11198 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11199 Similar to @code{isinf}, except the return value is -1 for
11200 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11201 Note while the parameter list is an
11202 ellipsis, this function only accepts exactly one floating-point
11203 argument. GCC treats this parameter as type-generic, which means it
11204 does not do default promotion from float to double.
11207 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11208 This is an implementation of the ISO C99 function @code{nan}.
11210 Since ISO C99 defines this function in terms of @code{strtod}, which we
11211 do not implement, a description of the parsing is in order. The string
11212 is parsed as by @code{strtol}; that is, the base is recognized by
11213 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11214 in the significand such that the least significant bit of the number
11215 is at the least significant bit of the significand. The number is
11216 truncated to fit the significand field provided. The significand is
11217 forced to be a quiet NaN@.
11219 This function, if given a string literal all of which would have been
11220 consumed by @code{strtol}, is evaluated early enough that it is considered a
11221 compile-time constant.
11224 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11225 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11228 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11229 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11232 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11233 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11236 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11237 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11240 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11241 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11244 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11245 Similar to @code{__builtin_nan}, except the significand is forced
11246 to be a signaling NaN@. The @code{nans} function is proposed by
11247 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11250 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11251 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11254 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11255 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11258 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11259 Returns one plus the index of the least significant 1-bit of @var{x}, or
11260 if @var{x} is zero, returns zero.
11263 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11264 Returns the number of leading 0-bits in @var{x}, starting at the most
11265 significant bit position. If @var{x} is 0, the result is undefined.
11268 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11269 Returns the number of trailing 0-bits in @var{x}, starting at the least
11270 significant bit position. If @var{x} is 0, the result is undefined.
11273 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11274 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11275 number of bits following the most significant bit that are identical
11276 to it. There are no special cases for 0 or other values.
11279 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11280 Returns the number of 1-bits in @var{x}.
11283 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11284 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11288 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11289 Similar to @code{__builtin_ffs}, except the argument type is
11293 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11294 Similar to @code{__builtin_clz}, except the argument type is
11295 @code{unsigned long}.
11298 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11299 Similar to @code{__builtin_ctz}, except the argument type is
11300 @code{unsigned long}.
11303 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11304 Similar to @code{__builtin_clrsb}, except the argument type is
11308 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11309 Similar to @code{__builtin_popcount}, except the argument type is
11310 @code{unsigned long}.
11313 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11314 Similar to @code{__builtin_parity}, except the argument type is
11315 @code{unsigned long}.
11318 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11319 Similar to @code{__builtin_ffs}, except the argument type is
11323 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11324 Similar to @code{__builtin_clz}, except the argument type is
11325 @code{unsigned long long}.
11328 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11329 Similar to @code{__builtin_ctz}, except the argument type is
11330 @code{unsigned long long}.
11333 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11334 Similar to @code{__builtin_clrsb}, except the argument type is
11338 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11339 Similar to @code{__builtin_popcount}, except the argument type is
11340 @code{unsigned long long}.
11343 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11344 Similar to @code{__builtin_parity}, except the argument type is
11345 @code{unsigned long long}.
11348 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11349 Returns the first argument raised to the power of the second. Unlike the
11350 @code{pow} function no guarantees about precision and rounding are made.
11353 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11354 Similar to @code{__builtin_powi}, except the argument and return types
11358 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11359 Similar to @code{__builtin_powi}, except the argument and return types
11360 are @code{long double}.
11363 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11364 Returns @var{x} with the order of the bytes reversed; for example,
11365 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11369 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11370 Similar to @code{__builtin_bswap16}, except the argument and return types
11374 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11375 Similar to @code{__builtin_bswap32}, except the argument and return types
11379 @node Target Builtins
11380 @section Built-in Functions Specific to Particular Target Machines
11382 On some target machines, GCC supports many built-in functions specific
11383 to those machines. Generally these generate calls to specific machine
11384 instructions, but allow the compiler to schedule those calls.
11387 * AArch64 Built-in Functions::
11388 * Alpha Built-in Functions::
11389 * Altera Nios II Built-in Functions::
11390 * ARC Built-in Functions::
11391 * ARC SIMD Built-in Functions::
11392 * ARM iWMMXt Built-in Functions::
11393 * ARM C Language Extensions (ACLE)::
11394 * ARM Floating Point Status and Control Intrinsics::
11395 * AVR Built-in Functions::
11396 * Blackfin Built-in Functions::
11397 * FR-V Built-in Functions::
11398 * MIPS DSP Built-in Functions::
11399 * MIPS Paired-Single Support::
11400 * MIPS Loongson Built-in Functions::
11401 * Other MIPS Built-in Functions::
11402 * MSP430 Built-in Functions::
11403 * NDS32 Built-in Functions::
11404 * picoChip Built-in Functions::
11405 * PowerPC Built-in Functions::
11406 * PowerPC AltiVec/VSX Built-in Functions::
11407 * PowerPC Hardware Transactional Memory Built-in Functions::
11408 * RX Built-in Functions::
11409 * S/390 System z Built-in Functions::
11410 * SH Built-in Functions::
11411 * SPARC VIS Built-in Functions::
11412 * SPU Built-in Functions::
11413 * TI C6X Built-in Functions::
11414 * TILE-Gx Built-in Functions::
11415 * TILEPro Built-in Functions::
11416 * x86 Built-in Functions::
11417 * x86 transactional memory intrinsics::
11420 @node AArch64 Built-in Functions
11421 @subsection AArch64 Built-in Functions
11423 These built-in functions are available for the AArch64 family of
11426 unsigned int __builtin_aarch64_get_fpcr ()
11427 void __builtin_aarch64_set_fpcr (unsigned int)
11428 unsigned int __builtin_aarch64_get_fpsr ()
11429 void __builtin_aarch64_set_fpsr (unsigned int)
11432 @node Alpha Built-in Functions
11433 @subsection Alpha Built-in Functions
11435 These built-in functions are available for the Alpha family of
11436 processors, depending on the command-line switches used.
11438 The following built-in functions are always available. They
11439 all generate the machine instruction that is part of the name.
11442 long __builtin_alpha_implver (void)
11443 long __builtin_alpha_rpcc (void)
11444 long __builtin_alpha_amask (long)
11445 long __builtin_alpha_cmpbge (long, long)
11446 long __builtin_alpha_extbl (long, long)
11447 long __builtin_alpha_extwl (long, long)
11448 long __builtin_alpha_extll (long, long)
11449 long __builtin_alpha_extql (long, long)
11450 long __builtin_alpha_extwh (long, long)
11451 long __builtin_alpha_extlh (long, long)
11452 long __builtin_alpha_extqh (long, long)
11453 long __builtin_alpha_insbl (long, long)
11454 long __builtin_alpha_inswl (long, long)
11455 long __builtin_alpha_insll (long, long)
11456 long __builtin_alpha_insql (long, long)
11457 long __builtin_alpha_inswh (long, long)
11458 long __builtin_alpha_inslh (long, long)
11459 long __builtin_alpha_insqh (long, long)
11460 long __builtin_alpha_mskbl (long, long)
11461 long __builtin_alpha_mskwl (long, long)
11462 long __builtin_alpha_mskll (long, long)
11463 long __builtin_alpha_mskql (long, long)
11464 long __builtin_alpha_mskwh (long, long)
11465 long __builtin_alpha_msklh (long, long)
11466 long __builtin_alpha_mskqh (long, long)
11467 long __builtin_alpha_umulh (long, long)
11468 long __builtin_alpha_zap (long, long)
11469 long __builtin_alpha_zapnot (long, long)
11472 The following built-in functions are always with @option{-mmax}
11473 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11474 later. They all generate the machine instruction that is part
11478 long __builtin_alpha_pklb (long)
11479 long __builtin_alpha_pkwb (long)
11480 long __builtin_alpha_unpkbl (long)
11481 long __builtin_alpha_unpkbw (long)
11482 long __builtin_alpha_minub8 (long, long)
11483 long __builtin_alpha_minsb8 (long, long)
11484 long __builtin_alpha_minuw4 (long, long)
11485 long __builtin_alpha_minsw4 (long, long)
11486 long __builtin_alpha_maxub8 (long, long)
11487 long __builtin_alpha_maxsb8 (long, long)
11488 long __builtin_alpha_maxuw4 (long, long)
11489 long __builtin_alpha_maxsw4 (long, long)
11490 long __builtin_alpha_perr (long, long)
11493 The following built-in functions are always with @option{-mcix}
11494 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11495 later. They all generate the machine instruction that is part
11499 long __builtin_alpha_cttz (long)
11500 long __builtin_alpha_ctlz (long)
11501 long __builtin_alpha_ctpop (long)
11504 The following built-in functions are available on systems that use the OSF/1
11505 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11506 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11507 @code{rdval} and @code{wrval}.
11510 void *__builtin_thread_pointer (void)
11511 void __builtin_set_thread_pointer (void *)
11514 @node Altera Nios II Built-in Functions
11515 @subsection Altera Nios II Built-in Functions
11517 These built-in functions are available for the Altera Nios II
11518 family of processors.
11520 The following built-in functions are always available. They
11521 all generate the machine instruction that is part of the name.
11524 int __builtin_ldbio (volatile const void *)
11525 int __builtin_ldbuio (volatile const void *)
11526 int __builtin_ldhio (volatile const void *)
11527 int __builtin_ldhuio (volatile const void *)
11528 int __builtin_ldwio (volatile const void *)
11529 void __builtin_stbio (volatile void *, int)
11530 void __builtin_sthio (volatile void *, int)
11531 void __builtin_stwio (volatile void *, int)
11532 void __builtin_sync (void)
11533 int __builtin_rdctl (int)
11534 int __builtin_rdprs (int, int)
11535 void __builtin_wrctl (int, int)
11536 void __builtin_flushd (volatile void *)
11537 void __builtin_flushda (volatile void *)
11538 int __builtin_wrpie (int);
11539 void __builtin_eni (int);
11540 int __builtin_ldex (volatile const void *)
11541 int __builtin_stex (volatile void *, int)
11542 int __builtin_ldsex (volatile const void *)
11543 int __builtin_stsex (volatile void *, int)
11546 The following built-in functions are always available. They
11547 all generate a Nios II Custom Instruction. The name of the
11548 function represents the types that the function takes and
11549 returns. The letter before the @code{n} is the return type
11550 or void if absent. The @code{n} represents the first parameter
11551 to all the custom instructions, the custom instruction number.
11552 The two letters after the @code{n} represent the up to two
11553 parameters to the function.
11555 The letters represent the following data types:
11558 @code{void} for return type and no parameter for parameter types.
11561 @code{int} for return type and parameter type
11564 @code{float} for return type and parameter type
11567 @code{void *} for return type and parameter type
11571 And the function names are:
11573 void __builtin_custom_n (void)
11574 void __builtin_custom_ni (int)
11575 void __builtin_custom_nf (float)
11576 void __builtin_custom_np (void *)
11577 void __builtin_custom_nii (int, int)
11578 void __builtin_custom_nif (int, float)
11579 void __builtin_custom_nip (int, void *)
11580 void __builtin_custom_nfi (float, int)
11581 void __builtin_custom_nff (float, float)
11582 void __builtin_custom_nfp (float, void *)
11583 void __builtin_custom_npi (void *, int)
11584 void __builtin_custom_npf (void *, float)
11585 void __builtin_custom_npp (void *, void *)
11586 int __builtin_custom_in (void)
11587 int __builtin_custom_ini (int)
11588 int __builtin_custom_inf (float)
11589 int __builtin_custom_inp (void *)
11590 int __builtin_custom_inii (int, int)
11591 int __builtin_custom_inif (int, float)
11592 int __builtin_custom_inip (int, void *)
11593 int __builtin_custom_infi (float, int)
11594 int __builtin_custom_inff (float, float)
11595 int __builtin_custom_infp (float, void *)
11596 int __builtin_custom_inpi (void *, int)
11597 int __builtin_custom_inpf (void *, float)
11598 int __builtin_custom_inpp (void *, void *)
11599 float __builtin_custom_fn (void)
11600 float __builtin_custom_fni (int)
11601 float __builtin_custom_fnf (float)
11602 float __builtin_custom_fnp (void *)
11603 float __builtin_custom_fnii (int, int)
11604 float __builtin_custom_fnif (int, float)
11605 float __builtin_custom_fnip (int, void *)
11606 float __builtin_custom_fnfi (float, int)
11607 float __builtin_custom_fnff (float, float)
11608 float __builtin_custom_fnfp (float, void *)
11609 float __builtin_custom_fnpi (void *, int)
11610 float __builtin_custom_fnpf (void *, float)
11611 float __builtin_custom_fnpp (void *, void *)
11612 void * __builtin_custom_pn (void)
11613 void * __builtin_custom_pni (int)
11614 void * __builtin_custom_pnf (float)
11615 void * __builtin_custom_pnp (void *)
11616 void * __builtin_custom_pnii (int, int)
11617 void * __builtin_custom_pnif (int, float)
11618 void * __builtin_custom_pnip (int, void *)
11619 void * __builtin_custom_pnfi (float, int)
11620 void * __builtin_custom_pnff (float, float)
11621 void * __builtin_custom_pnfp (float, void *)
11622 void * __builtin_custom_pnpi (void *, int)
11623 void * __builtin_custom_pnpf (void *, float)
11624 void * __builtin_custom_pnpp (void *, void *)
11627 @node ARC Built-in Functions
11628 @subsection ARC Built-in Functions
11630 The following built-in functions are provided for ARC targets. The
11631 built-ins generate the corresponding assembly instructions. In the
11632 examples given below, the generated code often requires an operand or
11633 result to be in a register. Where necessary further code will be
11634 generated to ensure this is true, but for brevity this is not
11635 described in each case.
11637 @emph{Note:} Using a built-in to generate an instruction not supported
11638 by a target may cause problems. At present the compiler is not
11639 guaranteed to detect such misuse, and as a result an internal compiler
11640 error may be generated.
11642 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11643 Return 1 if @var{val} is known to have the byte alignment given
11644 by @var{alignval}, otherwise return 0.
11645 Note that this is different from
11647 __alignof__(*(char *)@var{val}) >= alignval
11649 because __alignof__ sees only the type of the dereference, whereas
11650 __builtin_arc_align uses alignment information from the pointer
11651 as well as from the pointed-to type.
11652 The information available will depend on optimization level.
11655 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11662 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11663 The operand is the number of a register to be read. Generates:
11665 mov @var{dest}, r@var{regno}
11667 where the value in @var{dest} will be the result returned from the
11671 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11672 The first operand is the number of a register to be written, the
11673 second operand is a compile time constant to write into that
11674 register. Generates:
11676 mov r@var{regno}, @var{val}
11680 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11681 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11684 divaw @var{dest}, @var{a}, @var{b}
11686 where the value in @var{dest} will be the result returned from the
11690 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11697 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11698 The operand, @var{auxv}, is the address of an auxiliary register and
11699 must be a compile time constant. Generates:
11701 lr @var{dest}, [@var{auxr}]
11703 Where the value in @var{dest} will be the result returned from the
11707 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11708 Only available with @option{-mmul64}. Generates:
11710 mul64 @var{a}, @var{b}
11714 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11715 Only available with @option{-mmul64}. Generates:
11717 mulu64 @var{a}, @var{b}
11721 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11728 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11729 Only valid if the @samp{norm} instruction is available through the
11730 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11733 norm @var{dest}, @var{src}
11735 Where the value in @var{dest} will be the result returned from the
11739 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11740 Only valid if the @samp{normw} instruction is available through the
11741 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11744 normw @var{dest}, @var{src}
11746 Where the value in @var{dest} will be the result returned from the
11750 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11757 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11764 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11765 The first argument, @var{auxv}, is the address of an auxiliary
11766 register, the second argument, @var{val}, is a compile time constant
11767 to be written to the register. Generates:
11769 sr @var{auxr}, [@var{val}]
11773 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11774 Only valid with @option{-mswap}. Generates:
11776 swap @var{dest}, @var{src}
11778 Where the value in @var{dest} will be the result returned from the
11782 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11789 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11790 Only available with @option{-mcpu=ARC700}. Generates:
11796 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11797 Only available with @option{-mcpu=ARC700}. Generates:
11803 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11804 Only available with @option{-mcpu=ARC700}. Generates:
11810 The instructions generated by the following builtins are not
11811 considered as candidates for scheduling. They are not moved around by
11812 the compiler during scheduling, and thus can be expected to appear
11813 where they are put in the C code:
11815 __builtin_arc_brk()
11816 __builtin_arc_core_read()
11817 __builtin_arc_core_write()
11818 __builtin_arc_flag()
11820 __builtin_arc_sleep()
11822 __builtin_arc_swi()
11825 @node ARC SIMD Built-in Functions
11826 @subsection ARC SIMD Built-in Functions
11828 SIMD builtins provided by the compiler can be used to generate the
11829 vector instructions. This section describes the available builtins
11830 and their usage in programs. With the @option{-msimd} option, the
11831 compiler provides 128-bit vector types, which can be specified using
11832 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11833 can be included to use the following predefined types:
11835 typedef int __v4si __attribute__((vector_size(16)));
11836 typedef short __v8hi __attribute__((vector_size(16)));
11839 These types can be used to define 128-bit variables. The built-in
11840 functions listed in the following section can be used on these
11841 variables to generate the vector operations.
11843 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11844 @file{arc-simd.h} also provides equivalent macros called
11845 @code{_@var{someinsn}} that can be used for programming ease and
11846 improved readability. The following macros for DMA control are also
11849 #define _setup_dma_in_channel_reg _vdiwr
11850 #define _setup_dma_out_channel_reg _vdowr
11853 The following is a complete list of all the SIMD built-ins provided
11854 for ARC, grouped by calling signature.
11856 The following take two @code{__v8hi} arguments and return a
11857 @code{__v8hi} result:
11859 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11860 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11861 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11862 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11863 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11864 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11865 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11866 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11867 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11868 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11869 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11870 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11871 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11872 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11873 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11874 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11875 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11876 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11877 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11878 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11879 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11880 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11881 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11882 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11883 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11884 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11885 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11886 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11887 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11888 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11889 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11890 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11891 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11892 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11893 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11894 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11895 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11896 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11897 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11898 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11899 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11900 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11901 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11902 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11903 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11904 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11905 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11906 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11909 The following take one @code{__v8hi} and one @code{int} argument and return a
11910 @code{__v8hi} result:
11913 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11914 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11915 __v8hi __builtin_arc_vbminw (__v8hi, int)
11916 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11917 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11918 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11919 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11920 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11923 The following take one @code{__v8hi} argument and one @code{int} argument which
11924 must be a 3-bit compile time constant indicating a register number
11925 I0-I7. They return a @code{__v8hi} result.
11927 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11928 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11929 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11932 The following take one @code{__v8hi} argument and one @code{int}
11933 argument which must be a 6-bit compile time constant. They return a
11934 @code{__v8hi} result.
11936 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11937 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11938 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11939 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11940 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11941 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11942 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11945 The following take one @code{__v8hi} argument and one @code{int} argument which
11946 must be a 8-bit compile time constant. They return a @code{__v8hi}
11949 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11950 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11951 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11952 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11955 The following take two @code{int} arguments, the second of which which
11956 must be a 8-bit compile time constant. They return a @code{__v8hi}
11959 __v8hi __builtin_arc_vmovaw (int, const int)
11960 __v8hi __builtin_arc_vmovw (int, const int)
11961 __v8hi __builtin_arc_vmovzw (int, const int)
11964 The following take a single @code{__v8hi} argument and return a
11965 @code{__v8hi} result:
11967 __v8hi __builtin_arc_vabsaw (__v8hi)
11968 __v8hi __builtin_arc_vabsw (__v8hi)
11969 __v8hi __builtin_arc_vaddsuw (__v8hi)
11970 __v8hi __builtin_arc_vexch1 (__v8hi)
11971 __v8hi __builtin_arc_vexch2 (__v8hi)
11972 __v8hi __builtin_arc_vexch4 (__v8hi)
11973 __v8hi __builtin_arc_vsignw (__v8hi)
11974 __v8hi __builtin_arc_vupbaw (__v8hi)
11975 __v8hi __builtin_arc_vupbw (__v8hi)
11976 __v8hi __builtin_arc_vupsbaw (__v8hi)
11977 __v8hi __builtin_arc_vupsbw (__v8hi)
11980 The following take two @code{int} arguments and return no result:
11982 void __builtin_arc_vdirun (int, int)
11983 void __builtin_arc_vdorun (int, int)
11986 The following take two @code{int} arguments and return no result. The
11987 first argument must a 3-bit compile time constant indicating one of
11988 the DR0-DR7 DMA setup channels:
11990 void __builtin_arc_vdiwr (const int, int)
11991 void __builtin_arc_vdowr (const int, int)
11994 The following take an @code{int} argument and return no result:
11996 void __builtin_arc_vendrec (int)
11997 void __builtin_arc_vrec (int)
11998 void __builtin_arc_vrecrun (int)
11999 void __builtin_arc_vrun (int)
12002 The following take a @code{__v8hi} argument and two @code{int}
12003 arguments and return a @code{__v8hi} result. The second argument must
12004 be a 3-bit compile time constants, indicating one the registers I0-I7,
12005 and the third argument must be an 8-bit compile time constant.
12007 @emph{Note:} Although the equivalent hardware instructions do not take
12008 an SIMD register as an operand, these builtins overwrite the relevant
12009 bits of the @code{__v8hi} register provided as the first argument with
12010 the value loaded from the @code{[Ib, u8]} location in the SDM.
12013 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
12014 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
12015 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
12016 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
12019 The following take two @code{int} arguments and return a @code{__v8hi}
12020 result. The first argument must be a 3-bit compile time constants,
12021 indicating one the registers I0-I7, and the second argument must be an
12022 8-bit compile time constant.
12025 __v8hi __builtin_arc_vld128 (const int, const int)
12026 __v8hi __builtin_arc_vld64w (const int, const int)
12029 The following take a @code{__v8hi} argument and two @code{int}
12030 arguments and return no result. The second argument must be a 3-bit
12031 compile time constants, indicating one the registers I0-I7, and the
12032 third argument must be an 8-bit compile time constant.
12035 void __builtin_arc_vst128 (__v8hi, const int, const int)
12036 void __builtin_arc_vst64 (__v8hi, const int, const int)
12039 The following take a @code{__v8hi} argument and three @code{int}
12040 arguments and return no result. The second argument must be a 3-bit
12041 compile-time constant, identifying the 16-bit sub-register to be
12042 stored, the third argument must be a 3-bit compile time constants,
12043 indicating one the registers I0-I7, and the fourth argument must be an
12044 8-bit compile time constant.
12047 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
12048 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
12051 @node ARM iWMMXt Built-in Functions
12052 @subsection ARM iWMMXt Built-in Functions
12054 These built-in functions are available for the ARM family of
12055 processors when the @option{-mcpu=iwmmxt} switch is used:
12058 typedef int v2si __attribute__ ((vector_size (8)));
12059 typedef short v4hi __attribute__ ((vector_size (8)));
12060 typedef char v8qi __attribute__ ((vector_size (8)));
12062 int __builtin_arm_getwcgr0 (void)
12063 void __builtin_arm_setwcgr0 (int)
12064 int __builtin_arm_getwcgr1 (void)
12065 void __builtin_arm_setwcgr1 (int)
12066 int __builtin_arm_getwcgr2 (void)
12067 void __builtin_arm_setwcgr2 (int)
12068 int __builtin_arm_getwcgr3 (void)
12069 void __builtin_arm_setwcgr3 (int)
12070 int __builtin_arm_textrmsb (v8qi, int)
12071 int __builtin_arm_textrmsh (v4hi, int)
12072 int __builtin_arm_textrmsw (v2si, int)
12073 int __builtin_arm_textrmub (v8qi, int)
12074 int __builtin_arm_textrmuh (v4hi, int)
12075 int __builtin_arm_textrmuw (v2si, int)
12076 v8qi __builtin_arm_tinsrb (v8qi, int, int)
12077 v4hi __builtin_arm_tinsrh (v4hi, int, int)
12078 v2si __builtin_arm_tinsrw (v2si, int, int)
12079 long long __builtin_arm_tmia (long long, int, int)
12080 long long __builtin_arm_tmiabb (long long, int, int)
12081 long long __builtin_arm_tmiabt (long long, int, int)
12082 long long __builtin_arm_tmiaph (long long, int, int)
12083 long long __builtin_arm_tmiatb (long long, int, int)
12084 long long __builtin_arm_tmiatt (long long, int, int)
12085 int __builtin_arm_tmovmskb (v8qi)
12086 int __builtin_arm_tmovmskh (v4hi)
12087 int __builtin_arm_tmovmskw (v2si)
12088 long long __builtin_arm_waccb (v8qi)
12089 long long __builtin_arm_wacch (v4hi)
12090 long long __builtin_arm_waccw (v2si)
12091 v8qi __builtin_arm_waddb (v8qi, v8qi)
12092 v8qi __builtin_arm_waddbss (v8qi, v8qi)
12093 v8qi __builtin_arm_waddbus (v8qi, v8qi)
12094 v4hi __builtin_arm_waddh (v4hi, v4hi)
12095 v4hi __builtin_arm_waddhss (v4hi, v4hi)
12096 v4hi __builtin_arm_waddhus (v4hi, v4hi)
12097 v2si __builtin_arm_waddw (v2si, v2si)
12098 v2si __builtin_arm_waddwss (v2si, v2si)
12099 v2si __builtin_arm_waddwus (v2si, v2si)
12100 v8qi __builtin_arm_walign (v8qi, v8qi, int)
12101 long long __builtin_arm_wand(long long, long long)
12102 long long __builtin_arm_wandn (long long, long long)
12103 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
12104 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
12105 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
12106 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
12107 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
12108 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
12109 v2si __builtin_arm_wcmpeqw (v2si, v2si)
12110 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
12111 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
12112 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
12113 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
12114 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
12115 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
12116 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
12117 long long __builtin_arm_wmacsz (v4hi, v4hi)
12118 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
12119 long long __builtin_arm_wmacuz (v4hi, v4hi)
12120 v4hi __builtin_arm_wmadds (v4hi, v4hi)
12121 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
12122 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
12123 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
12124 v2si __builtin_arm_wmaxsw (v2si, v2si)
12125 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
12126 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
12127 v2si __builtin_arm_wmaxuw (v2si, v2si)
12128 v8qi __builtin_arm_wminsb (v8qi, v8qi)
12129 v4hi __builtin_arm_wminsh (v4hi, v4hi)
12130 v2si __builtin_arm_wminsw (v2si, v2si)
12131 v8qi __builtin_arm_wminub (v8qi, v8qi)
12132 v4hi __builtin_arm_wminuh (v4hi, v4hi)
12133 v2si __builtin_arm_wminuw (v2si, v2si)
12134 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
12135 v4hi __builtin_arm_wmulul (v4hi, v4hi)
12136 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12137 long long __builtin_arm_wor (long long, long long)
12138 v2si __builtin_arm_wpackdss (long long, long long)
12139 v2si __builtin_arm_wpackdus (long long, long long)
12140 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12141 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12142 v4hi __builtin_arm_wpackwss (v2si, v2si)
12143 v4hi __builtin_arm_wpackwus (v2si, v2si)
12144 long long __builtin_arm_wrord (long long, long long)
12145 long long __builtin_arm_wrordi (long long, int)
12146 v4hi __builtin_arm_wrorh (v4hi, long long)
12147 v4hi __builtin_arm_wrorhi (v4hi, int)
12148 v2si __builtin_arm_wrorw (v2si, long long)
12149 v2si __builtin_arm_wrorwi (v2si, int)
12150 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12151 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12152 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12153 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12154 v4hi __builtin_arm_wshufh (v4hi, int)
12155 long long __builtin_arm_wslld (long long, long long)
12156 long long __builtin_arm_wslldi (long long, int)
12157 v4hi __builtin_arm_wsllh (v4hi, long long)
12158 v4hi __builtin_arm_wsllhi (v4hi, int)
12159 v2si __builtin_arm_wsllw (v2si, long long)
12160 v2si __builtin_arm_wsllwi (v2si, int)
12161 long long __builtin_arm_wsrad (long long, long long)
12162 long long __builtin_arm_wsradi (long long, int)
12163 v4hi __builtin_arm_wsrah (v4hi, long long)
12164 v4hi __builtin_arm_wsrahi (v4hi, int)
12165 v2si __builtin_arm_wsraw (v2si, long long)
12166 v2si __builtin_arm_wsrawi (v2si, int)
12167 long long __builtin_arm_wsrld (long long, long long)
12168 long long __builtin_arm_wsrldi (long long, int)
12169 v4hi __builtin_arm_wsrlh (v4hi, long long)
12170 v4hi __builtin_arm_wsrlhi (v4hi, int)
12171 v2si __builtin_arm_wsrlw (v2si, long long)
12172 v2si __builtin_arm_wsrlwi (v2si, int)
12173 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12174 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12175 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12176 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12177 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12178 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12179 v2si __builtin_arm_wsubw (v2si, v2si)
12180 v2si __builtin_arm_wsubwss (v2si, v2si)
12181 v2si __builtin_arm_wsubwus (v2si, v2si)
12182 v4hi __builtin_arm_wunpckehsb (v8qi)
12183 v2si __builtin_arm_wunpckehsh (v4hi)
12184 long long __builtin_arm_wunpckehsw (v2si)
12185 v4hi __builtin_arm_wunpckehub (v8qi)
12186 v2si __builtin_arm_wunpckehuh (v4hi)
12187 long long __builtin_arm_wunpckehuw (v2si)
12188 v4hi __builtin_arm_wunpckelsb (v8qi)
12189 v2si __builtin_arm_wunpckelsh (v4hi)
12190 long long __builtin_arm_wunpckelsw (v2si)
12191 v4hi __builtin_arm_wunpckelub (v8qi)
12192 v2si __builtin_arm_wunpckeluh (v4hi)
12193 long long __builtin_arm_wunpckeluw (v2si)
12194 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12195 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12196 v2si __builtin_arm_wunpckihw (v2si, v2si)
12197 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12198 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12199 v2si __builtin_arm_wunpckilw (v2si, v2si)
12200 long long __builtin_arm_wxor (long long, long long)
12201 long long __builtin_arm_wzero ()
12205 @node ARM C Language Extensions (ACLE)
12206 @subsection ARM C Language Extensions (ACLE)
12208 GCC implements extensions for C as described in the ARM C Language
12209 Extensions (ACLE) specification, which can be found at
12210 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12212 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12213 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12214 intrinsics can be found at
12215 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12216 The built-in intrinsics for the Advanced SIMD extension are available when
12219 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12220 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12221 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12222 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12225 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12226 availability of extensions.
12228 @node ARM Floating Point Status and Control Intrinsics
12229 @subsection ARM Floating Point Status and Control Intrinsics
12231 These built-in functions are available for the ARM family of
12232 processors with floating-point unit.
12235 unsigned int __builtin_arm_get_fpscr ()
12236 void __builtin_arm_set_fpscr (unsigned int)
12239 @node AVR Built-in Functions
12240 @subsection AVR Built-in Functions
12242 For each built-in function for AVR, there is an equally named,
12243 uppercase built-in macro defined. That way users can easily query if
12244 or if not a specific built-in is implemented or not. For example, if
12245 @code{__builtin_avr_nop} is available the macro
12246 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12248 The following built-in functions map to the respective machine
12249 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12250 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12251 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12252 as library call if no hardware multiplier is available.
12255 void __builtin_avr_nop (void)
12256 void __builtin_avr_sei (void)
12257 void __builtin_avr_cli (void)
12258 void __builtin_avr_sleep (void)
12259 void __builtin_avr_wdr (void)
12260 unsigned char __builtin_avr_swap (unsigned char)
12261 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12262 int __builtin_avr_fmuls (char, char)
12263 int __builtin_avr_fmulsu (char, unsigned char)
12266 In order to delay execution for a specific number of cycles, GCC
12269 void __builtin_avr_delay_cycles (unsigned long ticks)
12273 @code{ticks} is the number of ticks to delay execution. Note that this
12274 built-in does not take into account the effect of interrupts that
12275 might increase delay time. @code{ticks} must be a compile-time
12276 integer constant; delays with a variable number of cycles are not supported.
12279 char __builtin_avr_flash_segment (const __memx void*)
12283 This built-in takes a byte address to the 24-bit
12284 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12285 the number of the flash segment (the 64 KiB chunk) where the address
12286 points to. Counting starts at @code{0}.
12287 If the address does not point to flash memory, return @code{-1}.
12290 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12294 Insert bits from @var{bits} into @var{val} and return the resulting
12295 value. The nibbles of @var{map} determine how the insertion is
12296 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12298 @item If @var{X} is @code{0xf},
12299 then the @var{n}-th bit of @var{val} is returned unaltered.
12301 @item If X is in the range 0@dots{}7,
12302 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12304 @item If X is in the range 8@dots{}@code{0xe},
12305 then the @var{n}-th result bit is undefined.
12309 One typical use case for this built-in is adjusting input and
12310 output values to non-contiguous port layouts. Some examples:
12313 // same as val, bits is unused
12314 __builtin_avr_insert_bits (0xffffffff, bits, val)
12318 // same as bits, val is unused
12319 __builtin_avr_insert_bits (0x76543210, bits, val)
12323 // same as rotating bits by 4
12324 __builtin_avr_insert_bits (0x32107654, bits, 0)
12328 // high nibble of result is the high nibble of val
12329 // low nibble of result is the low nibble of bits
12330 __builtin_avr_insert_bits (0xffff3210, bits, val)
12334 // reverse the bit order of bits
12335 __builtin_avr_insert_bits (0x01234567, bits, 0)
12338 @node Blackfin Built-in Functions
12339 @subsection Blackfin Built-in Functions
12341 Currently, there are two Blackfin-specific built-in functions. These are
12342 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12343 using inline assembly; by using these built-in functions the compiler can
12344 automatically add workarounds for hardware errata involving these
12345 instructions. These functions are named as follows:
12348 void __builtin_bfin_csync (void)
12349 void __builtin_bfin_ssync (void)
12352 @node FR-V Built-in Functions
12353 @subsection FR-V Built-in Functions
12355 GCC provides many FR-V-specific built-in functions. In general,
12356 these functions are intended to be compatible with those described
12357 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12358 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12359 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12360 pointer rather than by value.
12362 Most of the functions are named after specific FR-V instructions.
12363 Such functions are said to be ``directly mapped'' and are summarized
12364 here in tabular form.
12368 * Directly-mapped Integer Functions::
12369 * Directly-mapped Media Functions::
12370 * Raw read/write Functions::
12371 * Other Built-in Functions::
12374 @node Argument Types
12375 @subsubsection Argument Types
12377 The arguments to the built-in functions can be divided into three groups:
12378 register numbers, compile-time constants and run-time values. In order
12379 to make this classification clear at a glance, the arguments and return
12380 values are given the following pseudo types:
12382 @multitable @columnfractions .20 .30 .15 .35
12383 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12384 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12385 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12386 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12387 @item @code{uw2} @tab @code{unsigned long long} @tab No
12388 @tab an unsigned doubleword
12389 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12390 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12391 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12392 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12395 These pseudo types are not defined by GCC, they are simply a notational
12396 convenience used in this manual.
12398 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12399 and @code{sw2} are evaluated at run time. They correspond to
12400 register operands in the underlying FR-V instructions.
12402 @code{const} arguments represent immediate operands in the underlying
12403 FR-V instructions. They must be compile-time constants.
12405 @code{acc} arguments are evaluated at compile time and specify the number
12406 of an accumulator register. For example, an @code{acc} argument of 2
12407 selects the ACC2 register.
12409 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12410 number of an IACC register. See @pxref{Other Built-in Functions}
12413 @node Directly-mapped Integer Functions
12414 @subsubsection Directly-Mapped Integer Functions
12416 The functions listed below map directly to FR-V I-type instructions.
12418 @multitable @columnfractions .45 .32 .23
12419 @item Function prototype @tab Example usage @tab Assembly output
12420 @item @code{sw1 __ADDSS (sw1, sw1)}
12421 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12422 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12423 @item @code{sw1 __SCAN (sw1, sw1)}
12424 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12425 @tab @code{SCAN @var{a},@var{b},@var{c}}
12426 @item @code{sw1 __SCUTSS (sw1)}
12427 @tab @code{@var{b} = __SCUTSS (@var{a})}
12428 @tab @code{SCUTSS @var{a},@var{b}}
12429 @item @code{sw1 __SLASS (sw1, sw1)}
12430 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12431 @tab @code{SLASS @var{a},@var{b},@var{c}}
12432 @item @code{void __SMASS (sw1, sw1)}
12433 @tab @code{__SMASS (@var{a}, @var{b})}
12434 @tab @code{SMASS @var{a},@var{b}}
12435 @item @code{void __SMSSS (sw1, sw1)}
12436 @tab @code{__SMSSS (@var{a}, @var{b})}
12437 @tab @code{SMSSS @var{a},@var{b}}
12438 @item @code{void __SMU (sw1, sw1)}
12439 @tab @code{__SMU (@var{a}, @var{b})}
12440 @tab @code{SMU @var{a},@var{b}}
12441 @item @code{sw2 __SMUL (sw1, sw1)}
12442 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12443 @tab @code{SMUL @var{a},@var{b},@var{c}}
12444 @item @code{sw1 __SUBSS (sw1, sw1)}
12445 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12446 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12447 @item @code{uw2 __UMUL (uw1, uw1)}
12448 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12449 @tab @code{UMUL @var{a},@var{b},@var{c}}
12452 @node Directly-mapped Media Functions
12453 @subsubsection Directly-Mapped Media Functions
12455 The functions listed below map directly to FR-V M-type instructions.
12457 @multitable @columnfractions .45 .32 .23
12458 @item Function prototype @tab Example usage @tab Assembly output
12459 @item @code{uw1 __MABSHS (sw1)}
12460 @tab @code{@var{b} = __MABSHS (@var{a})}
12461 @tab @code{MABSHS @var{a},@var{b}}
12462 @item @code{void __MADDACCS (acc, acc)}
12463 @tab @code{__MADDACCS (@var{b}, @var{a})}
12464 @tab @code{MADDACCS @var{a},@var{b}}
12465 @item @code{sw1 __MADDHSS (sw1, sw1)}
12466 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12467 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12468 @item @code{uw1 __MADDHUS (uw1, uw1)}
12469 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12470 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12471 @item @code{uw1 __MAND (uw1, uw1)}
12472 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12473 @tab @code{MAND @var{a},@var{b},@var{c}}
12474 @item @code{void __MASACCS (acc, acc)}
12475 @tab @code{__MASACCS (@var{b}, @var{a})}
12476 @tab @code{MASACCS @var{a},@var{b}}
12477 @item @code{uw1 __MAVEH (uw1, uw1)}
12478 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12479 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12480 @item @code{uw2 __MBTOH (uw1)}
12481 @tab @code{@var{b} = __MBTOH (@var{a})}
12482 @tab @code{MBTOH @var{a},@var{b}}
12483 @item @code{void __MBTOHE (uw1 *, uw1)}
12484 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12485 @tab @code{MBTOHE @var{a},@var{b}}
12486 @item @code{void __MCLRACC (acc)}
12487 @tab @code{__MCLRACC (@var{a})}
12488 @tab @code{MCLRACC @var{a}}
12489 @item @code{void __MCLRACCA (void)}
12490 @tab @code{__MCLRACCA ()}
12491 @tab @code{MCLRACCA}
12492 @item @code{uw1 __Mcop1 (uw1, uw1)}
12493 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12494 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12495 @item @code{uw1 __Mcop2 (uw1, uw1)}
12496 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12497 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12498 @item @code{uw1 __MCPLHI (uw2, const)}
12499 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12500 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12501 @item @code{uw1 __MCPLI (uw2, const)}
12502 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12503 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12504 @item @code{void __MCPXIS (acc, sw1, sw1)}
12505 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12506 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12507 @item @code{void __MCPXIU (acc, uw1, uw1)}
12508 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12509 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12510 @item @code{void __MCPXRS (acc, sw1, sw1)}
12511 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12512 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12513 @item @code{void __MCPXRU (acc, uw1, uw1)}
12514 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12515 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12516 @item @code{uw1 __MCUT (acc, uw1)}
12517 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12518 @tab @code{MCUT @var{a},@var{b},@var{c}}
12519 @item @code{uw1 __MCUTSS (acc, sw1)}
12520 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12521 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12522 @item @code{void __MDADDACCS (acc, acc)}
12523 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12524 @tab @code{MDADDACCS @var{a},@var{b}}
12525 @item @code{void __MDASACCS (acc, acc)}
12526 @tab @code{__MDASACCS (@var{b}, @var{a})}
12527 @tab @code{MDASACCS @var{a},@var{b}}
12528 @item @code{uw2 __MDCUTSSI (acc, const)}
12529 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12530 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12531 @item @code{uw2 __MDPACKH (uw2, uw2)}
12532 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12533 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12534 @item @code{uw2 __MDROTLI (uw2, const)}
12535 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12536 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12537 @item @code{void __MDSUBACCS (acc, acc)}
12538 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12539 @tab @code{MDSUBACCS @var{a},@var{b}}
12540 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12541 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12542 @tab @code{MDUNPACKH @var{a},@var{b}}
12543 @item @code{uw2 __MEXPDHD (uw1, const)}
12544 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12545 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12546 @item @code{uw1 __MEXPDHW (uw1, const)}
12547 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12548 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12549 @item @code{uw1 __MHDSETH (uw1, const)}
12550 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12551 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12552 @item @code{sw1 __MHDSETS (const)}
12553 @tab @code{@var{b} = __MHDSETS (@var{a})}
12554 @tab @code{MHDSETS #@var{a},@var{b}}
12555 @item @code{uw1 __MHSETHIH (uw1, const)}
12556 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12557 @tab @code{MHSETHIH #@var{a},@var{b}}
12558 @item @code{sw1 __MHSETHIS (sw1, const)}
12559 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12560 @tab @code{MHSETHIS #@var{a},@var{b}}
12561 @item @code{uw1 __MHSETLOH (uw1, const)}
12562 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12563 @tab @code{MHSETLOH #@var{a},@var{b}}
12564 @item @code{sw1 __MHSETLOS (sw1, const)}
12565 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12566 @tab @code{MHSETLOS #@var{a},@var{b}}
12567 @item @code{uw1 __MHTOB (uw2)}
12568 @tab @code{@var{b} = __MHTOB (@var{a})}
12569 @tab @code{MHTOB @var{a},@var{b}}
12570 @item @code{void __MMACHS (acc, sw1, sw1)}
12571 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12572 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12573 @item @code{void __MMACHU (acc, uw1, uw1)}
12574 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12575 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12576 @item @code{void __MMRDHS (acc, sw1, sw1)}
12577 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12578 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12579 @item @code{void __MMRDHU (acc, uw1, uw1)}
12580 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12581 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12582 @item @code{void __MMULHS (acc, sw1, sw1)}
12583 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12584 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12585 @item @code{void __MMULHU (acc, uw1, uw1)}
12586 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12587 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12588 @item @code{void __MMULXHS (acc, sw1, sw1)}
12589 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12590 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12591 @item @code{void __MMULXHU (acc, uw1, uw1)}
12592 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12593 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12594 @item @code{uw1 __MNOT (uw1)}
12595 @tab @code{@var{b} = __MNOT (@var{a})}
12596 @tab @code{MNOT @var{a},@var{b}}
12597 @item @code{uw1 __MOR (uw1, uw1)}
12598 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12599 @tab @code{MOR @var{a},@var{b},@var{c}}
12600 @item @code{uw1 __MPACKH (uh, uh)}
12601 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12602 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12603 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12604 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12605 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12606 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12607 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12608 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12609 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12610 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12611 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12612 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12613 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12614 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12615 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12616 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12617 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12618 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12619 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12620 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12621 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12622 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12623 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12624 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12625 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12626 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12627 @item @code{void __MQMACHS (acc, sw2, sw2)}
12628 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12629 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12630 @item @code{void __MQMACHU (acc, uw2, uw2)}
12631 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12632 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12633 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12634 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12635 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12636 @item @code{void __MQMULHS (acc, sw2, sw2)}
12637 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12638 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12639 @item @code{void __MQMULHU (acc, uw2, uw2)}
12640 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12641 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12642 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12643 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12644 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12645 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12646 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12647 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12648 @item @code{sw2 __MQSATHS (sw2, sw2)}
12649 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12650 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12651 @item @code{uw2 __MQSLLHI (uw2, int)}
12652 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12653 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12654 @item @code{sw2 __MQSRAHI (sw2, int)}
12655 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12656 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12657 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12658 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12659 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12660 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12661 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12662 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12663 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12664 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12665 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12666 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12667 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12668 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12669 @item @code{uw1 __MRDACC (acc)}
12670 @tab @code{@var{b} = __MRDACC (@var{a})}
12671 @tab @code{MRDACC @var{a},@var{b}}
12672 @item @code{uw1 __MRDACCG (acc)}
12673 @tab @code{@var{b} = __MRDACCG (@var{a})}
12674 @tab @code{MRDACCG @var{a},@var{b}}
12675 @item @code{uw1 __MROTLI (uw1, const)}
12676 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12677 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12678 @item @code{uw1 __MROTRI (uw1, const)}
12679 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12680 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12681 @item @code{sw1 __MSATHS (sw1, sw1)}
12682 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12683 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12684 @item @code{uw1 __MSATHU (uw1, uw1)}
12685 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12686 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12687 @item @code{uw1 __MSLLHI (uw1, const)}
12688 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12689 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12690 @item @code{sw1 __MSRAHI (sw1, const)}
12691 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12692 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12693 @item @code{uw1 __MSRLHI (uw1, const)}
12694 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12695 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12696 @item @code{void __MSUBACCS (acc, acc)}
12697 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12698 @tab @code{MSUBACCS @var{a},@var{b}}
12699 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12700 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12701 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12702 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12703 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12704 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12705 @item @code{void __MTRAP (void)}
12706 @tab @code{__MTRAP ()}
12708 @item @code{uw2 __MUNPACKH (uw1)}
12709 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12710 @tab @code{MUNPACKH @var{a},@var{b}}
12711 @item @code{uw1 __MWCUT (uw2, uw1)}
12712 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12713 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12714 @item @code{void __MWTACC (acc, uw1)}
12715 @tab @code{__MWTACC (@var{b}, @var{a})}
12716 @tab @code{MWTACC @var{a},@var{b}}
12717 @item @code{void __MWTACCG (acc, uw1)}
12718 @tab @code{__MWTACCG (@var{b}, @var{a})}
12719 @tab @code{MWTACCG @var{a},@var{b}}
12720 @item @code{uw1 __MXOR (uw1, uw1)}
12721 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12722 @tab @code{MXOR @var{a},@var{b},@var{c}}
12725 @node Raw read/write Functions
12726 @subsubsection Raw Read/Write Functions
12728 This sections describes built-in functions related to read and write
12729 instructions to access memory. These functions generate
12730 @code{membar} instructions to flush the I/O load and stores where
12731 appropriate, as described in Fujitsu's manual described above.
12735 @item unsigned char __builtin_read8 (void *@var{data})
12736 @item unsigned short __builtin_read16 (void *@var{data})
12737 @item unsigned long __builtin_read32 (void *@var{data})
12738 @item unsigned long long __builtin_read64 (void *@var{data})
12740 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12741 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12742 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12743 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12746 @node Other Built-in Functions
12747 @subsubsection Other Built-in Functions
12749 This section describes built-in functions that are not named after
12750 a specific FR-V instruction.
12753 @item sw2 __IACCreadll (iacc @var{reg})
12754 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12755 for future expansion and must be 0.
12757 @item sw1 __IACCreadl (iacc @var{reg})
12758 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12759 Other values of @var{reg} are rejected as invalid.
12761 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12762 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12763 is reserved for future expansion and must be 0.
12765 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12766 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12767 is 1. Other values of @var{reg} are rejected as invalid.
12769 @item void __data_prefetch0 (const void *@var{x})
12770 Use the @code{dcpl} instruction to load the contents of address @var{x}
12771 into the data cache.
12773 @item void __data_prefetch (const void *@var{x})
12774 Use the @code{nldub} instruction to load the contents of address @var{x}
12775 into the data cache. The instruction is issued in slot I1@.
12778 @node MIPS DSP Built-in Functions
12779 @subsection MIPS DSP Built-in Functions
12781 The MIPS DSP Application-Specific Extension (ASE) includes new
12782 instructions that are designed to improve the performance of DSP and
12783 media applications. It provides instructions that operate on packed
12784 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12786 GCC supports MIPS DSP operations using both the generic
12787 vector extensions (@pxref{Vector Extensions}) and a collection of
12788 MIPS-specific built-in functions. Both kinds of support are
12789 enabled by the @option{-mdsp} command-line option.
12791 Revision 2 of the ASE was introduced in the second half of 2006.
12792 This revision adds extra instructions to the original ASE, but is
12793 otherwise backwards-compatible with it. You can select revision 2
12794 using the command-line option @option{-mdspr2}; this option implies
12797 The SCOUNT and POS bits of the DSP control register are global. The
12798 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12799 POS bits. During optimization, the compiler does not delete these
12800 instructions and it does not delete calls to functions containing
12801 these instructions.
12803 At present, GCC only provides support for operations on 32-bit
12804 vectors. The vector type associated with 8-bit integer data is
12805 usually called @code{v4i8}, the vector type associated with Q7
12806 is usually called @code{v4q7}, the vector type associated with 16-bit
12807 integer data is usually called @code{v2i16}, and the vector type
12808 associated with Q15 is usually called @code{v2q15}. They can be
12809 defined in C as follows:
12812 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12813 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12814 typedef short v2i16 __attribute__ ((vector_size(4)));
12815 typedef short v2q15 __attribute__ ((vector_size(4)));
12818 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12819 initialized in the same way as aggregates. For example:
12822 v4i8 a = @{1, 2, 3, 4@};
12824 b = (v4i8) @{5, 6, 7, 8@};
12826 v2q15 c = @{0x0fcb, 0x3a75@};
12828 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12831 @emph{Note:} The CPU's endianness determines the order in which values
12832 are packed. On little-endian targets, the first value is the least
12833 significant and the last value is the most significant. The opposite
12834 order applies to big-endian targets. For example, the code above
12835 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12836 and @code{4} on big-endian targets.
12838 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12839 representation. As shown in this example, the integer representation
12840 of a Q7 value can be obtained by multiplying the fractional value by
12841 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12842 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12845 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12846 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12847 and @code{c} and @code{d} are @code{v2q15} values.
12849 @multitable @columnfractions .50 .50
12850 @item C code @tab MIPS instruction
12851 @item @code{a + b} @tab @code{addu.qb}
12852 @item @code{c + d} @tab @code{addq.ph}
12853 @item @code{a - b} @tab @code{subu.qb}
12854 @item @code{c - d} @tab @code{subq.ph}
12857 The table below lists the @code{v2i16} operation for which
12858 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12859 @code{v2i16} values.
12861 @multitable @columnfractions .50 .50
12862 @item C code @tab MIPS instruction
12863 @item @code{e * f} @tab @code{mul.ph}
12866 It is easier to describe the DSP built-in functions if we first define
12867 the following types:
12872 typedef unsigned int ui32;
12873 typedef long long a64;
12876 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12877 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12878 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12879 @code{long long}, but we use @code{a64} to indicate values that are
12880 placed in one of the four DSP accumulators (@code{$ac0},
12881 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12883 Also, some built-in functions prefer or require immediate numbers as
12884 parameters, because the corresponding DSP instructions accept both immediate
12885 numbers and register operands, or accept immediate numbers only. The
12886 immediate parameters are listed as follows.
12894 imm0_255: 0 to 255.
12895 imm_n32_31: -32 to 31.
12896 imm_n512_511: -512 to 511.
12899 The following built-in functions map directly to a particular MIPS DSP
12900 instruction. Please refer to the architecture specification
12901 for details on what each instruction does.
12904 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12905 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12906 q31 __builtin_mips_addq_s_w (q31, q31)
12907 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12908 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12909 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12910 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12911 q31 __builtin_mips_subq_s_w (q31, q31)
12912 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12913 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12914 i32 __builtin_mips_addsc (i32, i32)
12915 i32 __builtin_mips_addwc (i32, i32)
12916 i32 __builtin_mips_modsub (i32, i32)
12917 i32 __builtin_mips_raddu_w_qb (v4i8)
12918 v2q15 __builtin_mips_absq_s_ph (v2q15)
12919 q31 __builtin_mips_absq_s_w (q31)
12920 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12921 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12922 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12923 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12924 q31 __builtin_mips_preceq_w_phl (v2q15)
12925 q31 __builtin_mips_preceq_w_phr (v2q15)
12926 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12927 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12928 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12929 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12930 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12931 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12932 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12933 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12934 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12935 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12936 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12937 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12938 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12939 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12940 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12941 q31 __builtin_mips_shll_s_w (q31, i32)
12942 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12943 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12944 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12945 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12946 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12947 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12948 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12949 q31 __builtin_mips_shra_r_w (q31, i32)
12950 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12951 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12952 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12953 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12954 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12955 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12956 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12957 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12958 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12959 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12960 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12961 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12962 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12963 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12964 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12965 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12966 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12967 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12968 i32 __builtin_mips_bitrev (i32)
12969 i32 __builtin_mips_insv (i32, i32)
12970 v4i8 __builtin_mips_repl_qb (imm0_255)
12971 v4i8 __builtin_mips_repl_qb (i32)
12972 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12973 v2q15 __builtin_mips_repl_ph (i32)
12974 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12975 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12976 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12977 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12978 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12979 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12980 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12981 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12982 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12983 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12984 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12985 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12986 i32 __builtin_mips_extr_w (a64, imm0_31)
12987 i32 __builtin_mips_extr_w (a64, i32)
12988 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12989 i32 __builtin_mips_extr_s_h (a64, i32)
12990 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12991 i32 __builtin_mips_extr_rs_w (a64, i32)
12992 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12993 i32 __builtin_mips_extr_r_w (a64, i32)
12994 i32 __builtin_mips_extp (a64, imm0_31)
12995 i32 __builtin_mips_extp (a64, i32)
12996 i32 __builtin_mips_extpdp (a64, imm0_31)
12997 i32 __builtin_mips_extpdp (a64, i32)
12998 a64 __builtin_mips_shilo (a64, imm_n32_31)
12999 a64 __builtin_mips_shilo (a64, i32)
13000 a64 __builtin_mips_mthlip (a64, i32)
13001 void __builtin_mips_wrdsp (i32, imm0_63)
13002 i32 __builtin_mips_rddsp (imm0_63)
13003 i32 __builtin_mips_lbux (void *, i32)
13004 i32 __builtin_mips_lhx (void *, i32)
13005 i32 __builtin_mips_lwx (void *, i32)
13006 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
13007 i32 __builtin_mips_bposge32 (void)
13008 a64 __builtin_mips_madd (a64, i32, i32);
13009 a64 __builtin_mips_maddu (a64, ui32, ui32);
13010 a64 __builtin_mips_msub (a64, i32, i32);
13011 a64 __builtin_mips_msubu (a64, ui32, ui32);
13012 a64 __builtin_mips_mult (i32, i32);
13013 a64 __builtin_mips_multu (ui32, ui32);
13016 The following built-in functions map directly to a particular MIPS DSP REV 2
13017 instruction. Please refer to the architecture specification
13018 for details on what each instruction does.
13021 v4q7 __builtin_mips_absq_s_qb (v4q7);
13022 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
13023 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
13024 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
13025 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
13026 i32 __builtin_mips_append (i32, i32, imm0_31);
13027 i32 __builtin_mips_balign (i32, i32, imm0_3);
13028 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
13029 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
13030 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
13031 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
13032 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
13033 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
13034 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
13035 q31 __builtin_mips_mulq_rs_w (q31, q31);
13036 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
13037 q31 __builtin_mips_mulq_s_w (q31, q31);
13038 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
13039 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
13040 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
13041 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
13042 i32 __builtin_mips_prepend (i32, i32, imm0_31);
13043 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
13044 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
13045 v4i8 __builtin_mips_shra_qb (v4i8, i32);
13046 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
13047 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
13048 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
13049 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
13050 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
13051 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
13052 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
13053 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
13054 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
13055 q31 __builtin_mips_addqh_w (q31, q31);
13056 q31 __builtin_mips_addqh_r_w (q31, q31);
13057 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
13058 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
13059 q31 __builtin_mips_subqh_w (q31, q31);
13060 q31 __builtin_mips_subqh_r_w (q31, q31);
13061 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
13062 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
13063 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
13064 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
13065 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
13066 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
13070 @node MIPS Paired-Single Support
13071 @subsection MIPS Paired-Single Support
13073 The MIPS64 architecture includes a number of instructions that
13074 operate on pairs of single-precision floating-point values.
13075 Each pair is packed into a 64-bit floating-point register,
13076 with one element being designated the ``upper half'' and
13077 the other being designated the ``lower half''.
13079 GCC supports paired-single operations using both the generic
13080 vector extensions (@pxref{Vector Extensions}) and a collection of
13081 MIPS-specific built-in functions. Both kinds of support are
13082 enabled by the @option{-mpaired-single} command-line option.
13084 The vector type associated with paired-single values is usually
13085 called @code{v2sf}. It can be defined in C as follows:
13088 typedef float v2sf __attribute__ ((vector_size (8)));
13091 @code{v2sf} values are initialized in the same way as aggregates.
13095 v2sf a = @{1.5, 9.1@};
13098 b = (v2sf) @{e, f@};
13101 @emph{Note:} The CPU's endianness determines which value is stored in
13102 the upper half of a register and which value is stored in the lower half.
13103 On little-endian targets, the first value is the lower one and the second
13104 value is the upper one. The opposite order applies to big-endian targets.
13105 For example, the code above sets the lower half of @code{a} to
13106 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
13108 @node MIPS Loongson Built-in Functions
13109 @subsection MIPS Loongson Built-in Functions
13111 GCC provides intrinsics to access the SIMD instructions provided by the
13112 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
13113 available after inclusion of the @code{loongson.h} header file,
13114 operate on the following 64-bit vector types:
13117 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
13118 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
13119 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
13120 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
13121 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
13122 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
13125 The intrinsics provided are listed below; each is named after the
13126 machine instruction to which it corresponds, with suffixes added as
13127 appropriate to distinguish intrinsics that expand to the same machine
13128 instruction yet have different argument types. Refer to the architecture
13129 documentation for a description of the functionality of each
13133 int16x4_t packsswh (int32x2_t s, int32x2_t t);
13134 int8x8_t packsshb (int16x4_t s, int16x4_t t);
13135 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
13136 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13137 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13138 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13139 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13140 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13141 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13142 uint64_t paddd_u (uint64_t s, uint64_t t);
13143 int64_t paddd_s (int64_t s, int64_t t);
13144 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13145 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13146 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13147 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13148 uint64_t pandn_ud (uint64_t s, uint64_t t);
13149 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13150 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13151 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13152 int64_t pandn_sd (int64_t s, int64_t t);
13153 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13154 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13155 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13156 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13157 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13158 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13159 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13160 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13161 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13162 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13163 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13164 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13165 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13166 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13167 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13168 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13169 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13170 uint16x4_t pextrh_u (uint16x4_t s, int field);
13171 int16x4_t pextrh_s (int16x4_t s, int field);
13172 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13173 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13174 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13175 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13176 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13177 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13178 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13179 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13180 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13181 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13182 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13183 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13184 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13185 uint8x8_t pmovmskb_u (uint8x8_t s);
13186 int8x8_t pmovmskb_s (int8x8_t s);
13187 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13188 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13189 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13190 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13191 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13192 uint16x4_t biadd (uint8x8_t s);
13193 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13194 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13195 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13196 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13197 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13198 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13199 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13200 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13201 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13202 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13203 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13204 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13205 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13206 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13207 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13208 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13209 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13210 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13211 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13212 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13213 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13214 uint64_t psubd_u (uint64_t s, uint64_t t);
13215 int64_t psubd_s (int64_t s, int64_t t);
13216 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13217 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13218 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13219 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13220 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13221 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13222 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13223 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13224 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13225 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13226 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13227 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13228 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13229 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13230 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13231 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13235 * Paired-Single Arithmetic::
13236 * Paired-Single Built-in Functions::
13237 * MIPS-3D Built-in Functions::
13240 @node Paired-Single Arithmetic
13241 @subsubsection Paired-Single Arithmetic
13243 The table below lists the @code{v2sf} operations for which hardware
13244 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13245 values and @code{x} is an integral value.
13247 @multitable @columnfractions .50 .50
13248 @item C code @tab MIPS instruction
13249 @item @code{a + b} @tab @code{add.ps}
13250 @item @code{a - b} @tab @code{sub.ps}
13251 @item @code{-a} @tab @code{neg.ps}
13252 @item @code{a * b} @tab @code{mul.ps}
13253 @item @code{a * b + c} @tab @code{madd.ps}
13254 @item @code{a * b - c} @tab @code{msub.ps}
13255 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13256 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13257 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13260 Note that the multiply-accumulate instructions can be disabled
13261 using the command-line option @code{-mno-fused-madd}.
13263 @node Paired-Single Built-in Functions
13264 @subsubsection Paired-Single Built-in Functions
13266 The following paired-single functions map directly to a particular
13267 MIPS instruction. Please refer to the architecture specification
13268 for details on what each instruction does.
13271 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13272 Pair lower lower (@code{pll.ps}).
13274 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13275 Pair upper lower (@code{pul.ps}).
13277 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13278 Pair lower upper (@code{plu.ps}).
13280 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13281 Pair upper upper (@code{puu.ps}).
13283 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13284 Convert pair to paired single (@code{cvt.ps.s}).
13286 @item float __builtin_mips_cvt_s_pl (v2sf)
13287 Convert pair lower to single (@code{cvt.s.pl}).
13289 @item float __builtin_mips_cvt_s_pu (v2sf)
13290 Convert pair upper to single (@code{cvt.s.pu}).
13292 @item v2sf __builtin_mips_abs_ps (v2sf)
13293 Absolute value (@code{abs.ps}).
13295 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13296 Align variable (@code{alnv.ps}).
13298 @emph{Note:} The value of the third parameter must be 0 or 4
13299 modulo 8, otherwise the result is unpredictable. Please read the
13300 instruction description for details.
13303 The following multi-instruction functions are also available.
13304 In each case, @var{cond} can be any of the 16 floating-point conditions:
13305 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13306 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13307 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13310 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13311 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13312 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13313 @code{movt.ps}/@code{movf.ps}).
13315 The @code{movt} functions return the value @var{x} computed by:
13318 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13319 mov.ps @var{x},@var{c}
13320 movt.ps @var{x},@var{d},@var{cc}
13323 The @code{movf} functions are similar but use @code{movf.ps} instead
13326 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13327 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13328 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13329 @code{bc1t}/@code{bc1f}).
13331 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13332 and return either the upper or lower half of the result. For example:
13336 if (__builtin_mips_upper_c_eq_ps (a, b))
13337 upper_halves_are_equal ();
13339 upper_halves_are_unequal ();
13341 if (__builtin_mips_lower_c_eq_ps (a, b))
13342 lower_halves_are_equal ();
13344 lower_halves_are_unequal ();
13348 @node MIPS-3D Built-in Functions
13349 @subsubsection MIPS-3D Built-in Functions
13351 The MIPS-3D Application-Specific Extension (ASE) includes additional
13352 paired-single instructions that are designed to improve the performance
13353 of 3D graphics operations. Support for these instructions is controlled
13354 by the @option{-mips3d} command-line option.
13356 The functions listed below map directly to a particular MIPS-3D
13357 instruction. Please refer to the architecture specification for
13358 more details on what each instruction does.
13361 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13362 Reduction add (@code{addr.ps}).
13364 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13365 Reduction multiply (@code{mulr.ps}).
13367 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13368 Convert paired single to paired word (@code{cvt.pw.ps}).
13370 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13371 Convert paired word to paired single (@code{cvt.ps.pw}).
13373 @item float __builtin_mips_recip1_s (float)
13374 @itemx double __builtin_mips_recip1_d (double)
13375 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13376 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13378 @item float __builtin_mips_recip2_s (float, float)
13379 @itemx double __builtin_mips_recip2_d (double, double)
13380 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13381 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13383 @item float __builtin_mips_rsqrt1_s (float)
13384 @itemx double __builtin_mips_rsqrt1_d (double)
13385 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13386 Reduced-precision reciprocal square root (sequence step 1)
13387 (@code{rsqrt1.@var{fmt}}).
13389 @item float __builtin_mips_rsqrt2_s (float, float)
13390 @itemx double __builtin_mips_rsqrt2_d (double, double)
13391 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13392 Reduced-precision reciprocal square root (sequence step 2)
13393 (@code{rsqrt2.@var{fmt}}).
13396 The following multi-instruction functions are also available.
13397 In each case, @var{cond} can be any of the 16 floating-point conditions:
13398 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13399 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13400 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13403 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13404 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13405 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13406 @code{bc1t}/@code{bc1f}).
13408 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13409 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13414 if (__builtin_mips_cabs_eq_s (a, b))
13420 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13421 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13422 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13423 @code{bc1t}/@code{bc1f}).
13425 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13426 and return either the upper or lower half of the result. For example:
13430 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13431 upper_halves_are_equal ();
13433 upper_halves_are_unequal ();
13435 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13436 lower_halves_are_equal ();
13438 lower_halves_are_unequal ();
13441 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13442 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13443 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13444 @code{movt.ps}/@code{movf.ps}).
13446 The @code{movt} functions return the value @var{x} computed by:
13449 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13450 mov.ps @var{x},@var{c}
13451 movt.ps @var{x},@var{d},@var{cc}
13454 The @code{movf} functions are similar but use @code{movf.ps} instead
13457 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13458 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13459 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13460 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13461 Comparison of two paired-single values
13462 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13463 @code{bc1any2t}/@code{bc1any2f}).
13465 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13466 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13467 result is true and the @code{all} forms return true if both results are true.
13472 if (__builtin_mips_any_c_eq_ps (a, b))
13477 if (__builtin_mips_all_c_eq_ps (a, b))
13483 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13484 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13485 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13486 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13487 Comparison of four paired-single values
13488 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13489 @code{bc1any4t}/@code{bc1any4f}).
13491 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13492 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13493 The @code{any} forms return true if any of the four results are true
13494 and the @code{all} forms return true if all four results are true.
13499 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13504 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13511 @node Other MIPS Built-in Functions
13512 @subsection Other MIPS Built-in Functions
13514 GCC provides other MIPS-specific built-in functions:
13517 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13518 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13519 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13520 when this function is available.
13522 @item unsigned int __builtin_mips_get_fcsr (void)
13523 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13524 Get and set the contents of the floating-point control and status register
13525 (FPU control register 31). These functions are only available in hard-float
13526 code but can be called in both MIPS16 and non-MIPS16 contexts.
13528 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13529 register except the condition codes, which GCC assumes are preserved.
13532 @node MSP430 Built-in Functions
13533 @subsection MSP430 Built-in Functions
13535 GCC provides a couple of special builtin functions to aid in the
13536 writing of interrupt handlers in C.
13539 @item __bic_SR_register_on_exit (int @var{mask})
13540 This clears the indicated bits in the saved copy of the status register
13541 currently residing on the stack. This only works inside interrupt
13542 handlers and the changes to the status register will only take affect
13543 once the handler returns.
13545 @item __bis_SR_register_on_exit (int @var{mask})
13546 This sets the indicated bits in the saved copy of the status register
13547 currently residing on the stack. This only works inside interrupt
13548 handlers and the changes to the status register will only take affect
13549 once the handler returns.
13551 @item __delay_cycles (long long @var{cycles})
13552 This inserts an instruction sequence that takes exactly @var{cycles}
13553 cycles (between 0 and about 17E9) to complete. The inserted sequence
13554 may use jumps, loops, or no-ops, and does not interfere with any other
13555 instructions. Note that @var{cycles} must be a compile-time constant
13556 integer - that is, you must pass a number, not a variable that may be
13557 optimized to a constant later. The number of cycles delayed by this
13561 @node NDS32 Built-in Functions
13562 @subsection NDS32 Built-in Functions
13564 These built-in functions are available for the NDS32 target:
13566 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13567 Insert an ISYNC instruction into the instruction stream where
13568 @var{addr} is an instruction address for serialization.
13571 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13572 Insert an ISB instruction into the instruction stream.
13575 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13576 Return the content of a system register which is mapped by @var{sr}.
13579 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13580 Return the content of a user space register which is mapped by @var{usr}.
13583 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13584 Move the @var{value} to a system register which is mapped by @var{sr}.
13587 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13588 Move the @var{value} to a user space register which is mapped by @var{usr}.
13591 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13592 Enable global interrupt.
13595 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13596 Disable global interrupt.
13599 @node picoChip Built-in Functions
13600 @subsection picoChip Built-in Functions
13602 GCC provides an interface to selected machine instructions from the
13603 picoChip instruction set.
13606 @item int __builtin_sbc (int @var{value})
13607 Sign bit count. Return the number of consecutive bits in @var{value}
13608 that have the same value as the sign bit. The result is the number of
13609 leading sign bits minus one, giving the number of redundant sign bits in
13612 @item int __builtin_byteswap (int @var{value})
13613 Byte swap. Return the result of swapping the upper and lower bytes of
13616 @item int __builtin_brev (int @var{value})
13617 Bit reversal. Return the result of reversing the bits in
13618 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13621 @item int __builtin_adds (int @var{x}, int @var{y})
13622 Saturating addition. Return the result of adding @var{x} and @var{y},
13623 storing the value 32767 if the result overflows.
13625 @item int __builtin_subs (int @var{x}, int @var{y})
13626 Saturating subtraction. Return the result of subtracting @var{y} from
13627 @var{x}, storing the value @minus{}32768 if the result overflows.
13629 @item void __builtin_halt (void)
13630 Halt. The processor stops execution. This built-in is useful for
13631 implementing assertions.
13635 @node PowerPC Built-in Functions
13636 @subsection PowerPC Built-in Functions
13638 The following built-in functions are always available and can be used to
13639 check the PowerPC target platform type:
13641 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
13642 This function is a @code{nop} on the PowerPC platform and is included solely
13643 to maintain API compatibility with the x86 builtins.
13646 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
13647 This function returns a value of @code{1} if the run-time CPU is of type
13648 @var{cpuname} and returns @code{0} otherwise. The following CPU names can be
13653 IBM POWER9 Server CPU.
13655 IBM POWER8 Server CPU.
13657 IBM POWER7 Server CPU.
13659 IBM POWER6 Server CPU (RAW mode).
13661 IBM POWER6 Server CPU (Architected mode).
13663 IBM POWER5+ Server CPU.
13665 IBM POWER5 Server CPU.
13667 IBM 970 Server CPU (ie, Apple G5).
13669 IBM POWER4 Server CPU.
13671 IBM A2 64-bit Embedded CPU
13673 IBM PowerPC 476FP 32-bit Embedded CPU.
13675 IBM PowerPC 464 32-bit Embedded CPU.
13677 PowerPC 440 32-bit Embedded CPU.
13679 PowerPC 405 32-bit Embedded CPU.
13681 IBM PowerPC Cell Broadband Engine Architecture CPU.
13684 Here is an example:
13686 if (__builtin_cpu_is ("power8"))
13688 do_power8 (); // POWER8 specific implementation.
13692 do_generic (); // Generic implementation.
13697 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
13698 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
13699 feature @var{feature} and returns @code{0} otherwise. The following features can be
13704 4xx CPU has a Multiply Accumulator.
13706 CPU has a SIMD/Vector Unit.
13708 CPU supports ISA 2.05 (eg, POWER6)
13710 CPU supports ISA 2.06 (eg, POWER7)
13712 CPU supports ISA 2.07 (eg, POWER8)
13714 CPU supports ISA 3.00 (eg, POWER9)
13716 CPU supports the set of compatible performance monitoring events.
13718 CPU supports the Embedded ISA category.
13720 CPU has a CELL broadband engine.
13722 CPU has a decimal floating point unit.
13724 CPU supports the data stream control register.
13726 CPU supports event base branching.
13728 CPU has a SPE double precision floating point unit.
13730 CPU has a SPE single precision floating point unit.
13732 CPU has a floating point unit.
13734 CPU has hardware transaction memory instructions.
13736 Kernel aborts hardware transactions when a syscall is made.
13738 CPU supports icache snooping capabilities.
13740 CPU supports 128-bit IEEE binary floating point instructions.
13742 CPU supports the integer select instruction.
13744 CPU has a memory management unit.
13746 CPU does not have a timebase (eg, 601 and 403gx).
13748 CPU supports the PA Semi 6T CORE ISA.
13750 CPU supports ISA 2.00 (eg, POWER4)
13752 CPU supports ISA 2.02 (eg, POWER5)
13754 CPU supports ISA 2.03 (eg, POWER5+)
13756 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
13758 CPU supports 32-bit mode execution.
13760 CPU supports the old POWER ISA (eg, 601)
13762 CPU supports 64-bit mode execution.
13764 CPU supports a little-endian mode that uses address swizzling.
13766 CPU support simultaneous multi-threading.
13768 CPU has a signal processing extension unit.
13770 CPU supports the target address register.
13772 CPU supports true little-endian mode.
13774 CPU has unified I/D cache.
13776 CPU supports the vector cryptography instructions.
13778 CPU supports the vector-scalar extension.
13781 Here is an example:
13783 if (__builtin_cpu_supports ("fpu"))
13785 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
13789 dst = __fadd (src1, src2); // Software FP addition function.
13794 These built-in functions are available for the PowerPC family of
13797 float __builtin_recipdivf (float, float);
13798 float __builtin_rsqrtf (float);
13799 double __builtin_recipdiv (double, double);
13800 double __builtin_rsqrt (double);
13801 uint64_t __builtin_ppc_get_timebase ();
13802 unsigned long __builtin_ppc_mftb ();
13803 double __builtin_unpack_longdouble (long double, int);
13804 long double __builtin_pack_longdouble (double, double);
13807 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13808 @code{__builtin_rsqrtf} functions generate multiple instructions to
13809 implement the reciprocal sqrt functionality using reciprocal sqrt
13810 estimate instructions.
13812 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13813 functions generate multiple instructions to implement division using
13814 the reciprocal estimate instructions.
13816 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13817 functions generate instructions to read the Time Base Register. The
13818 @code{__builtin_ppc_get_timebase} function may generate multiple
13819 instructions and always returns the 64 bits of the Time Base Register.
13820 The @code{__builtin_ppc_mftb} function always generates one instruction and
13821 returns the Time Base Register value as an unsigned long, throwing away
13822 the most significant word on 32-bit environments.
13824 The following built-in functions are available for the PowerPC family
13825 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13826 or @option{-mpopcntd}):
13828 long __builtin_bpermd (long, long);
13829 int __builtin_divwe (int, int);
13830 int __builtin_divweo (int, int);
13831 unsigned int __builtin_divweu (unsigned int, unsigned int);
13832 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13833 long __builtin_divde (long, long);
13834 long __builtin_divdeo (long, long);
13835 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13836 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13837 unsigned int cdtbcd (unsigned int);
13838 unsigned int cbcdtd (unsigned int);
13839 unsigned int addg6s (unsigned int, unsigned int);
13842 The @code{__builtin_divde}, @code{__builtin_divdeo},
13843 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13844 64-bit environment support ISA 2.06 or later.
13846 The following built-in functions are available for the PowerPC family
13847 of processors when hardware decimal floating point
13848 (@option{-mhard-dfp}) is available:
13850 _Decimal64 __builtin_dxex (_Decimal64);
13851 _Decimal128 __builtin_dxexq (_Decimal128);
13852 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13853 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13854 _Decimal64 __builtin_denbcd (int, _Decimal64);
13855 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13856 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13857 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13858 _Decimal64 __builtin_dscli (_Decimal64, int);
13859 _Decimal128 __builtin_dscliq (_Decimal128, int);
13860 _Decimal64 __builtin_dscri (_Decimal64, int);
13861 _Decimal128 __builtin_dscriq (_Decimal128, int);
13862 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13863 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13866 The following built-in functions are available for the PowerPC family
13867 of processors when the Vector Scalar (vsx) instruction set is
13870 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13871 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13872 unsigned long long);
13875 @node PowerPC AltiVec/VSX Built-in Functions
13876 @subsection PowerPC AltiVec Built-in Functions
13878 GCC provides an interface for the PowerPC family of processors to access
13879 the AltiVec operations described in Motorola's AltiVec Programming
13880 Interface Manual. The interface is made available by including
13881 @code{<altivec.h>} and using @option{-maltivec} and
13882 @option{-mabi=altivec}. The interface supports the following vector
13886 vector unsigned char
13890 vector unsigned short
13891 vector signed short
13895 vector unsigned int
13901 If @option{-mvsx} is used the following additional vector types are
13905 vector unsigned long
13910 The long types are only implemented for 64-bit code generation, and
13911 the long type is only used in the floating point/integer conversion
13914 GCC's implementation of the high-level language interface available from
13915 C and C++ code differs from Motorola's documentation in several ways.
13920 A vector constant is a list of constant expressions within curly braces.
13923 A vector initializer requires no cast if the vector constant is of the
13924 same type as the variable it is initializing.
13927 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13928 vector type is the default signedness of the base type. The default
13929 varies depending on the operating system, so a portable program should
13930 always specify the signedness.
13933 Compiling with @option{-maltivec} adds keywords @code{__vector},
13934 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13935 @code{bool}. When compiling ISO C, the context-sensitive substitution
13936 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13937 disabled. To use them, you must include @code{<altivec.h>} instead.
13940 GCC allows using a @code{typedef} name as the type specifier for a
13944 For C, overloaded functions are implemented with macros so the following
13948 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13952 Since @code{vec_add} is a macro, the vector constant in the example
13953 is treated as four separate arguments. Wrap the entire argument in
13954 parentheses for this to work.
13957 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13958 Internally, GCC uses built-in functions to achieve the functionality in
13959 the aforementioned header file, but they are not supported and are
13960 subject to change without notice.
13962 The following interfaces are supported for the generic and specific
13963 AltiVec operations and the AltiVec predicates. In cases where there
13964 is a direct mapping between generic and specific operations, only the
13965 generic names are shown here, although the specific operations can also
13968 Arguments that are documented as @code{const int} require literal
13969 integral values within the range required for that operation.
13972 vector signed char vec_abs (vector signed char);
13973 vector signed short vec_abs (vector signed short);
13974 vector signed int vec_abs (vector signed int);
13975 vector float vec_abs (vector float);
13977 vector signed char vec_abss (vector signed char);
13978 vector signed short vec_abss (vector signed short);
13979 vector signed int vec_abss (vector signed int);
13981 vector signed char vec_add (vector bool char, vector signed char);
13982 vector signed char vec_add (vector signed char, vector bool char);
13983 vector signed char vec_add (vector signed char, vector signed char);
13984 vector unsigned char vec_add (vector bool char, vector unsigned char);
13985 vector unsigned char vec_add (vector unsigned char, vector bool char);
13986 vector unsigned char vec_add (vector unsigned char,
13987 vector unsigned char);
13988 vector signed short vec_add (vector bool short, vector signed short);
13989 vector signed short vec_add (vector signed short, vector bool short);
13990 vector signed short vec_add (vector signed short, vector signed short);
13991 vector unsigned short vec_add (vector bool short,
13992 vector unsigned short);
13993 vector unsigned short vec_add (vector unsigned short,
13994 vector bool short);
13995 vector unsigned short vec_add (vector unsigned short,
13996 vector unsigned short);
13997 vector signed int vec_add (vector bool int, vector signed int);
13998 vector signed int vec_add (vector signed int, vector bool int);
13999 vector signed int vec_add (vector signed int, vector signed int);
14000 vector unsigned int vec_add (vector bool int, vector unsigned int);
14001 vector unsigned int vec_add (vector unsigned int, vector bool int);
14002 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
14003 vector float vec_add (vector float, vector float);
14005 vector float vec_vaddfp (vector float, vector float);
14007 vector signed int vec_vadduwm (vector bool int, vector signed int);
14008 vector signed int vec_vadduwm (vector signed int, vector bool int);
14009 vector signed int vec_vadduwm (vector signed int, vector signed int);
14010 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
14011 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
14012 vector unsigned int vec_vadduwm (vector unsigned int,
14013 vector unsigned int);
14015 vector signed short vec_vadduhm (vector bool short,
14016 vector signed short);
14017 vector signed short vec_vadduhm (vector signed short,
14018 vector bool short);
14019 vector signed short vec_vadduhm (vector signed short,
14020 vector signed short);
14021 vector unsigned short vec_vadduhm (vector bool short,
14022 vector unsigned short);
14023 vector unsigned short vec_vadduhm (vector unsigned short,
14024 vector bool short);
14025 vector unsigned short vec_vadduhm (vector unsigned short,
14026 vector unsigned short);
14028 vector signed char vec_vaddubm (vector bool char, vector signed char);
14029 vector signed char vec_vaddubm (vector signed char, vector bool char);
14030 vector signed char vec_vaddubm (vector signed char, vector signed char);
14031 vector unsigned char vec_vaddubm (vector bool char,
14032 vector unsigned char);
14033 vector unsigned char vec_vaddubm (vector unsigned char,
14035 vector unsigned char vec_vaddubm (vector unsigned char,
14036 vector unsigned char);
14038 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
14040 vector unsigned char vec_adds (vector bool char, vector unsigned char);
14041 vector unsigned char vec_adds (vector unsigned char, vector bool char);
14042 vector unsigned char vec_adds (vector unsigned char,
14043 vector unsigned char);
14044 vector signed char vec_adds (vector bool char, vector signed char);
14045 vector signed char vec_adds (vector signed char, vector bool char);
14046 vector signed char vec_adds (vector signed char, vector signed char);
14047 vector unsigned short vec_adds (vector bool short,
14048 vector unsigned short);
14049 vector unsigned short vec_adds (vector unsigned short,
14050 vector bool short);
14051 vector unsigned short vec_adds (vector unsigned short,
14052 vector unsigned short);
14053 vector signed short vec_adds (vector bool short, vector signed short);
14054 vector signed short vec_adds (vector signed short, vector bool short);
14055 vector signed short vec_adds (vector signed short, vector signed short);
14056 vector unsigned int vec_adds (vector bool int, vector unsigned int);
14057 vector unsigned int vec_adds (vector unsigned int, vector bool int);
14058 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
14059 vector signed int vec_adds (vector bool int, vector signed int);
14060 vector signed int vec_adds (vector signed int, vector bool int);
14061 vector signed int vec_adds (vector signed int, vector signed int);
14063 vector signed int vec_vaddsws (vector bool int, vector signed int);
14064 vector signed int vec_vaddsws (vector signed int, vector bool int);
14065 vector signed int vec_vaddsws (vector signed int, vector signed int);
14067 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
14068 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
14069 vector unsigned int vec_vadduws (vector unsigned int,
14070 vector unsigned int);
14072 vector signed short vec_vaddshs (vector bool short,
14073 vector signed short);
14074 vector signed short vec_vaddshs (vector signed short,
14075 vector bool short);
14076 vector signed short vec_vaddshs (vector signed short,
14077 vector signed short);
14079 vector unsigned short vec_vadduhs (vector bool short,
14080 vector unsigned short);
14081 vector unsigned short vec_vadduhs (vector unsigned short,
14082 vector bool short);
14083 vector unsigned short vec_vadduhs (vector unsigned short,
14084 vector unsigned short);
14086 vector signed char vec_vaddsbs (vector bool char, vector signed char);
14087 vector signed char vec_vaddsbs (vector signed char, vector bool char);
14088 vector signed char vec_vaddsbs (vector signed char, vector signed char);
14090 vector unsigned char vec_vaddubs (vector bool char,
14091 vector unsigned char);
14092 vector unsigned char vec_vaddubs (vector unsigned char,
14094 vector unsigned char vec_vaddubs (vector unsigned char,
14095 vector unsigned char);
14097 vector float vec_and (vector float, vector float);
14098 vector float vec_and (vector float, vector bool int);
14099 vector float vec_and (vector bool int, vector float);
14100 vector bool int vec_and (vector bool int, vector bool int);
14101 vector signed int vec_and (vector bool int, vector signed int);
14102 vector signed int vec_and (vector signed int, vector bool int);
14103 vector signed int vec_and (vector signed int, vector signed int);
14104 vector unsigned int vec_and (vector bool int, vector unsigned int);
14105 vector unsigned int vec_and (vector unsigned int, vector bool int);
14106 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
14107 vector bool short vec_and (vector bool short, vector bool short);
14108 vector signed short vec_and (vector bool short, vector signed short);
14109 vector signed short vec_and (vector signed short, vector bool short);
14110 vector signed short vec_and (vector signed short, vector signed short);
14111 vector unsigned short vec_and (vector bool short,
14112 vector unsigned short);
14113 vector unsigned short vec_and (vector unsigned short,
14114 vector bool short);
14115 vector unsigned short vec_and (vector unsigned short,
14116 vector unsigned short);
14117 vector signed char vec_and (vector bool char, vector signed char);
14118 vector bool char vec_and (vector bool char, vector bool char);
14119 vector signed char vec_and (vector signed char, vector bool char);
14120 vector signed char vec_and (vector signed char, vector signed char);
14121 vector unsigned char vec_and (vector bool char, vector unsigned char);
14122 vector unsigned char vec_and (vector unsigned char, vector bool char);
14123 vector unsigned char vec_and (vector unsigned char,
14124 vector unsigned char);
14126 vector float vec_andc (vector float, vector float);
14127 vector float vec_andc (vector float, vector bool int);
14128 vector float vec_andc (vector bool int, vector float);
14129 vector bool int vec_andc (vector bool int, vector bool int);
14130 vector signed int vec_andc (vector bool int, vector signed int);
14131 vector signed int vec_andc (vector signed int, vector bool int);
14132 vector signed int vec_andc (vector signed int, vector signed int);
14133 vector unsigned int vec_andc (vector bool int, vector unsigned int);
14134 vector unsigned int vec_andc (vector unsigned int, vector bool int);
14135 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
14136 vector bool short vec_andc (vector bool short, vector bool short);
14137 vector signed short vec_andc (vector bool short, vector signed short);
14138 vector signed short vec_andc (vector signed short, vector bool short);
14139 vector signed short vec_andc (vector signed short, vector signed short);
14140 vector unsigned short vec_andc (vector bool short,
14141 vector unsigned short);
14142 vector unsigned short vec_andc (vector unsigned short,
14143 vector bool short);
14144 vector unsigned short vec_andc (vector unsigned short,
14145 vector unsigned short);
14146 vector signed char vec_andc (vector bool char, vector signed char);
14147 vector bool char vec_andc (vector bool char, vector bool char);
14148 vector signed char vec_andc (vector signed char, vector bool char);
14149 vector signed char vec_andc (vector signed char, vector signed char);
14150 vector unsigned char vec_andc (vector bool char, vector unsigned char);
14151 vector unsigned char vec_andc (vector unsigned char, vector bool char);
14152 vector unsigned char vec_andc (vector unsigned char,
14153 vector unsigned char);
14155 vector unsigned char vec_avg (vector unsigned char,
14156 vector unsigned char);
14157 vector signed char vec_avg (vector signed char, vector signed char);
14158 vector unsigned short vec_avg (vector unsigned short,
14159 vector unsigned short);
14160 vector signed short vec_avg (vector signed short, vector signed short);
14161 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
14162 vector signed int vec_avg (vector signed int, vector signed int);
14164 vector signed int vec_vavgsw (vector signed int, vector signed int);
14166 vector unsigned int vec_vavguw (vector unsigned int,
14167 vector unsigned int);
14169 vector signed short vec_vavgsh (vector signed short,
14170 vector signed short);
14172 vector unsigned short vec_vavguh (vector unsigned short,
14173 vector unsigned short);
14175 vector signed char vec_vavgsb (vector signed char, vector signed char);
14177 vector unsigned char vec_vavgub (vector unsigned char,
14178 vector unsigned char);
14180 vector float vec_copysign (vector float);
14182 vector float vec_ceil (vector float);
14184 vector signed int vec_cmpb (vector float, vector float);
14186 vector bool char vec_cmpeq (vector signed char, vector signed char);
14187 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
14188 vector bool short vec_cmpeq (vector signed short, vector signed short);
14189 vector bool short vec_cmpeq (vector unsigned short,
14190 vector unsigned short);
14191 vector bool int vec_cmpeq (vector signed int, vector signed int);
14192 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
14193 vector bool int vec_cmpeq (vector float, vector float);
14195 vector bool int vec_vcmpeqfp (vector float, vector float);
14197 vector bool int vec_vcmpequw (vector signed int, vector signed int);
14198 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
14200 vector bool short vec_vcmpequh (vector signed short,
14201 vector signed short);
14202 vector bool short vec_vcmpequh (vector unsigned short,
14203 vector unsigned short);
14205 vector bool char vec_vcmpequb (vector signed char, vector signed char);
14206 vector bool char vec_vcmpequb (vector unsigned char,
14207 vector unsigned char);
14209 vector bool int vec_cmpge (vector float, vector float);
14211 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
14212 vector bool char vec_cmpgt (vector signed char, vector signed char);
14213 vector bool short vec_cmpgt (vector unsigned short,
14214 vector unsigned short);
14215 vector bool short vec_cmpgt (vector signed short, vector signed short);
14216 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
14217 vector bool int vec_cmpgt (vector signed int, vector signed int);
14218 vector bool int vec_cmpgt (vector float, vector float);
14220 vector bool int vec_vcmpgtfp (vector float, vector float);
14222 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
14224 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
14226 vector bool short vec_vcmpgtsh (vector signed short,
14227 vector signed short);
14229 vector bool short vec_vcmpgtuh (vector unsigned short,
14230 vector unsigned short);
14232 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
14234 vector bool char vec_vcmpgtub (vector unsigned char,
14235 vector unsigned char);
14237 vector bool int vec_cmple (vector float, vector float);
14239 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
14240 vector bool char vec_cmplt (vector signed char, vector signed char);
14241 vector bool short vec_cmplt (vector unsigned short,
14242 vector unsigned short);
14243 vector bool short vec_cmplt (vector signed short, vector signed short);
14244 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
14245 vector bool int vec_cmplt (vector signed int, vector signed int);
14246 vector bool int vec_cmplt (vector float, vector float);
14248 vector float vec_cpsgn (vector float, vector float);
14250 vector float vec_ctf (vector unsigned int, const int);
14251 vector float vec_ctf (vector signed int, const int);
14252 vector double vec_ctf (vector unsigned long, const int);
14253 vector double vec_ctf (vector signed long, const int);
14255 vector float vec_vcfsx (vector signed int, const int);
14257 vector float vec_vcfux (vector unsigned int, const int);
14259 vector signed int vec_cts (vector float, const int);
14260 vector signed long vec_cts (vector double, const int);
14262 vector unsigned int vec_ctu (vector float, const int);
14263 vector unsigned long vec_ctu (vector double, const int);
14265 void vec_dss (const int);
14267 void vec_dssall (void);
14269 void vec_dst (const vector unsigned char *, int, const int);
14270 void vec_dst (const vector signed char *, int, const int);
14271 void vec_dst (const vector bool char *, int, const int);
14272 void vec_dst (const vector unsigned short *, int, const int);
14273 void vec_dst (const vector signed short *, int, const int);
14274 void vec_dst (const vector bool short *, int, const int);
14275 void vec_dst (const vector pixel *, int, const int);
14276 void vec_dst (const vector unsigned int *, int, const int);
14277 void vec_dst (const vector signed int *, int, const int);
14278 void vec_dst (const vector bool int *, int, const int);
14279 void vec_dst (const vector float *, int, const int);
14280 void vec_dst (const unsigned char *, int, const int);
14281 void vec_dst (const signed char *, int, const int);
14282 void vec_dst (const unsigned short *, int, const int);
14283 void vec_dst (const short *, int, const int);
14284 void vec_dst (const unsigned int *, int, const int);
14285 void vec_dst (const int *, int, const int);
14286 void vec_dst (const unsigned long *, int, const int);
14287 void vec_dst (const long *, int, const int);
14288 void vec_dst (const float *, int, const int);
14290 void vec_dstst (const vector unsigned char *, int, const int);
14291 void vec_dstst (const vector signed char *, int, const int);
14292 void vec_dstst (const vector bool char *, int, const int);
14293 void vec_dstst (const vector unsigned short *, int, const int);
14294 void vec_dstst (const vector signed short *, int, const int);
14295 void vec_dstst (const vector bool short *, int, const int);
14296 void vec_dstst (const vector pixel *, int, const int);
14297 void vec_dstst (const vector unsigned int *, int, const int);
14298 void vec_dstst (const vector signed int *, int, const int);
14299 void vec_dstst (const vector bool int *, int, const int);
14300 void vec_dstst (const vector float *, int, const int);
14301 void vec_dstst (const unsigned char *, int, const int);
14302 void vec_dstst (const signed char *, int, const int);
14303 void vec_dstst (const unsigned short *, int, const int);
14304 void vec_dstst (const short *, int, const int);
14305 void vec_dstst (const unsigned int *, int, const int);
14306 void vec_dstst (const int *, int, const int);
14307 void vec_dstst (const unsigned long *, int, const int);
14308 void vec_dstst (const long *, int, const int);
14309 void vec_dstst (const float *, int, const int);
14311 void vec_dststt (const vector unsigned char *, int, const int);
14312 void vec_dststt (const vector signed char *, int, const int);
14313 void vec_dststt (const vector bool char *, int, const int);
14314 void vec_dststt (const vector unsigned short *, int, const int);
14315 void vec_dststt (const vector signed short *, int, const int);
14316 void vec_dststt (const vector bool short *, int, const int);
14317 void vec_dststt (const vector pixel *, int, const int);
14318 void vec_dststt (const vector unsigned int *, int, const int);
14319 void vec_dststt (const vector signed int *, int, const int);
14320 void vec_dststt (const vector bool int *, int, const int);
14321 void vec_dststt (const vector float *, int, const int);
14322 void vec_dststt (const unsigned char *, int, const int);
14323 void vec_dststt (const signed char *, int, const int);
14324 void vec_dststt (const unsigned short *, int, const int);
14325 void vec_dststt (const short *, int, const int);
14326 void vec_dststt (const unsigned int *, int, const int);
14327 void vec_dststt (const int *, int, const int);
14328 void vec_dststt (const unsigned long *, int, const int);
14329 void vec_dststt (const long *, int, const int);
14330 void vec_dststt (const float *, int, const int);
14332 void vec_dstt (const vector unsigned char *, int, const int);
14333 void vec_dstt (const vector signed char *, int, const int);
14334 void vec_dstt (const vector bool char *, int, const int);
14335 void vec_dstt (const vector unsigned short *, int, const int);
14336 void vec_dstt (const vector signed short *, int, const int);
14337 void vec_dstt (const vector bool short *, int, const int);
14338 void vec_dstt (const vector pixel *, int, const int);
14339 void vec_dstt (const vector unsigned int *, int, const int);
14340 void vec_dstt (const vector signed int *, int, const int);
14341 void vec_dstt (const vector bool int *, int, const int);
14342 void vec_dstt (const vector float *, int, const int);
14343 void vec_dstt (const unsigned char *, int, const int);
14344 void vec_dstt (const signed char *, int, const int);
14345 void vec_dstt (const unsigned short *, int, const int);
14346 void vec_dstt (const short *, int, const int);
14347 void vec_dstt (const unsigned int *, int, const int);
14348 void vec_dstt (const int *, int, const int);
14349 void vec_dstt (const unsigned long *, int, const int);
14350 void vec_dstt (const long *, int, const int);
14351 void vec_dstt (const float *, int, const int);
14353 vector float vec_expte (vector float);
14355 vector float vec_floor (vector float);
14357 vector float vec_ld (int, const vector float *);
14358 vector float vec_ld (int, const float *);
14359 vector bool int vec_ld (int, const vector bool int *);
14360 vector signed int vec_ld (int, const vector signed int *);
14361 vector signed int vec_ld (int, const int *);
14362 vector signed int vec_ld (int, const long *);
14363 vector unsigned int vec_ld (int, const vector unsigned int *);
14364 vector unsigned int vec_ld (int, const unsigned int *);
14365 vector unsigned int vec_ld (int, const unsigned long *);
14366 vector bool short vec_ld (int, const vector bool short *);
14367 vector pixel vec_ld (int, const vector pixel *);
14368 vector signed short vec_ld (int, const vector signed short *);
14369 vector signed short vec_ld (int, const short *);
14370 vector unsigned short vec_ld (int, const vector unsigned short *);
14371 vector unsigned short vec_ld (int, const unsigned short *);
14372 vector bool char vec_ld (int, const vector bool char *);
14373 vector signed char vec_ld (int, const vector signed char *);
14374 vector signed char vec_ld (int, const signed char *);
14375 vector unsigned char vec_ld (int, const vector unsigned char *);
14376 vector unsigned char vec_ld (int, const unsigned char *);
14378 vector signed char vec_lde (int, const signed char *);
14379 vector unsigned char vec_lde (int, const unsigned char *);
14380 vector signed short vec_lde (int, const short *);
14381 vector unsigned short vec_lde (int, const unsigned short *);
14382 vector float vec_lde (int, const float *);
14383 vector signed int vec_lde (int, const int *);
14384 vector unsigned int vec_lde (int, const unsigned int *);
14385 vector signed int vec_lde (int, const long *);
14386 vector unsigned int vec_lde (int, const unsigned long *);
14388 vector float vec_lvewx (int, float *);
14389 vector signed int vec_lvewx (int, int *);
14390 vector unsigned int vec_lvewx (int, unsigned int *);
14391 vector signed int vec_lvewx (int, long *);
14392 vector unsigned int vec_lvewx (int, unsigned long *);
14394 vector signed short vec_lvehx (int, short *);
14395 vector unsigned short vec_lvehx (int, unsigned short *);
14397 vector signed char vec_lvebx (int, char *);
14398 vector unsigned char vec_lvebx (int, unsigned char *);
14400 vector float vec_ldl (int, const vector float *);
14401 vector float vec_ldl (int, const float *);
14402 vector bool int vec_ldl (int, const vector bool int *);
14403 vector signed int vec_ldl (int, const vector signed int *);
14404 vector signed int vec_ldl (int, const int *);
14405 vector signed int vec_ldl (int, const long *);
14406 vector unsigned int vec_ldl (int, const vector unsigned int *);
14407 vector unsigned int vec_ldl (int, const unsigned int *);
14408 vector unsigned int vec_ldl (int, const unsigned long *);
14409 vector bool short vec_ldl (int, const vector bool short *);
14410 vector pixel vec_ldl (int, const vector pixel *);
14411 vector signed short vec_ldl (int, const vector signed short *);
14412 vector signed short vec_ldl (int, const short *);
14413 vector unsigned short vec_ldl (int, const vector unsigned short *);
14414 vector unsigned short vec_ldl (int, const unsigned short *);
14415 vector bool char vec_ldl (int, const vector bool char *);
14416 vector signed char vec_ldl (int, const vector signed char *);
14417 vector signed char vec_ldl (int, const signed char *);
14418 vector unsigned char vec_ldl (int, const vector unsigned char *);
14419 vector unsigned char vec_ldl (int, const unsigned char *);
14421 vector float vec_loge (vector float);
14423 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
14424 vector unsigned char vec_lvsl (int, const volatile signed char *);
14425 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
14426 vector unsigned char vec_lvsl (int, const volatile short *);
14427 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
14428 vector unsigned char vec_lvsl (int, const volatile int *);
14429 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
14430 vector unsigned char vec_lvsl (int, const volatile long *);
14431 vector unsigned char vec_lvsl (int, const volatile float *);
14433 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
14434 vector unsigned char vec_lvsr (int, const volatile signed char *);
14435 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
14436 vector unsigned char vec_lvsr (int, const volatile short *);
14437 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
14438 vector unsigned char vec_lvsr (int, const volatile int *);
14439 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
14440 vector unsigned char vec_lvsr (int, const volatile long *);
14441 vector unsigned char vec_lvsr (int, const volatile float *);
14443 vector float vec_madd (vector float, vector float, vector float);
14445 vector signed short vec_madds (vector signed short,
14446 vector signed short,
14447 vector signed short);
14449 vector unsigned char vec_max (vector bool char, vector unsigned char);
14450 vector unsigned char vec_max (vector unsigned char, vector bool char);
14451 vector unsigned char vec_max (vector unsigned char,
14452 vector unsigned char);
14453 vector signed char vec_max (vector bool char, vector signed char);
14454 vector signed char vec_max (vector signed char, vector bool char);
14455 vector signed char vec_max (vector signed char, vector signed char);
14456 vector unsigned short vec_max (vector bool short,
14457 vector unsigned short);
14458 vector unsigned short vec_max (vector unsigned short,
14459 vector bool short);
14460 vector unsigned short vec_max (vector unsigned short,
14461 vector unsigned short);
14462 vector signed short vec_max (vector bool short, vector signed short);
14463 vector signed short vec_max (vector signed short, vector bool short);
14464 vector signed short vec_max (vector signed short, vector signed short);
14465 vector unsigned int vec_max (vector bool int, vector unsigned int);
14466 vector unsigned int vec_max (vector unsigned int, vector bool int);
14467 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
14468 vector signed int vec_max (vector bool int, vector signed int);
14469 vector signed int vec_max (vector signed int, vector bool int);
14470 vector signed int vec_max (vector signed int, vector signed int);
14471 vector float vec_max (vector float, vector float);
14473 vector float vec_vmaxfp (vector float, vector float);
14475 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14476 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14477 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14479 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14480 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14481 vector unsigned int vec_vmaxuw (vector unsigned int,
14482 vector unsigned int);
14484 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14485 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14486 vector signed short vec_vmaxsh (vector signed short,
14487 vector signed short);
14489 vector unsigned short vec_vmaxuh (vector bool short,
14490 vector unsigned short);
14491 vector unsigned short vec_vmaxuh (vector unsigned short,
14492 vector bool short);
14493 vector unsigned short vec_vmaxuh (vector unsigned short,
14494 vector unsigned short);
14496 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14497 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14498 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14500 vector unsigned char vec_vmaxub (vector bool char,
14501 vector unsigned char);
14502 vector unsigned char vec_vmaxub (vector unsigned char,
14504 vector unsigned char vec_vmaxub (vector unsigned char,
14505 vector unsigned char);
14507 vector bool char vec_mergeh (vector bool char, vector bool char);
14508 vector signed char vec_mergeh (vector signed char, vector signed char);
14509 vector unsigned char vec_mergeh (vector unsigned char,
14510 vector unsigned char);
14511 vector bool short vec_mergeh (vector bool short, vector bool short);
14512 vector pixel vec_mergeh (vector pixel, vector pixel);
14513 vector signed short vec_mergeh (vector signed short,
14514 vector signed short);
14515 vector unsigned short vec_mergeh (vector unsigned short,
14516 vector unsigned short);
14517 vector float vec_mergeh (vector float, vector float);
14518 vector bool int vec_mergeh (vector bool int, vector bool int);
14519 vector signed int vec_mergeh (vector signed int, vector signed int);
14520 vector unsigned int vec_mergeh (vector unsigned int,
14521 vector unsigned int);
14523 vector float vec_vmrghw (vector float, vector float);
14524 vector bool int vec_vmrghw (vector bool int, vector bool int);
14525 vector signed int vec_vmrghw (vector signed int, vector signed int);
14526 vector unsigned int vec_vmrghw (vector unsigned int,
14527 vector unsigned int);
14529 vector bool short vec_vmrghh (vector bool short, vector bool short);
14530 vector signed short vec_vmrghh (vector signed short,
14531 vector signed short);
14532 vector unsigned short vec_vmrghh (vector unsigned short,
14533 vector unsigned short);
14534 vector pixel vec_vmrghh (vector pixel, vector pixel);
14536 vector bool char vec_vmrghb (vector bool char, vector bool char);
14537 vector signed char vec_vmrghb (vector signed char, vector signed char);
14538 vector unsigned char vec_vmrghb (vector unsigned char,
14539 vector unsigned char);
14541 vector bool char vec_mergel (vector bool char, vector bool char);
14542 vector signed char vec_mergel (vector signed char, vector signed char);
14543 vector unsigned char vec_mergel (vector unsigned char,
14544 vector unsigned char);
14545 vector bool short vec_mergel (vector bool short, vector bool short);
14546 vector pixel vec_mergel (vector pixel, vector pixel);
14547 vector signed short vec_mergel (vector signed short,
14548 vector signed short);
14549 vector unsigned short vec_mergel (vector unsigned short,
14550 vector unsigned short);
14551 vector float vec_mergel (vector float, vector float);
14552 vector bool int vec_mergel (vector bool int, vector bool int);
14553 vector signed int vec_mergel (vector signed int, vector signed int);
14554 vector unsigned int vec_mergel (vector unsigned int,
14555 vector unsigned int);
14557 vector float vec_vmrglw (vector float, vector float);
14558 vector signed int vec_vmrglw (vector signed int, vector signed int);
14559 vector unsigned int vec_vmrglw (vector unsigned int,
14560 vector unsigned int);
14561 vector bool int vec_vmrglw (vector bool int, vector bool int);
14563 vector bool short vec_vmrglh (vector bool short, vector bool short);
14564 vector signed short vec_vmrglh (vector signed short,
14565 vector signed short);
14566 vector unsigned short vec_vmrglh (vector unsigned short,
14567 vector unsigned short);
14568 vector pixel vec_vmrglh (vector pixel, vector pixel);
14570 vector bool char vec_vmrglb (vector bool char, vector bool char);
14571 vector signed char vec_vmrglb (vector signed char, vector signed char);
14572 vector unsigned char vec_vmrglb (vector unsigned char,
14573 vector unsigned char);
14575 vector unsigned short vec_mfvscr (void);
14577 vector unsigned char vec_min (vector bool char, vector unsigned char);
14578 vector unsigned char vec_min (vector unsigned char, vector bool char);
14579 vector unsigned char vec_min (vector unsigned char,
14580 vector unsigned char);
14581 vector signed char vec_min (vector bool char, vector signed char);
14582 vector signed char vec_min (vector signed char, vector bool char);
14583 vector signed char vec_min (vector signed char, vector signed char);
14584 vector unsigned short vec_min (vector bool short,
14585 vector unsigned short);
14586 vector unsigned short vec_min (vector unsigned short,
14587 vector bool short);
14588 vector unsigned short vec_min (vector unsigned short,
14589 vector unsigned short);
14590 vector signed short vec_min (vector bool short, vector signed short);
14591 vector signed short vec_min (vector signed short, vector bool short);
14592 vector signed short vec_min (vector signed short, vector signed short);
14593 vector unsigned int vec_min (vector bool int, vector unsigned int);
14594 vector unsigned int vec_min (vector unsigned int, vector bool int);
14595 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14596 vector signed int vec_min (vector bool int, vector signed int);
14597 vector signed int vec_min (vector signed int, vector bool int);
14598 vector signed int vec_min (vector signed int, vector signed int);
14599 vector float vec_min (vector float, vector float);
14601 vector float vec_vminfp (vector float, vector float);
14603 vector signed int vec_vminsw (vector bool int, vector signed int);
14604 vector signed int vec_vminsw (vector signed int, vector bool int);
14605 vector signed int vec_vminsw (vector signed int, vector signed int);
14607 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14608 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14609 vector unsigned int vec_vminuw (vector unsigned int,
14610 vector unsigned int);
14612 vector signed short vec_vminsh (vector bool short, vector signed short);
14613 vector signed short vec_vminsh (vector signed short, vector bool short);
14614 vector signed short vec_vminsh (vector signed short,
14615 vector signed short);
14617 vector unsigned short vec_vminuh (vector bool short,
14618 vector unsigned short);
14619 vector unsigned short vec_vminuh (vector unsigned short,
14620 vector bool short);
14621 vector unsigned short vec_vminuh (vector unsigned short,
14622 vector unsigned short);
14624 vector signed char vec_vminsb (vector bool char, vector signed char);
14625 vector signed char vec_vminsb (vector signed char, vector bool char);
14626 vector signed char vec_vminsb (vector signed char, vector signed char);
14628 vector unsigned char vec_vminub (vector bool char,
14629 vector unsigned char);
14630 vector unsigned char vec_vminub (vector unsigned char,
14632 vector unsigned char vec_vminub (vector unsigned char,
14633 vector unsigned char);
14635 vector signed short vec_mladd (vector signed short,
14636 vector signed short,
14637 vector signed short);
14638 vector signed short vec_mladd (vector signed short,
14639 vector unsigned short,
14640 vector unsigned short);
14641 vector signed short vec_mladd (vector unsigned short,
14642 vector signed short,
14643 vector signed short);
14644 vector unsigned short vec_mladd (vector unsigned short,
14645 vector unsigned short,
14646 vector unsigned short);
14648 vector signed short vec_mradds (vector signed short,
14649 vector signed short,
14650 vector signed short);
14652 vector unsigned int vec_msum (vector unsigned char,
14653 vector unsigned char,
14654 vector unsigned int);
14655 vector signed int vec_msum (vector signed char,
14656 vector unsigned char,
14657 vector signed int);
14658 vector unsigned int vec_msum (vector unsigned short,
14659 vector unsigned short,
14660 vector unsigned int);
14661 vector signed int vec_msum (vector signed short,
14662 vector signed short,
14663 vector signed int);
14665 vector signed int vec_vmsumshm (vector signed short,
14666 vector signed short,
14667 vector signed int);
14669 vector unsigned int vec_vmsumuhm (vector unsigned short,
14670 vector unsigned short,
14671 vector unsigned int);
14673 vector signed int vec_vmsummbm (vector signed char,
14674 vector unsigned char,
14675 vector signed int);
14677 vector unsigned int vec_vmsumubm (vector unsigned char,
14678 vector unsigned char,
14679 vector unsigned int);
14681 vector unsigned int vec_msums (vector unsigned short,
14682 vector unsigned short,
14683 vector unsigned int);
14684 vector signed int vec_msums (vector signed short,
14685 vector signed short,
14686 vector signed int);
14688 vector signed int vec_vmsumshs (vector signed short,
14689 vector signed short,
14690 vector signed int);
14692 vector unsigned int vec_vmsumuhs (vector unsigned short,
14693 vector unsigned short,
14694 vector unsigned int);
14696 void vec_mtvscr (vector signed int);
14697 void vec_mtvscr (vector unsigned int);
14698 void vec_mtvscr (vector bool int);
14699 void vec_mtvscr (vector signed short);
14700 void vec_mtvscr (vector unsigned short);
14701 void vec_mtvscr (vector bool short);
14702 void vec_mtvscr (vector pixel);
14703 void vec_mtvscr (vector signed char);
14704 void vec_mtvscr (vector unsigned char);
14705 void vec_mtvscr (vector bool char);
14707 vector unsigned short vec_mule (vector unsigned char,
14708 vector unsigned char);
14709 vector signed short vec_mule (vector signed char,
14710 vector signed char);
14711 vector unsigned int vec_mule (vector unsigned short,
14712 vector unsigned short);
14713 vector signed int vec_mule (vector signed short, vector signed short);
14715 vector signed int vec_vmulesh (vector signed short,
14716 vector signed short);
14718 vector unsigned int vec_vmuleuh (vector unsigned short,
14719 vector unsigned short);
14721 vector signed short vec_vmulesb (vector signed char,
14722 vector signed char);
14724 vector unsigned short vec_vmuleub (vector unsigned char,
14725 vector unsigned char);
14727 vector unsigned short vec_mulo (vector unsigned char,
14728 vector unsigned char);
14729 vector signed short vec_mulo (vector signed char, vector signed char);
14730 vector unsigned int vec_mulo (vector unsigned short,
14731 vector unsigned short);
14732 vector signed int vec_mulo (vector signed short, vector signed short);
14734 vector signed int vec_vmulosh (vector signed short,
14735 vector signed short);
14737 vector unsigned int vec_vmulouh (vector unsigned short,
14738 vector unsigned short);
14740 vector signed short vec_vmulosb (vector signed char,
14741 vector signed char);
14743 vector unsigned short vec_vmuloub (vector unsigned char,
14744 vector unsigned char);
14746 vector float vec_nmsub (vector float, vector float, vector float);
14748 vector float vec_nor (vector float, vector float);
14749 vector signed int vec_nor (vector signed int, vector signed int);
14750 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14751 vector bool int vec_nor (vector bool int, vector bool int);
14752 vector signed short vec_nor (vector signed short, vector signed short);
14753 vector unsigned short vec_nor (vector unsigned short,
14754 vector unsigned short);
14755 vector bool short vec_nor (vector bool short, vector bool short);
14756 vector signed char vec_nor (vector signed char, vector signed char);
14757 vector unsigned char vec_nor (vector unsigned char,
14758 vector unsigned char);
14759 vector bool char vec_nor (vector bool char, vector bool char);
14761 vector float vec_or (vector float, vector float);
14762 vector float vec_or (vector float, vector bool int);
14763 vector float vec_or (vector bool int, vector float);
14764 vector bool int vec_or (vector bool int, vector bool int);
14765 vector signed int vec_or (vector bool int, vector signed int);
14766 vector signed int vec_or (vector signed int, vector bool int);
14767 vector signed int vec_or (vector signed int, vector signed int);
14768 vector unsigned int vec_or (vector bool int, vector unsigned int);
14769 vector unsigned int vec_or (vector unsigned int, vector bool int);
14770 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14771 vector bool short vec_or (vector bool short, vector bool short);
14772 vector signed short vec_or (vector bool short, vector signed short);
14773 vector signed short vec_or (vector signed short, vector bool short);
14774 vector signed short vec_or (vector signed short, vector signed short);
14775 vector unsigned short vec_or (vector bool short, vector unsigned short);
14776 vector unsigned short vec_or (vector unsigned short, vector bool short);
14777 vector unsigned short vec_or (vector unsigned short,
14778 vector unsigned short);
14779 vector signed char vec_or (vector bool char, vector signed char);
14780 vector bool char vec_or (vector bool char, vector bool char);
14781 vector signed char vec_or (vector signed char, vector bool char);
14782 vector signed char vec_or (vector signed char, vector signed char);
14783 vector unsigned char vec_or (vector bool char, vector unsigned char);
14784 vector unsigned char vec_or (vector unsigned char, vector bool char);
14785 vector unsigned char vec_or (vector unsigned char,
14786 vector unsigned char);
14788 vector signed char vec_pack (vector signed short, vector signed short);
14789 vector unsigned char vec_pack (vector unsigned short,
14790 vector unsigned short);
14791 vector bool char vec_pack (vector bool short, vector bool short);
14792 vector signed short vec_pack (vector signed int, vector signed int);
14793 vector unsigned short vec_pack (vector unsigned int,
14794 vector unsigned int);
14795 vector bool short vec_pack (vector bool int, vector bool int);
14797 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14798 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14799 vector unsigned short vec_vpkuwum (vector unsigned int,
14800 vector unsigned int);
14802 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14803 vector signed char vec_vpkuhum (vector signed short,
14804 vector signed short);
14805 vector unsigned char vec_vpkuhum (vector unsigned short,
14806 vector unsigned short);
14808 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14810 vector unsigned char vec_packs (vector unsigned short,
14811 vector unsigned short);
14812 vector signed char vec_packs (vector signed short, vector signed short);
14813 vector unsigned short vec_packs (vector unsigned int,
14814 vector unsigned int);
14815 vector signed short vec_packs (vector signed int, vector signed int);
14817 vector signed short vec_vpkswss (vector signed int, vector signed int);
14819 vector unsigned short vec_vpkuwus (vector unsigned int,
14820 vector unsigned int);
14822 vector signed char vec_vpkshss (vector signed short,
14823 vector signed short);
14825 vector unsigned char vec_vpkuhus (vector unsigned short,
14826 vector unsigned short);
14828 vector unsigned char vec_packsu (vector unsigned short,
14829 vector unsigned short);
14830 vector unsigned char vec_packsu (vector signed short,
14831 vector signed short);
14832 vector unsigned short vec_packsu (vector unsigned int,
14833 vector unsigned int);
14834 vector unsigned short vec_packsu (vector signed int, vector signed int);
14836 vector unsigned short vec_vpkswus (vector signed int,
14837 vector signed int);
14839 vector unsigned char vec_vpkshus (vector signed short,
14840 vector signed short);
14842 vector float vec_perm (vector float,
14844 vector unsigned char);
14845 vector signed int vec_perm (vector signed int,
14847 vector unsigned char);
14848 vector unsigned int vec_perm (vector unsigned int,
14849 vector unsigned int,
14850 vector unsigned char);
14851 vector bool int vec_perm (vector bool int,
14853 vector unsigned char);
14854 vector signed short vec_perm (vector signed short,
14855 vector signed short,
14856 vector unsigned char);
14857 vector unsigned short vec_perm (vector unsigned short,
14858 vector unsigned short,
14859 vector unsigned char);
14860 vector bool short vec_perm (vector bool short,
14862 vector unsigned char);
14863 vector pixel vec_perm (vector pixel,
14865 vector unsigned char);
14866 vector signed char vec_perm (vector signed char,
14867 vector signed char,
14868 vector unsigned char);
14869 vector unsigned char vec_perm (vector unsigned char,
14870 vector unsigned char,
14871 vector unsigned char);
14872 vector bool char vec_perm (vector bool char,
14874 vector unsigned char);
14876 vector float vec_re (vector float);
14878 vector signed char vec_rl (vector signed char,
14879 vector unsigned char);
14880 vector unsigned char vec_rl (vector unsigned char,
14881 vector unsigned char);
14882 vector signed short vec_rl (vector signed short, vector unsigned short);
14883 vector unsigned short vec_rl (vector unsigned short,
14884 vector unsigned short);
14885 vector signed int vec_rl (vector signed int, vector unsigned int);
14886 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14888 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14889 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14891 vector signed short vec_vrlh (vector signed short,
14892 vector unsigned short);
14893 vector unsigned short vec_vrlh (vector unsigned short,
14894 vector unsigned short);
14896 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14897 vector unsigned char vec_vrlb (vector unsigned char,
14898 vector unsigned char);
14900 vector float vec_round (vector float);
14902 vector float vec_recip (vector float, vector float);
14904 vector float vec_rsqrt (vector float);
14906 vector float vec_rsqrte (vector float);
14908 vector float vec_sel (vector float, vector float, vector bool int);
14909 vector float vec_sel (vector float, vector float, vector unsigned int);
14910 vector signed int vec_sel (vector signed int,
14913 vector signed int vec_sel (vector signed int,
14915 vector unsigned int);
14916 vector unsigned int vec_sel (vector unsigned int,
14917 vector unsigned int,
14919 vector unsigned int vec_sel (vector unsigned int,
14920 vector unsigned int,
14921 vector unsigned int);
14922 vector bool int vec_sel (vector bool int,
14925 vector bool int vec_sel (vector bool int,
14927 vector unsigned int);
14928 vector signed short vec_sel (vector signed short,
14929 vector signed short,
14930 vector bool short);
14931 vector signed short vec_sel (vector signed short,
14932 vector signed short,
14933 vector unsigned short);
14934 vector unsigned short vec_sel (vector unsigned short,
14935 vector unsigned short,
14936 vector bool short);
14937 vector unsigned short vec_sel (vector unsigned short,
14938 vector unsigned short,
14939 vector unsigned short);
14940 vector bool short vec_sel (vector bool short,
14942 vector bool short);
14943 vector bool short vec_sel (vector bool short,
14945 vector unsigned short);
14946 vector signed char vec_sel (vector signed char,
14947 vector signed char,
14949 vector signed char vec_sel (vector signed char,
14950 vector signed char,
14951 vector unsigned char);
14952 vector unsigned char vec_sel (vector unsigned char,
14953 vector unsigned char,
14955 vector unsigned char vec_sel (vector unsigned char,
14956 vector unsigned char,
14957 vector unsigned char);
14958 vector bool char vec_sel (vector bool char,
14961 vector bool char vec_sel (vector bool char,
14963 vector unsigned char);
14965 vector signed char vec_sl (vector signed char,
14966 vector unsigned char);
14967 vector unsigned char vec_sl (vector unsigned char,
14968 vector unsigned char);
14969 vector signed short vec_sl (vector signed short, vector unsigned short);
14970 vector unsigned short vec_sl (vector unsigned short,
14971 vector unsigned short);
14972 vector signed int vec_sl (vector signed int, vector unsigned int);
14973 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14975 vector signed int vec_vslw (vector signed int, vector unsigned int);
14976 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14978 vector signed short vec_vslh (vector signed short,
14979 vector unsigned short);
14980 vector unsigned short vec_vslh (vector unsigned short,
14981 vector unsigned short);
14983 vector signed char vec_vslb (vector signed char, vector unsigned char);
14984 vector unsigned char vec_vslb (vector unsigned char,
14985 vector unsigned char);
14987 vector float vec_sld (vector float, vector float, const int);
14988 vector signed int vec_sld (vector signed int,
14991 vector unsigned int vec_sld (vector unsigned int,
14992 vector unsigned int,
14994 vector bool int vec_sld (vector bool int,
14997 vector signed short vec_sld (vector signed short,
14998 vector signed short,
15000 vector unsigned short vec_sld (vector unsigned short,
15001 vector unsigned short,
15003 vector bool short vec_sld (vector bool short,
15006 vector pixel vec_sld (vector pixel,
15009 vector signed char vec_sld (vector signed char,
15010 vector signed char,
15012 vector unsigned char vec_sld (vector unsigned char,
15013 vector unsigned char,
15015 vector bool char vec_sld (vector bool char,
15019 vector signed int vec_sll (vector signed int,
15020 vector unsigned int);
15021 vector signed int vec_sll (vector signed int,
15022 vector unsigned short);
15023 vector signed int vec_sll (vector signed int,
15024 vector unsigned char);
15025 vector unsigned int vec_sll (vector unsigned int,
15026 vector unsigned int);
15027 vector unsigned int vec_sll (vector unsigned int,
15028 vector unsigned short);
15029 vector unsigned int vec_sll (vector unsigned int,
15030 vector unsigned char);
15031 vector bool int vec_sll (vector bool int,
15032 vector unsigned int);
15033 vector bool int vec_sll (vector bool int,
15034 vector unsigned short);
15035 vector bool int vec_sll (vector bool int,
15036 vector unsigned char);
15037 vector signed short vec_sll (vector signed short,
15038 vector unsigned int);
15039 vector signed short vec_sll (vector signed short,
15040 vector unsigned short);
15041 vector signed short vec_sll (vector signed short,
15042 vector unsigned char);
15043 vector unsigned short vec_sll (vector unsigned short,
15044 vector unsigned int);
15045 vector unsigned short vec_sll (vector unsigned short,
15046 vector unsigned short);
15047 vector unsigned short vec_sll (vector unsigned short,
15048 vector unsigned char);
15049 vector bool short vec_sll (vector bool short, vector unsigned int);
15050 vector bool short vec_sll (vector bool short, vector unsigned short);
15051 vector bool short vec_sll (vector bool short, vector unsigned char);
15052 vector pixel vec_sll (vector pixel, vector unsigned int);
15053 vector pixel vec_sll (vector pixel, vector unsigned short);
15054 vector pixel vec_sll (vector pixel, vector unsigned char);
15055 vector signed char vec_sll (vector signed char, vector unsigned int);
15056 vector signed char vec_sll (vector signed char, vector unsigned short);
15057 vector signed char vec_sll (vector signed char, vector unsigned char);
15058 vector unsigned char vec_sll (vector unsigned char,
15059 vector unsigned int);
15060 vector unsigned char vec_sll (vector unsigned char,
15061 vector unsigned short);
15062 vector unsigned char vec_sll (vector unsigned char,
15063 vector unsigned char);
15064 vector bool char vec_sll (vector bool char, vector unsigned int);
15065 vector bool char vec_sll (vector bool char, vector unsigned short);
15066 vector bool char vec_sll (vector bool char, vector unsigned char);
15068 vector float vec_slo (vector float, vector signed char);
15069 vector float vec_slo (vector float, vector unsigned char);
15070 vector signed int vec_slo (vector signed int, vector signed char);
15071 vector signed int vec_slo (vector signed int, vector unsigned char);
15072 vector unsigned int vec_slo (vector unsigned int, vector signed char);
15073 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
15074 vector signed short vec_slo (vector signed short, vector signed char);
15075 vector signed short vec_slo (vector signed short, vector unsigned char);
15076 vector unsigned short vec_slo (vector unsigned short,
15077 vector signed char);
15078 vector unsigned short vec_slo (vector unsigned short,
15079 vector unsigned char);
15080 vector pixel vec_slo (vector pixel, vector signed char);
15081 vector pixel vec_slo (vector pixel, vector unsigned char);
15082 vector signed char vec_slo (vector signed char, vector signed char);
15083 vector signed char vec_slo (vector signed char, vector unsigned char);
15084 vector unsigned char vec_slo (vector unsigned char, vector signed char);
15085 vector unsigned char vec_slo (vector unsigned char,
15086 vector unsigned char);
15088 vector signed char vec_splat (vector signed char, const int);
15089 vector unsigned char vec_splat (vector unsigned char, const int);
15090 vector bool char vec_splat (vector bool char, const int);
15091 vector signed short vec_splat (vector signed short, const int);
15092 vector unsigned short vec_splat (vector unsigned short, const int);
15093 vector bool short vec_splat (vector bool short, const int);
15094 vector pixel vec_splat (vector pixel, const int);
15095 vector float vec_splat (vector float, const int);
15096 vector signed int vec_splat (vector signed int, const int);
15097 vector unsigned int vec_splat (vector unsigned int, const int);
15098 vector bool int vec_splat (vector bool int, const int);
15099 vector signed long vec_splat (vector signed long, const int);
15100 vector unsigned long vec_splat (vector unsigned long, const int);
15102 vector signed char vec_splats (signed char);
15103 vector unsigned char vec_splats (unsigned char);
15104 vector signed short vec_splats (signed short);
15105 vector unsigned short vec_splats (unsigned short);
15106 vector signed int vec_splats (signed int);
15107 vector unsigned int vec_splats (unsigned int);
15108 vector float vec_splats (float);
15110 vector float vec_vspltw (vector float, const int);
15111 vector signed int vec_vspltw (vector signed int, const int);
15112 vector unsigned int vec_vspltw (vector unsigned int, const int);
15113 vector bool int vec_vspltw (vector bool int, const int);
15115 vector bool short vec_vsplth (vector bool short, const int);
15116 vector signed short vec_vsplth (vector signed short, const int);
15117 vector unsigned short vec_vsplth (vector unsigned short, const int);
15118 vector pixel vec_vsplth (vector pixel, const int);
15120 vector signed char vec_vspltb (vector signed char, const int);
15121 vector unsigned char vec_vspltb (vector unsigned char, const int);
15122 vector bool char vec_vspltb (vector bool char, const int);
15124 vector signed char vec_splat_s8 (const int);
15126 vector signed short vec_splat_s16 (const int);
15128 vector signed int vec_splat_s32 (const int);
15130 vector unsigned char vec_splat_u8 (const int);
15132 vector unsigned short vec_splat_u16 (const int);
15134 vector unsigned int vec_splat_u32 (const int);
15136 vector signed char vec_sr (vector signed char, vector unsigned char);
15137 vector unsigned char vec_sr (vector unsigned char,
15138 vector unsigned char);
15139 vector signed short vec_sr (vector signed short,
15140 vector unsigned short);
15141 vector unsigned short vec_sr (vector unsigned short,
15142 vector unsigned short);
15143 vector signed int vec_sr (vector signed int, vector unsigned int);
15144 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
15146 vector signed int vec_vsrw (vector signed int, vector unsigned int);
15147 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
15149 vector signed short vec_vsrh (vector signed short,
15150 vector unsigned short);
15151 vector unsigned short vec_vsrh (vector unsigned short,
15152 vector unsigned short);
15154 vector signed char vec_vsrb (vector signed char, vector unsigned char);
15155 vector unsigned char vec_vsrb (vector unsigned char,
15156 vector unsigned char);
15158 vector signed char vec_sra (vector signed char, vector unsigned char);
15159 vector unsigned char vec_sra (vector unsigned char,
15160 vector unsigned char);
15161 vector signed short vec_sra (vector signed short,
15162 vector unsigned short);
15163 vector unsigned short vec_sra (vector unsigned short,
15164 vector unsigned short);
15165 vector signed int vec_sra (vector signed int, vector unsigned int);
15166 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
15168 vector signed int vec_vsraw (vector signed int, vector unsigned int);
15169 vector unsigned int vec_vsraw (vector unsigned int,
15170 vector unsigned int);
15172 vector signed short vec_vsrah (vector signed short,
15173 vector unsigned short);
15174 vector unsigned short vec_vsrah (vector unsigned short,
15175 vector unsigned short);
15177 vector signed char vec_vsrab (vector signed char, vector unsigned char);
15178 vector unsigned char vec_vsrab (vector unsigned char,
15179 vector unsigned char);
15181 vector signed int vec_srl (vector signed int, vector unsigned int);
15182 vector signed int vec_srl (vector signed int, vector unsigned short);
15183 vector signed int vec_srl (vector signed int, vector unsigned char);
15184 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
15185 vector unsigned int vec_srl (vector unsigned int,
15186 vector unsigned short);
15187 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
15188 vector bool int vec_srl (vector bool int, vector unsigned int);
15189 vector bool int vec_srl (vector bool int, vector unsigned short);
15190 vector bool int vec_srl (vector bool int, vector unsigned char);
15191 vector signed short vec_srl (vector signed short, vector unsigned int);
15192 vector signed short vec_srl (vector signed short,
15193 vector unsigned short);
15194 vector signed short vec_srl (vector signed short, vector unsigned char);
15195 vector unsigned short vec_srl (vector unsigned short,
15196 vector unsigned int);
15197 vector unsigned short vec_srl (vector unsigned short,
15198 vector unsigned short);
15199 vector unsigned short vec_srl (vector unsigned short,
15200 vector unsigned char);
15201 vector bool short vec_srl (vector bool short, vector unsigned int);
15202 vector bool short vec_srl (vector bool short, vector unsigned short);
15203 vector bool short vec_srl (vector bool short, vector unsigned char);
15204 vector pixel vec_srl (vector pixel, vector unsigned int);
15205 vector pixel vec_srl (vector pixel, vector unsigned short);
15206 vector pixel vec_srl (vector pixel, vector unsigned char);
15207 vector signed char vec_srl (vector signed char, vector unsigned int);
15208 vector signed char vec_srl (vector signed char, vector unsigned short);
15209 vector signed char vec_srl (vector signed char, vector unsigned char);
15210 vector unsigned char vec_srl (vector unsigned char,
15211 vector unsigned int);
15212 vector unsigned char vec_srl (vector unsigned char,
15213 vector unsigned short);
15214 vector unsigned char vec_srl (vector unsigned char,
15215 vector unsigned char);
15216 vector bool char vec_srl (vector bool char, vector unsigned int);
15217 vector bool char vec_srl (vector bool char, vector unsigned short);
15218 vector bool char vec_srl (vector bool char, vector unsigned char);
15220 vector float vec_sro (vector float, vector signed char);
15221 vector float vec_sro (vector float, vector unsigned char);
15222 vector signed int vec_sro (vector signed int, vector signed char);
15223 vector signed int vec_sro (vector signed int, vector unsigned char);
15224 vector unsigned int vec_sro (vector unsigned int, vector signed char);
15225 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
15226 vector signed short vec_sro (vector signed short, vector signed char);
15227 vector signed short vec_sro (vector signed short, vector unsigned char);
15228 vector unsigned short vec_sro (vector unsigned short,
15229 vector signed char);
15230 vector unsigned short vec_sro (vector unsigned short,
15231 vector unsigned char);
15232 vector pixel vec_sro (vector pixel, vector signed char);
15233 vector pixel vec_sro (vector pixel, vector unsigned char);
15234 vector signed char vec_sro (vector signed char, vector signed char);
15235 vector signed char vec_sro (vector signed char, vector unsigned char);
15236 vector unsigned char vec_sro (vector unsigned char, vector signed char);
15237 vector unsigned char vec_sro (vector unsigned char,
15238 vector unsigned char);
15240 void vec_st (vector float, int, vector float *);
15241 void vec_st (vector float, int, float *);
15242 void vec_st (vector signed int, int, vector signed int *);
15243 void vec_st (vector signed int, int, int *);
15244 void vec_st (vector unsigned int, int, vector unsigned int *);
15245 void vec_st (vector unsigned int, int, unsigned int *);
15246 void vec_st (vector bool int, int, vector bool int *);
15247 void vec_st (vector bool int, int, unsigned int *);
15248 void vec_st (vector bool int, int, int *);
15249 void vec_st (vector signed short, int, vector signed short *);
15250 void vec_st (vector signed short, int, short *);
15251 void vec_st (vector unsigned short, int, vector unsigned short *);
15252 void vec_st (vector unsigned short, int, unsigned short *);
15253 void vec_st (vector bool short, int, vector bool short *);
15254 void vec_st (vector bool short, int, unsigned short *);
15255 void vec_st (vector pixel, int, vector pixel *);
15256 void vec_st (vector pixel, int, unsigned short *);
15257 void vec_st (vector pixel, int, short *);
15258 void vec_st (vector bool short, int, short *);
15259 void vec_st (vector signed char, int, vector signed char *);
15260 void vec_st (vector signed char, int, signed char *);
15261 void vec_st (vector unsigned char, int, vector unsigned char *);
15262 void vec_st (vector unsigned char, int, unsigned char *);
15263 void vec_st (vector bool char, int, vector bool char *);
15264 void vec_st (vector bool char, int, unsigned char *);
15265 void vec_st (vector bool char, int, signed char *);
15267 void vec_ste (vector signed char, int, signed char *);
15268 void vec_ste (vector unsigned char, int, unsigned char *);
15269 void vec_ste (vector bool char, int, signed char *);
15270 void vec_ste (vector bool char, int, unsigned char *);
15271 void vec_ste (vector signed short, int, short *);
15272 void vec_ste (vector unsigned short, int, unsigned short *);
15273 void vec_ste (vector bool short, int, short *);
15274 void vec_ste (vector bool short, int, unsigned short *);
15275 void vec_ste (vector pixel, int, short *);
15276 void vec_ste (vector pixel, int, unsigned short *);
15277 void vec_ste (vector float, int, float *);
15278 void vec_ste (vector signed int, int, int *);
15279 void vec_ste (vector unsigned int, int, unsigned int *);
15280 void vec_ste (vector bool int, int, int *);
15281 void vec_ste (vector bool int, int, unsigned int *);
15283 void vec_stvewx (vector float, int, float *);
15284 void vec_stvewx (vector signed int, int, int *);
15285 void vec_stvewx (vector unsigned int, int, unsigned int *);
15286 void vec_stvewx (vector bool int, int, int *);
15287 void vec_stvewx (vector bool int, int, unsigned int *);
15289 void vec_stvehx (vector signed short, int, short *);
15290 void vec_stvehx (vector unsigned short, int, unsigned short *);
15291 void vec_stvehx (vector bool short, int, short *);
15292 void vec_stvehx (vector bool short, int, unsigned short *);
15293 void vec_stvehx (vector pixel, int, short *);
15294 void vec_stvehx (vector pixel, int, unsigned short *);
15296 void vec_stvebx (vector signed char, int, signed char *);
15297 void vec_stvebx (vector unsigned char, int, unsigned char *);
15298 void vec_stvebx (vector bool char, int, signed char *);
15299 void vec_stvebx (vector bool char, int, unsigned char *);
15301 void vec_stl (vector float, int, vector float *);
15302 void vec_stl (vector float, int, float *);
15303 void vec_stl (vector signed int, int, vector signed int *);
15304 void vec_stl (vector signed int, int, int *);
15305 void vec_stl (vector unsigned int, int, vector unsigned int *);
15306 void vec_stl (vector unsigned int, int, unsigned int *);
15307 void vec_stl (vector bool int, int, vector bool int *);
15308 void vec_stl (vector bool int, int, unsigned int *);
15309 void vec_stl (vector bool int, int, int *);
15310 void vec_stl (vector signed short, int, vector signed short *);
15311 void vec_stl (vector signed short, int, short *);
15312 void vec_stl (vector unsigned short, int, vector unsigned short *);
15313 void vec_stl (vector unsigned short, int, unsigned short *);
15314 void vec_stl (vector bool short, int, vector bool short *);
15315 void vec_stl (vector bool short, int, unsigned short *);
15316 void vec_stl (vector bool short, int, short *);
15317 void vec_stl (vector pixel, int, vector pixel *);
15318 void vec_stl (vector pixel, int, unsigned short *);
15319 void vec_stl (vector pixel, int, short *);
15320 void vec_stl (vector signed char, int, vector signed char *);
15321 void vec_stl (vector signed char, int, signed char *);
15322 void vec_stl (vector unsigned char, int, vector unsigned char *);
15323 void vec_stl (vector unsigned char, int, unsigned char *);
15324 void vec_stl (vector bool char, int, vector bool char *);
15325 void vec_stl (vector bool char, int, unsigned char *);
15326 void vec_stl (vector bool char, int, signed char *);
15328 vector signed char vec_sub (vector bool char, vector signed char);
15329 vector signed char vec_sub (vector signed char, vector bool char);
15330 vector signed char vec_sub (vector signed char, vector signed char);
15331 vector unsigned char vec_sub (vector bool char, vector unsigned char);
15332 vector unsigned char vec_sub (vector unsigned char, vector bool char);
15333 vector unsigned char vec_sub (vector unsigned char,
15334 vector unsigned char);
15335 vector signed short vec_sub (vector bool short, vector signed short);
15336 vector signed short vec_sub (vector signed short, vector bool short);
15337 vector signed short vec_sub (vector signed short, vector signed short);
15338 vector unsigned short vec_sub (vector bool short,
15339 vector unsigned short);
15340 vector unsigned short vec_sub (vector unsigned short,
15341 vector bool short);
15342 vector unsigned short vec_sub (vector unsigned short,
15343 vector unsigned short);
15344 vector signed int vec_sub (vector bool int, vector signed int);
15345 vector signed int vec_sub (vector signed int, vector bool int);
15346 vector signed int vec_sub (vector signed int, vector signed int);
15347 vector unsigned int vec_sub (vector bool int, vector unsigned int);
15348 vector unsigned int vec_sub (vector unsigned int, vector bool int);
15349 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
15350 vector float vec_sub (vector float, vector float);
15352 vector float vec_vsubfp (vector float, vector float);
15354 vector signed int vec_vsubuwm (vector bool int, vector signed int);
15355 vector signed int vec_vsubuwm (vector signed int, vector bool int);
15356 vector signed int vec_vsubuwm (vector signed int, vector signed int);
15357 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
15358 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
15359 vector unsigned int vec_vsubuwm (vector unsigned int,
15360 vector unsigned int);
15362 vector signed short vec_vsubuhm (vector bool short,
15363 vector signed short);
15364 vector signed short vec_vsubuhm (vector signed short,
15365 vector bool short);
15366 vector signed short vec_vsubuhm (vector signed short,
15367 vector signed short);
15368 vector unsigned short vec_vsubuhm (vector bool short,
15369 vector unsigned short);
15370 vector unsigned short vec_vsubuhm (vector unsigned short,
15371 vector bool short);
15372 vector unsigned short vec_vsubuhm (vector unsigned short,
15373 vector unsigned short);
15375 vector signed char vec_vsububm (vector bool char, vector signed char);
15376 vector signed char vec_vsububm (vector signed char, vector bool char);
15377 vector signed char vec_vsububm (vector signed char, vector signed char);
15378 vector unsigned char vec_vsububm (vector bool char,
15379 vector unsigned char);
15380 vector unsigned char vec_vsububm (vector unsigned char,
15382 vector unsigned char vec_vsububm (vector unsigned char,
15383 vector unsigned char);
15385 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
15387 vector unsigned char vec_subs (vector bool char, vector unsigned char);
15388 vector unsigned char vec_subs (vector unsigned char, vector bool char);
15389 vector unsigned char vec_subs (vector unsigned char,
15390 vector unsigned char);
15391 vector signed char vec_subs (vector bool char, vector signed char);
15392 vector signed char vec_subs (vector signed char, vector bool char);
15393 vector signed char vec_subs (vector signed char, vector signed char);
15394 vector unsigned short vec_subs (vector bool short,
15395 vector unsigned short);
15396 vector unsigned short vec_subs (vector unsigned short,
15397 vector bool short);
15398 vector unsigned short vec_subs (vector unsigned short,
15399 vector unsigned short);
15400 vector signed short vec_subs (vector bool short, vector signed short);
15401 vector signed short vec_subs (vector signed short, vector bool short);
15402 vector signed short vec_subs (vector signed short, vector signed short);
15403 vector unsigned int vec_subs (vector bool int, vector unsigned int);
15404 vector unsigned int vec_subs (vector unsigned int, vector bool int);
15405 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
15406 vector signed int vec_subs (vector bool int, vector signed int);
15407 vector signed int vec_subs (vector signed int, vector bool int);
15408 vector signed int vec_subs (vector signed int, vector signed int);
15410 vector signed int vec_vsubsws (vector bool int, vector signed int);
15411 vector signed int vec_vsubsws (vector signed int, vector bool int);
15412 vector signed int vec_vsubsws (vector signed int, vector signed int);
15414 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
15415 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
15416 vector unsigned int vec_vsubuws (vector unsigned int,
15417 vector unsigned int);
15419 vector signed short vec_vsubshs (vector bool short,
15420 vector signed short);
15421 vector signed short vec_vsubshs (vector signed short,
15422 vector bool short);
15423 vector signed short vec_vsubshs (vector signed short,
15424 vector signed short);
15426 vector unsigned short vec_vsubuhs (vector bool short,
15427 vector unsigned short);
15428 vector unsigned short vec_vsubuhs (vector unsigned short,
15429 vector bool short);
15430 vector unsigned short vec_vsubuhs (vector unsigned short,
15431 vector unsigned short);
15433 vector signed char vec_vsubsbs (vector bool char, vector signed char);
15434 vector signed char vec_vsubsbs (vector signed char, vector bool char);
15435 vector signed char vec_vsubsbs (vector signed char, vector signed char);
15437 vector unsigned char vec_vsububs (vector bool char,
15438 vector unsigned char);
15439 vector unsigned char vec_vsububs (vector unsigned char,
15441 vector unsigned char vec_vsububs (vector unsigned char,
15442 vector unsigned char);
15444 vector unsigned int vec_sum4s (vector unsigned char,
15445 vector unsigned int);
15446 vector signed int vec_sum4s (vector signed char, vector signed int);
15447 vector signed int vec_sum4s (vector signed short, vector signed int);
15449 vector signed int vec_vsum4shs (vector signed short, vector signed int);
15451 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
15453 vector unsigned int vec_vsum4ubs (vector unsigned char,
15454 vector unsigned int);
15456 vector signed int vec_sum2s (vector signed int, vector signed int);
15458 vector signed int vec_sums (vector signed int, vector signed int);
15460 vector float vec_trunc (vector float);
15462 vector signed short vec_unpackh (vector signed char);
15463 vector bool short vec_unpackh (vector bool char);
15464 vector signed int vec_unpackh (vector signed short);
15465 vector bool int vec_unpackh (vector bool short);
15466 vector unsigned int vec_unpackh (vector pixel);
15468 vector bool int vec_vupkhsh (vector bool short);
15469 vector signed int vec_vupkhsh (vector signed short);
15471 vector unsigned int vec_vupkhpx (vector pixel);
15473 vector bool short vec_vupkhsb (vector bool char);
15474 vector signed short vec_vupkhsb (vector signed char);
15476 vector signed short vec_unpackl (vector signed char);
15477 vector bool short vec_unpackl (vector bool char);
15478 vector unsigned int vec_unpackl (vector pixel);
15479 vector signed int vec_unpackl (vector signed short);
15480 vector bool int vec_unpackl (vector bool short);
15482 vector unsigned int vec_vupklpx (vector pixel);
15484 vector bool int vec_vupklsh (vector bool short);
15485 vector signed int vec_vupklsh (vector signed short);
15487 vector bool short vec_vupklsb (vector bool char);
15488 vector signed short vec_vupklsb (vector signed char);
15490 vector float vec_xor (vector float, vector float);
15491 vector float vec_xor (vector float, vector bool int);
15492 vector float vec_xor (vector bool int, vector float);
15493 vector bool int vec_xor (vector bool int, vector bool int);
15494 vector signed int vec_xor (vector bool int, vector signed int);
15495 vector signed int vec_xor (vector signed int, vector bool int);
15496 vector signed int vec_xor (vector signed int, vector signed int);
15497 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15498 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15499 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15500 vector bool short vec_xor (vector bool short, vector bool short);
15501 vector signed short vec_xor (vector bool short, vector signed short);
15502 vector signed short vec_xor (vector signed short, vector bool short);
15503 vector signed short vec_xor (vector signed short, vector signed short);
15504 vector unsigned short vec_xor (vector bool short,
15505 vector unsigned short);
15506 vector unsigned short vec_xor (vector unsigned short,
15507 vector bool short);
15508 vector unsigned short vec_xor (vector unsigned short,
15509 vector unsigned short);
15510 vector signed char vec_xor (vector bool char, vector signed char);
15511 vector bool char vec_xor (vector bool char, vector bool char);
15512 vector signed char vec_xor (vector signed char, vector bool char);
15513 vector signed char vec_xor (vector signed char, vector signed char);
15514 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15515 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15516 vector unsigned char vec_xor (vector unsigned char,
15517 vector unsigned char);
15519 int vec_all_eq (vector signed char, vector bool char);
15520 int vec_all_eq (vector signed char, vector signed char);
15521 int vec_all_eq (vector unsigned char, vector bool char);
15522 int vec_all_eq (vector unsigned char, vector unsigned char);
15523 int vec_all_eq (vector bool char, vector bool char);
15524 int vec_all_eq (vector bool char, vector unsigned char);
15525 int vec_all_eq (vector bool char, vector signed char);
15526 int vec_all_eq (vector signed short, vector bool short);
15527 int vec_all_eq (vector signed short, vector signed short);
15528 int vec_all_eq (vector unsigned short, vector bool short);
15529 int vec_all_eq (vector unsigned short, vector unsigned short);
15530 int vec_all_eq (vector bool short, vector bool short);
15531 int vec_all_eq (vector bool short, vector unsigned short);
15532 int vec_all_eq (vector bool short, vector signed short);
15533 int vec_all_eq (vector pixel, vector pixel);
15534 int vec_all_eq (vector signed int, vector bool int);
15535 int vec_all_eq (vector signed int, vector signed int);
15536 int vec_all_eq (vector unsigned int, vector bool int);
15537 int vec_all_eq (vector unsigned int, vector unsigned int);
15538 int vec_all_eq (vector bool int, vector bool int);
15539 int vec_all_eq (vector bool int, vector unsigned int);
15540 int vec_all_eq (vector bool int, vector signed int);
15541 int vec_all_eq (vector float, vector float);
15543 int vec_all_ge (vector bool char, vector unsigned char);
15544 int vec_all_ge (vector unsigned char, vector bool char);
15545 int vec_all_ge (vector unsigned char, vector unsigned char);
15546 int vec_all_ge (vector bool char, vector signed char);
15547 int vec_all_ge (vector signed char, vector bool char);
15548 int vec_all_ge (vector signed char, vector signed char);
15549 int vec_all_ge (vector bool short, vector unsigned short);
15550 int vec_all_ge (vector unsigned short, vector bool short);
15551 int vec_all_ge (vector unsigned short, vector unsigned short);
15552 int vec_all_ge (vector signed short, vector signed short);
15553 int vec_all_ge (vector bool short, vector signed short);
15554 int vec_all_ge (vector signed short, vector bool short);
15555 int vec_all_ge (vector bool int, vector unsigned int);
15556 int vec_all_ge (vector unsigned int, vector bool int);
15557 int vec_all_ge (vector unsigned int, vector unsigned int);
15558 int vec_all_ge (vector bool int, vector signed int);
15559 int vec_all_ge (vector signed int, vector bool int);
15560 int vec_all_ge (vector signed int, vector signed int);
15561 int vec_all_ge (vector float, vector float);
15563 int vec_all_gt (vector bool char, vector unsigned char);
15564 int vec_all_gt (vector unsigned char, vector bool char);
15565 int vec_all_gt (vector unsigned char, vector unsigned char);
15566 int vec_all_gt (vector bool char, vector signed char);
15567 int vec_all_gt (vector signed char, vector bool char);
15568 int vec_all_gt (vector signed char, vector signed char);
15569 int vec_all_gt (vector bool short, vector unsigned short);
15570 int vec_all_gt (vector unsigned short, vector bool short);
15571 int vec_all_gt (vector unsigned short, vector unsigned short);
15572 int vec_all_gt (vector bool short, vector signed short);
15573 int vec_all_gt (vector signed short, vector bool short);
15574 int vec_all_gt (vector signed short, vector signed short);
15575 int vec_all_gt (vector bool int, vector unsigned int);
15576 int vec_all_gt (vector unsigned int, vector bool int);
15577 int vec_all_gt (vector unsigned int, vector unsigned int);
15578 int vec_all_gt (vector bool int, vector signed int);
15579 int vec_all_gt (vector signed int, vector bool int);
15580 int vec_all_gt (vector signed int, vector signed int);
15581 int vec_all_gt (vector float, vector float);
15583 int vec_all_in (vector float, vector float);
15585 int vec_all_le (vector bool char, vector unsigned char);
15586 int vec_all_le (vector unsigned char, vector bool char);
15587 int vec_all_le (vector unsigned char, vector unsigned char);
15588 int vec_all_le (vector bool char, vector signed char);
15589 int vec_all_le (vector signed char, vector bool char);
15590 int vec_all_le (vector signed char, vector signed char);
15591 int vec_all_le (vector bool short, vector unsigned short);
15592 int vec_all_le (vector unsigned short, vector bool short);
15593 int vec_all_le (vector unsigned short, vector unsigned short);
15594 int vec_all_le (vector bool short, vector signed short);
15595 int vec_all_le (vector signed short, vector bool short);
15596 int vec_all_le (vector signed short, vector signed short);
15597 int vec_all_le (vector bool int, vector unsigned int);
15598 int vec_all_le (vector unsigned int, vector bool int);
15599 int vec_all_le (vector unsigned int, vector unsigned int);
15600 int vec_all_le (vector bool int, vector signed int);
15601 int vec_all_le (vector signed int, vector bool int);
15602 int vec_all_le (vector signed int, vector signed int);
15603 int vec_all_le (vector float, vector float);
15605 int vec_all_lt (vector bool char, vector unsigned char);
15606 int vec_all_lt (vector unsigned char, vector bool char);
15607 int vec_all_lt (vector unsigned char, vector unsigned char);
15608 int vec_all_lt (vector bool char, vector signed char);
15609 int vec_all_lt (vector signed char, vector bool char);
15610 int vec_all_lt (vector signed char, vector signed char);
15611 int vec_all_lt (vector bool short, vector unsigned short);
15612 int vec_all_lt (vector unsigned short, vector bool short);
15613 int vec_all_lt (vector unsigned short, vector unsigned short);
15614 int vec_all_lt (vector bool short, vector signed short);
15615 int vec_all_lt (vector signed short, vector bool short);
15616 int vec_all_lt (vector signed short, vector signed short);
15617 int vec_all_lt (vector bool int, vector unsigned int);
15618 int vec_all_lt (vector unsigned int, vector bool int);
15619 int vec_all_lt (vector unsigned int, vector unsigned int);
15620 int vec_all_lt (vector bool int, vector signed int);
15621 int vec_all_lt (vector signed int, vector bool int);
15622 int vec_all_lt (vector signed int, vector signed int);
15623 int vec_all_lt (vector float, vector float);
15625 int vec_all_nan (vector float);
15627 int vec_all_ne (vector signed char, vector bool char);
15628 int vec_all_ne (vector signed char, vector signed char);
15629 int vec_all_ne (vector unsigned char, vector bool char);
15630 int vec_all_ne (vector unsigned char, vector unsigned char);
15631 int vec_all_ne (vector bool char, vector bool char);
15632 int vec_all_ne (vector bool char, vector unsigned char);
15633 int vec_all_ne (vector bool char, vector signed char);
15634 int vec_all_ne (vector signed short, vector bool short);
15635 int vec_all_ne (vector signed short, vector signed short);
15636 int vec_all_ne (vector unsigned short, vector bool short);
15637 int vec_all_ne (vector unsigned short, vector unsigned short);
15638 int vec_all_ne (vector bool short, vector bool short);
15639 int vec_all_ne (vector bool short, vector unsigned short);
15640 int vec_all_ne (vector bool short, vector signed short);
15641 int vec_all_ne (vector pixel, vector pixel);
15642 int vec_all_ne (vector signed int, vector bool int);
15643 int vec_all_ne (vector signed int, vector signed int);
15644 int vec_all_ne (vector unsigned int, vector bool int);
15645 int vec_all_ne (vector unsigned int, vector unsigned int);
15646 int vec_all_ne (vector bool int, vector bool int);
15647 int vec_all_ne (vector bool int, vector unsigned int);
15648 int vec_all_ne (vector bool int, vector signed int);
15649 int vec_all_ne (vector float, vector float);
15651 int vec_all_nge (vector float, vector float);
15653 int vec_all_ngt (vector float, vector float);
15655 int vec_all_nle (vector float, vector float);
15657 int vec_all_nlt (vector float, vector float);
15659 int vec_all_numeric (vector float);
15661 int vec_any_eq (vector signed char, vector bool char);
15662 int vec_any_eq (vector signed char, vector signed char);
15663 int vec_any_eq (vector unsigned char, vector bool char);
15664 int vec_any_eq (vector unsigned char, vector unsigned char);
15665 int vec_any_eq (vector bool char, vector bool char);
15666 int vec_any_eq (vector bool char, vector unsigned char);
15667 int vec_any_eq (vector bool char, vector signed char);
15668 int vec_any_eq (vector signed short, vector bool short);
15669 int vec_any_eq (vector signed short, vector signed short);
15670 int vec_any_eq (vector unsigned short, vector bool short);
15671 int vec_any_eq (vector unsigned short, vector unsigned short);
15672 int vec_any_eq (vector bool short, vector bool short);
15673 int vec_any_eq (vector bool short, vector unsigned short);
15674 int vec_any_eq (vector bool short, vector signed short);
15675 int vec_any_eq (vector pixel, vector pixel);
15676 int vec_any_eq (vector signed int, vector bool int);
15677 int vec_any_eq (vector signed int, vector signed int);
15678 int vec_any_eq (vector unsigned int, vector bool int);
15679 int vec_any_eq (vector unsigned int, vector unsigned int);
15680 int vec_any_eq (vector bool int, vector bool int);
15681 int vec_any_eq (vector bool int, vector unsigned int);
15682 int vec_any_eq (vector bool int, vector signed int);
15683 int vec_any_eq (vector float, vector float);
15685 int vec_any_ge (vector signed char, vector bool char);
15686 int vec_any_ge (vector unsigned char, vector bool char);
15687 int vec_any_ge (vector unsigned char, vector unsigned char);
15688 int vec_any_ge (vector signed char, vector signed char);
15689 int vec_any_ge (vector bool char, vector unsigned char);
15690 int vec_any_ge (vector bool char, vector signed char);
15691 int vec_any_ge (vector unsigned short, vector bool short);
15692 int vec_any_ge (vector unsigned short, vector unsigned short);
15693 int vec_any_ge (vector signed short, vector signed short);
15694 int vec_any_ge (vector signed short, vector bool short);
15695 int vec_any_ge (vector bool short, vector unsigned short);
15696 int vec_any_ge (vector bool short, vector signed short);
15697 int vec_any_ge (vector signed int, vector bool int);
15698 int vec_any_ge (vector unsigned int, vector bool int);
15699 int vec_any_ge (vector unsigned int, vector unsigned int);
15700 int vec_any_ge (vector signed int, vector signed int);
15701 int vec_any_ge (vector bool int, vector unsigned int);
15702 int vec_any_ge (vector bool int, vector signed int);
15703 int vec_any_ge (vector float, vector float);
15705 int vec_any_gt (vector bool char, vector unsigned char);
15706 int vec_any_gt (vector unsigned char, vector bool char);
15707 int vec_any_gt (vector unsigned char, vector unsigned char);
15708 int vec_any_gt (vector bool char, vector signed char);
15709 int vec_any_gt (vector signed char, vector bool char);
15710 int vec_any_gt (vector signed char, vector signed char);
15711 int vec_any_gt (vector bool short, vector unsigned short);
15712 int vec_any_gt (vector unsigned short, vector bool short);
15713 int vec_any_gt (vector unsigned short, vector unsigned short);
15714 int vec_any_gt (vector bool short, vector signed short);
15715 int vec_any_gt (vector signed short, vector bool short);
15716 int vec_any_gt (vector signed short, vector signed short);
15717 int vec_any_gt (vector bool int, vector unsigned int);
15718 int vec_any_gt (vector unsigned int, vector bool int);
15719 int vec_any_gt (vector unsigned int, vector unsigned int);
15720 int vec_any_gt (vector bool int, vector signed int);
15721 int vec_any_gt (vector signed int, vector bool int);
15722 int vec_any_gt (vector signed int, vector signed int);
15723 int vec_any_gt (vector float, vector float);
15725 int vec_any_le (vector bool char, vector unsigned char);
15726 int vec_any_le (vector unsigned char, vector bool char);
15727 int vec_any_le (vector unsigned char, vector unsigned char);
15728 int vec_any_le (vector bool char, vector signed char);
15729 int vec_any_le (vector signed char, vector bool char);
15730 int vec_any_le (vector signed char, vector signed char);
15731 int vec_any_le (vector bool short, vector unsigned short);
15732 int vec_any_le (vector unsigned short, vector bool short);
15733 int vec_any_le (vector unsigned short, vector unsigned short);
15734 int vec_any_le (vector bool short, vector signed short);
15735 int vec_any_le (vector signed short, vector bool short);
15736 int vec_any_le (vector signed short, vector signed short);
15737 int vec_any_le (vector bool int, vector unsigned int);
15738 int vec_any_le (vector unsigned int, vector bool int);
15739 int vec_any_le (vector unsigned int, vector unsigned int);
15740 int vec_any_le (vector bool int, vector signed int);
15741 int vec_any_le (vector signed int, vector bool int);
15742 int vec_any_le (vector signed int, vector signed int);
15743 int vec_any_le (vector float, vector float);
15745 int vec_any_lt (vector bool char, vector unsigned char);
15746 int vec_any_lt (vector unsigned char, vector bool char);
15747 int vec_any_lt (vector unsigned char, vector unsigned char);
15748 int vec_any_lt (vector bool char, vector signed char);
15749 int vec_any_lt (vector signed char, vector bool char);
15750 int vec_any_lt (vector signed char, vector signed char);
15751 int vec_any_lt (vector bool short, vector unsigned short);
15752 int vec_any_lt (vector unsigned short, vector bool short);
15753 int vec_any_lt (vector unsigned short, vector unsigned short);
15754 int vec_any_lt (vector bool short, vector signed short);
15755 int vec_any_lt (vector signed short, vector bool short);
15756 int vec_any_lt (vector signed short, vector signed short);
15757 int vec_any_lt (vector bool int, vector unsigned int);
15758 int vec_any_lt (vector unsigned int, vector bool int);
15759 int vec_any_lt (vector unsigned int, vector unsigned int);
15760 int vec_any_lt (vector bool int, vector signed int);
15761 int vec_any_lt (vector signed int, vector bool int);
15762 int vec_any_lt (vector signed int, vector signed int);
15763 int vec_any_lt (vector float, vector float);
15765 int vec_any_nan (vector float);
15767 int vec_any_ne (vector signed char, vector bool char);
15768 int vec_any_ne (vector signed char, vector signed char);
15769 int vec_any_ne (vector unsigned char, vector bool char);
15770 int vec_any_ne (vector unsigned char, vector unsigned char);
15771 int vec_any_ne (vector bool char, vector bool char);
15772 int vec_any_ne (vector bool char, vector unsigned char);
15773 int vec_any_ne (vector bool char, vector signed char);
15774 int vec_any_ne (vector signed short, vector bool short);
15775 int vec_any_ne (vector signed short, vector signed short);
15776 int vec_any_ne (vector unsigned short, vector bool short);
15777 int vec_any_ne (vector unsigned short, vector unsigned short);
15778 int vec_any_ne (vector bool short, vector bool short);
15779 int vec_any_ne (vector bool short, vector unsigned short);
15780 int vec_any_ne (vector bool short, vector signed short);
15781 int vec_any_ne (vector pixel, vector pixel);
15782 int vec_any_ne (vector signed int, vector bool int);
15783 int vec_any_ne (vector signed int, vector signed int);
15784 int vec_any_ne (vector unsigned int, vector bool int);
15785 int vec_any_ne (vector unsigned int, vector unsigned int);
15786 int vec_any_ne (vector bool int, vector bool int);
15787 int vec_any_ne (vector bool int, vector unsigned int);
15788 int vec_any_ne (vector bool int, vector signed int);
15789 int vec_any_ne (vector float, vector float);
15791 int vec_any_nge (vector float, vector float);
15793 int vec_any_ngt (vector float, vector float);
15795 int vec_any_nle (vector float, vector float);
15797 int vec_any_nlt (vector float, vector float);
15799 int vec_any_numeric (vector float);
15801 int vec_any_out (vector float, vector float);
15804 If the vector/scalar (VSX) instruction set is available, the following
15805 additional functions are available:
15808 vector double vec_abs (vector double);
15809 vector double vec_add (vector double, vector double);
15810 vector double vec_and (vector double, vector double);
15811 vector double vec_and (vector double, vector bool long);
15812 vector double vec_and (vector bool long, vector double);
15813 vector long vec_and (vector long, vector long);
15814 vector long vec_and (vector long, vector bool long);
15815 vector long vec_and (vector bool long, vector long);
15816 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15817 vector unsigned long vec_and (vector unsigned long, vector bool long);
15818 vector unsigned long vec_and (vector bool long, vector unsigned long);
15819 vector double vec_andc (vector double, vector double);
15820 vector double vec_andc (vector double, vector bool long);
15821 vector double vec_andc (vector bool long, vector double);
15822 vector long vec_andc (vector long, vector long);
15823 vector long vec_andc (vector long, vector bool long);
15824 vector long vec_andc (vector bool long, vector long);
15825 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15826 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15827 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15828 vector double vec_ceil (vector double);
15829 vector bool long vec_cmpeq (vector double, vector double);
15830 vector bool long vec_cmpge (vector double, vector double);
15831 vector bool long vec_cmpgt (vector double, vector double);
15832 vector bool long vec_cmple (vector double, vector double);
15833 vector bool long vec_cmplt (vector double, vector double);
15834 vector double vec_cpsgn (vector double, vector double);
15835 vector float vec_div (vector float, vector float);
15836 vector double vec_div (vector double, vector double);
15837 vector long vec_div (vector long, vector long);
15838 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15839 vector double vec_floor (vector double);
15840 vector double vec_ld (int, const vector double *);
15841 vector double vec_ld (int, const double *);
15842 vector double vec_ldl (int, const vector double *);
15843 vector double vec_ldl (int, const double *);
15844 vector unsigned char vec_lvsl (int, const volatile double *);
15845 vector unsigned char vec_lvsr (int, const volatile double *);
15846 vector double vec_madd (vector double, vector double, vector double);
15847 vector double vec_max (vector double, vector double);
15848 vector signed long vec_mergeh (vector signed long, vector signed long);
15849 vector signed long vec_mergeh (vector signed long, vector bool long);
15850 vector signed long vec_mergeh (vector bool long, vector signed long);
15851 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15852 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15853 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15854 vector signed long vec_mergel (vector signed long, vector signed long);
15855 vector signed long vec_mergel (vector signed long, vector bool long);
15856 vector signed long vec_mergel (vector bool long, vector signed long);
15857 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15858 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15859 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15860 vector double vec_min (vector double, vector double);
15861 vector float vec_msub (vector float, vector float, vector float);
15862 vector double vec_msub (vector double, vector double, vector double);
15863 vector float vec_mul (vector float, vector float);
15864 vector double vec_mul (vector double, vector double);
15865 vector long vec_mul (vector long, vector long);
15866 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15867 vector float vec_nearbyint (vector float);
15868 vector double vec_nearbyint (vector double);
15869 vector float vec_nmadd (vector float, vector float, vector float);
15870 vector double vec_nmadd (vector double, vector double, vector double);
15871 vector double vec_nmsub (vector double, vector double, vector double);
15872 vector double vec_nor (vector double, vector double);
15873 vector long vec_nor (vector long, vector long);
15874 vector long vec_nor (vector long, vector bool long);
15875 vector long vec_nor (vector bool long, vector long);
15876 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15877 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15878 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15879 vector double vec_or (vector double, vector double);
15880 vector double vec_or (vector double, vector bool long);
15881 vector double vec_or (vector bool long, vector double);
15882 vector long vec_or (vector long, vector long);
15883 vector long vec_or (vector long, vector bool long);
15884 vector long vec_or (vector bool long, vector long);
15885 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15886 vector unsigned long vec_or (vector unsigned long, vector bool long);
15887 vector unsigned long vec_or (vector bool long, vector unsigned long);
15888 vector double vec_perm (vector double, vector double, vector unsigned char);
15889 vector long vec_perm (vector long, vector long, vector unsigned char);
15890 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15891 vector unsigned char);
15892 vector double vec_rint (vector double);
15893 vector double vec_recip (vector double, vector double);
15894 vector double vec_rsqrt (vector double);
15895 vector double vec_rsqrte (vector double);
15896 vector double vec_sel (vector double, vector double, vector bool long);
15897 vector double vec_sel (vector double, vector double, vector unsigned long);
15898 vector long vec_sel (vector long, vector long, vector long);
15899 vector long vec_sel (vector long, vector long, vector unsigned long);
15900 vector long vec_sel (vector long, vector long, vector bool long);
15901 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15903 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15904 vector unsigned long);
15905 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15907 vector double vec_splats (double);
15908 vector signed long vec_splats (signed long);
15909 vector unsigned long vec_splats (unsigned long);
15910 vector float vec_sqrt (vector float);
15911 vector double vec_sqrt (vector double);
15912 void vec_st (vector double, int, vector double *);
15913 void vec_st (vector double, int, double *);
15914 vector double vec_sub (vector double, vector double);
15915 vector double vec_trunc (vector double);
15916 vector double vec_xor (vector double, vector double);
15917 vector double vec_xor (vector double, vector bool long);
15918 vector double vec_xor (vector bool long, vector double);
15919 vector long vec_xor (vector long, vector long);
15920 vector long vec_xor (vector long, vector bool long);
15921 vector long vec_xor (vector bool long, vector long);
15922 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15923 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15924 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15925 int vec_all_eq (vector double, vector double);
15926 int vec_all_ge (vector double, vector double);
15927 int vec_all_gt (vector double, vector double);
15928 int vec_all_le (vector double, vector double);
15929 int vec_all_lt (vector double, vector double);
15930 int vec_all_nan (vector double);
15931 int vec_all_ne (vector double, vector double);
15932 int vec_all_nge (vector double, vector double);
15933 int vec_all_ngt (vector double, vector double);
15934 int vec_all_nle (vector double, vector double);
15935 int vec_all_nlt (vector double, vector double);
15936 int vec_all_numeric (vector double);
15937 int vec_any_eq (vector double, vector double);
15938 int vec_any_ge (vector double, vector double);
15939 int vec_any_gt (vector double, vector double);
15940 int vec_any_le (vector double, vector double);
15941 int vec_any_lt (vector double, vector double);
15942 int vec_any_nan (vector double);
15943 int vec_any_ne (vector double, vector double);
15944 int vec_any_nge (vector double, vector double);
15945 int vec_any_ngt (vector double, vector double);
15946 int vec_any_nle (vector double, vector double);
15947 int vec_any_nlt (vector double, vector double);
15948 int vec_any_numeric (vector double);
15950 vector double vec_vsx_ld (int, const vector double *);
15951 vector double vec_vsx_ld (int, const double *);
15952 vector float vec_vsx_ld (int, const vector float *);
15953 vector float vec_vsx_ld (int, const float *);
15954 vector bool int vec_vsx_ld (int, const vector bool int *);
15955 vector signed int vec_vsx_ld (int, const vector signed int *);
15956 vector signed int vec_vsx_ld (int, const int *);
15957 vector signed int vec_vsx_ld (int, const long *);
15958 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15959 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15960 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15961 vector bool short vec_vsx_ld (int, const vector bool short *);
15962 vector pixel vec_vsx_ld (int, const vector pixel *);
15963 vector signed short vec_vsx_ld (int, const vector signed short *);
15964 vector signed short vec_vsx_ld (int, const short *);
15965 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15966 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15967 vector bool char vec_vsx_ld (int, const vector bool char *);
15968 vector signed char vec_vsx_ld (int, const vector signed char *);
15969 vector signed char vec_vsx_ld (int, const signed char *);
15970 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15971 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15973 void vec_vsx_st (vector double, int, vector double *);
15974 void vec_vsx_st (vector double, int, double *);
15975 void vec_vsx_st (vector float, int, vector float *);
15976 void vec_vsx_st (vector float, int, float *);
15977 void vec_vsx_st (vector signed int, int, vector signed int *);
15978 void vec_vsx_st (vector signed int, int, int *);
15979 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15980 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15981 void vec_vsx_st (vector bool int, int, vector bool int *);
15982 void vec_vsx_st (vector bool int, int, unsigned int *);
15983 void vec_vsx_st (vector bool int, int, int *);
15984 void vec_vsx_st (vector signed short, int, vector signed short *);
15985 void vec_vsx_st (vector signed short, int, short *);
15986 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15987 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15988 void vec_vsx_st (vector bool short, int, vector bool short *);
15989 void vec_vsx_st (vector bool short, int, unsigned short *);
15990 void vec_vsx_st (vector pixel, int, vector pixel *);
15991 void vec_vsx_st (vector pixel, int, unsigned short *);
15992 void vec_vsx_st (vector pixel, int, short *);
15993 void vec_vsx_st (vector bool short, int, short *);
15994 void vec_vsx_st (vector signed char, int, vector signed char *);
15995 void vec_vsx_st (vector signed char, int, signed char *);
15996 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15997 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15998 void vec_vsx_st (vector bool char, int, vector bool char *);
15999 void vec_vsx_st (vector bool char, int, unsigned char *);
16000 void vec_vsx_st (vector bool char, int, signed char *);
16002 vector double vec_xxpermdi (vector double, vector double, int);
16003 vector float vec_xxpermdi (vector float, vector float, int);
16004 vector long long vec_xxpermdi (vector long long, vector long long, int);
16005 vector unsigned long long vec_xxpermdi (vector unsigned long long,
16006 vector unsigned long long, int);
16007 vector int vec_xxpermdi (vector int, vector int, int);
16008 vector unsigned int vec_xxpermdi (vector unsigned int,
16009 vector unsigned int, int);
16010 vector short vec_xxpermdi (vector short, vector short, int);
16011 vector unsigned short vec_xxpermdi (vector unsigned short,
16012 vector unsigned short, int);
16013 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
16014 vector unsigned char vec_xxpermdi (vector unsigned char,
16015 vector unsigned char, int);
16017 vector double vec_xxsldi (vector double, vector double, int);
16018 vector float vec_xxsldi (vector float, vector float, int);
16019 vector long long vec_xxsldi (vector long long, vector long long, int);
16020 vector unsigned long long vec_xxsldi (vector unsigned long long,
16021 vector unsigned long long, int);
16022 vector int vec_xxsldi (vector int, vector int, int);
16023 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
16024 vector short vec_xxsldi (vector short, vector short, int);
16025 vector unsigned short vec_xxsldi (vector unsigned short,
16026 vector unsigned short, int);
16027 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
16028 vector unsigned char vec_xxsldi (vector unsigned char,
16029 vector unsigned char, int);
16032 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
16033 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
16034 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
16035 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
16036 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
16038 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16039 instruction set is available, the following additional functions are
16040 available for both 32-bit and 64-bit targets. For 64-bit targets, you
16041 can use @var{vector long} instead of @var{vector long long},
16042 @var{vector bool long} instead of @var{vector bool long long}, and
16043 @var{vector unsigned long} instead of @var{vector unsigned long long}.
16046 vector long long vec_abs (vector long long);
16048 vector long long vec_add (vector long long, vector long long);
16049 vector unsigned long long vec_add (vector unsigned long long,
16050 vector unsigned long long);
16052 int vec_all_eq (vector long long, vector long long);
16053 int vec_all_eq (vector unsigned long long, vector unsigned long long);
16054 int vec_all_ge (vector long long, vector long long);
16055 int vec_all_ge (vector unsigned long long, vector unsigned long long);
16056 int vec_all_gt (vector long long, vector long long);
16057 int vec_all_gt (vector unsigned long long, vector unsigned long long);
16058 int vec_all_le (vector long long, vector long long);
16059 int vec_all_le (vector unsigned long long, vector unsigned long long);
16060 int vec_all_lt (vector long long, vector long long);
16061 int vec_all_lt (vector unsigned long long, vector unsigned long long);
16062 int vec_all_ne (vector long long, vector long long);
16063 int vec_all_ne (vector unsigned long long, vector unsigned long long);
16065 int vec_any_eq (vector long long, vector long long);
16066 int vec_any_eq (vector unsigned long long, vector unsigned long long);
16067 int vec_any_ge (vector long long, vector long long);
16068 int vec_any_ge (vector unsigned long long, vector unsigned long long);
16069 int vec_any_gt (vector long long, vector long long);
16070 int vec_any_gt (vector unsigned long long, vector unsigned long long);
16071 int vec_any_le (vector long long, vector long long);
16072 int vec_any_le (vector unsigned long long, vector unsigned long long);
16073 int vec_any_lt (vector long long, vector long long);
16074 int vec_any_lt (vector unsigned long long, vector unsigned long long);
16075 int vec_any_ne (vector long long, vector long long);
16076 int vec_any_ne (vector unsigned long long, vector unsigned long long);
16078 vector long long vec_eqv (vector long long, vector long long);
16079 vector long long vec_eqv (vector bool long long, vector long long);
16080 vector long long vec_eqv (vector long long, vector bool long long);
16081 vector unsigned long long vec_eqv (vector unsigned long long,
16082 vector unsigned long long);
16083 vector unsigned long long vec_eqv (vector bool long long,
16084 vector unsigned long long);
16085 vector unsigned long long vec_eqv (vector unsigned long long,
16086 vector bool long long);
16087 vector int vec_eqv (vector int, vector int);
16088 vector int vec_eqv (vector bool int, vector int);
16089 vector int vec_eqv (vector int, vector bool int);
16090 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
16091 vector unsigned int vec_eqv (vector bool unsigned int,
16092 vector unsigned int);
16093 vector unsigned int vec_eqv (vector unsigned int,
16094 vector bool unsigned int);
16095 vector short vec_eqv (vector short, vector short);
16096 vector short vec_eqv (vector bool short, vector short);
16097 vector short vec_eqv (vector short, vector bool short);
16098 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
16099 vector unsigned short vec_eqv (vector bool unsigned short,
16100 vector unsigned short);
16101 vector unsigned short vec_eqv (vector unsigned short,
16102 vector bool unsigned short);
16103 vector signed char vec_eqv (vector signed char, vector signed char);
16104 vector signed char vec_eqv (vector bool signed char, vector signed char);
16105 vector signed char vec_eqv (vector signed char, vector bool signed char);
16106 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
16107 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
16108 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
16110 vector long long vec_max (vector long long, vector long long);
16111 vector unsigned long long vec_max (vector unsigned long long,
16112 vector unsigned long long);
16114 vector signed int vec_mergee (vector signed int, vector signed int);
16115 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
16116 vector bool int vec_mergee (vector bool int, vector bool int);
16118 vector signed int vec_mergeo (vector signed int, vector signed int);
16119 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
16120 vector bool int vec_mergeo (vector bool int, vector bool int);
16122 vector long long vec_min (vector long long, vector long long);
16123 vector unsigned long long vec_min (vector unsigned long long,
16124 vector unsigned long long);
16126 vector long long vec_nand (vector long long, vector long long);
16127 vector long long vec_nand (vector bool long long, vector long long);
16128 vector long long vec_nand (vector long long, vector bool long long);
16129 vector unsigned long long vec_nand (vector unsigned long long,
16130 vector unsigned long long);
16131 vector unsigned long long vec_nand (vector bool long long,
16132 vector unsigned long long);
16133 vector unsigned long long vec_nand (vector unsigned long long,
16134 vector bool long long);
16135 vector int vec_nand (vector int, vector int);
16136 vector int vec_nand (vector bool int, vector int);
16137 vector int vec_nand (vector int, vector bool int);
16138 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
16139 vector unsigned int vec_nand (vector bool unsigned int,
16140 vector unsigned int);
16141 vector unsigned int vec_nand (vector unsigned int,
16142 vector bool unsigned int);
16143 vector short vec_nand (vector short, vector short);
16144 vector short vec_nand (vector bool short, vector short);
16145 vector short vec_nand (vector short, vector bool short);
16146 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
16147 vector unsigned short vec_nand (vector bool unsigned short,
16148 vector unsigned short);
16149 vector unsigned short vec_nand (vector unsigned short,
16150 vector bool unsigned short);
16151 vector signed char vec_nand (vector signed char, vector signed char);
16152 vector signed char vec_nand (vector bool signed char, vector signed char);
16153 vector signed char vec_nand (vector signed char, vector bool signed char);
16154 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
16155 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
16156 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
16158 vector long long vec_orc (vector long long, vector long long);
16159 vector long long vec_orc (vector bool long long, vector long long);
16160 vector long long vec_orc (vector long long, vector bool long long);
16161 vector unsigned long long vec_orc (vector unsigned long long,
16162 vector unsigned long long);
16163 vector unsigned long long vec_orc (vector bool long long,
16164 vector unsigned long long);
16165 vector unsigned long long vec_orc (vector unsigned long long,
16166 vector bool long long);
16167 vector int vec_orc (vector int, vector int);
16168 vector int vec_orc (vector bool int, vector int);
16169 vector int vec_orc (vector int, vector bool int);
16170 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
16171 vector unsigned int vec_orc (vector bool unsigned int,
16172 vector unsigned int);
16173 vector unsigned int vec_orc (vector unsigned int,
16174 vector bool unsigned int);
16175 vector short vec_orc (vector short, vector short);
16176 vector short vec_orc (vector bool short, vector short);
16177 vector short vec_orc (vector short, vector bool short);
16178 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
16179 vector unsigned short vec_orc (vector bool unsigned short,
16180 vector unsigned short);
16181 vector unsigned short vec_orc (vector unsigned short,
16182 vector bool unsigned short);
16183 vector signed char vec_orc (vector signed char, vector signed char);
16184 vector signed char vec_orc (vector bool signed char, vector signed char);
16185 vector signed char vec_orc (vector signed char, vector bool signed char);
16186 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
16187 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
16188 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
16190 vector int vec_pack (vector long long, vector long long);
16191 vector unsigned int vec_pack (vector unsigned long long,
16192 vector unsigned long long);
16193 vector bool int vec_pack (vector bool long long, vector bool long long);
16195 vector int vec_packs (vector long long, vector long long);
16196 vector unsigned int vec_packs (vector unsigned long long,
16197 vector unsigned long long);
16199 vector unsigned int vec_packsu (vector long long, vector long long);
16200 vector unsigned int vec_packsu (vector unsigned long long,
16201 vector unsigned long long);
16203 vector long long vec_rl (vector long long,
16204 vector unsigned long long);
16205 vector long long vec_rl (vector unsigned long long,
16206 vector unsigned long long);
16208 vector long long vec_sl (vector long long, vector unsigned long long);
16209 vector long long vec_sl (vector unsigned long long,
16210 vector unsigned long long);
16212 vector long long vec_sr (vector long long, vector unsigned long long);
16213 vector unsigned long long char vec_sr (vector unsigned long long,
16214 vector unsigned long long);
16216 vector long long vec_sra (vector long long, vector unsigned long long);
16217 vector unsigned long long vec_sra (vector unsigned long long,
16218 vector unsigned long long);
16220 vector long long vec_sub (vector long long, vector long long);
16221 vector unsigned long long vec_sub (vector unsigned long long,
16222 vector unsigned long long);
16224 vector long long vec_unpackh (vector int);
16225 vector unsigned long long vec_unpackh (vector unsigned int);
16227 vector long long vec_unpackl (vector int);
16228 vector unsigned long long vec_unpackl (vector unsigned int);
16230 vector long long vec_vaddudm (vector long long, vector long long);
16231 vector long long vec_vaddudm (vector bool long long, vector long long);
16232 vector long long vec_vaddudm (vector long long, vector bool long long);
16233 vector unsigned long long vec_vaddudm (vector unsigned long long,
16234 vector unsigned long long);
16235 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
16236 vector unsigned long long);
16237 vector unsigned long long vec_vaddudm (vector unsigned long long,
16238 vector bool unsigned long long);
16240 vector long long vec_vbpermq (vector signed char, vector signed char);
16241 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
16243 vector long long vec_cntlz (vector long long);
16244 vector unsigned long long vec_cntlz (vector unsigned long long);
16245 vector int vec_cntlz (vector int);
16246 vector unsigned int vec_cntlz (vector int);
16247 vector short vec_cntlz (vector short);
16248 vector unsigned short vec_cntlz (vector unsigned short);
16249 vector signed char vec_cntlz (vector signed char);
16250 vector unsigned char vec_cntlz (vector unsigned char);
16252 vector long long vec_vclz (vector long long);
16253 vector unsigned long long vec_vclz (vector unsigned long long);
16254 vector int vec_vclz (vector int);
16255 vector unsigned int vec_vclz (vector int);
16256 vector short vec_vclz (vector short);
16257 vector unsigned short vec_vclz (vector unsigned short);
16258 vector signed char vec_vclz (vector signed char);
16259 vector unsigned char vec_vclz (vector unsigned char);
16261 vector signed char vec_vclzb (vector signed char);
16262 vector unsigned char vec_vclzb (vector unsigned char);
16264 vector long long vec_vclzd (vector long long);
16265 vector unsigned long long vec_vclzd (vector unsigned long long);
16267 vector short vec_vclzh (vector short);
16268 vector unsigned short vec_vclzh (vector unsigned short);
16270 vector int vec_vclzw (vector int);
16271 vector unsigned int vec_vclzw (vector int);
16273 vector signed char vec_vgbbd (vector signed char);
16274 vector unsigned char vec_vgbbd (vector unsigned char);
16276 vector long long vec_vmaxsd (vector long long, vector long long);
16278 vector unsigned long long vec_vmaxud (vector unsigned long long,
16279 unsigned vector long long);
16281 vector long long vec_vminsd (vector long long, vector long long);
16283 vector unsigned long long vec_vminud (vector long long,
16286 vector int vec_vpksdss (vector long long, vector long long);
16287 vector unsigned int vec_vpksdss (vector long long, vector long long);
16289 vector unsigned int vec_vpkudus (vector unsigned long long,
16290 vector unsigned long long);
16292 vector int vec_vpkudum (vector long long, vector long long);
16293 vector unsigned int vec_vpkudum (vector unsigned long long,
16294 vector unsigned long long);
16295 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
16297 vector long long vec_vpopcnt (vector long long);
16298 vector unsigned long long vec_vpopcnt (vector unsigned long long);
16299 vector int vec_vpopcnt (vector int);
16300 vector unsigned int vec_vpopcnt (vector int);
16301 vector short vec_vpopcnt (vector short);
16302 vector unsigned short vec_vpopcnt (vector unsigned short);
16303 vector signed char vec_vpopcnt (vector signed char);
16304 vector unsigned char vec_vpopcnt (vector unsigned char);
16306 vector signed char vec_vpopcntb (vector signed char);
16307 vector unsigned char vec_vpopcntb (vector unsigned char);
16309 vector long long vec_vpopcntd (vector long long);
16310 vector unsigned long long vec_vpopcntd (vector unsigned long long);
16312 vector short vec_vpopcnth (vector short);
16313 vector unsigned short vec_vpopcnth (vector unsigned short);
16315 vector int vec_vpopcntw (vector int);
16316 vector unsigned int vec_vpopcntw (vector int);
16318 vector long long vec_vrld (vector long long, vector unsigned long long);
16319 vector unsigned long long vec_vrld (vector unsigned long long,
16320 vector unsigned long long);
16322 vector long long vec_vsld (vector long long, vector unsigned long long);
16323 vector long long vec_vsld (vector unsigned long long,
16324 vector unsigned long long);
16326 vector long long vec_vsrad (vector long long, vector unsigned long long);
16327 vector unsigned long long vec_vsrad (vector unsigned long long,
16328 vector unsigned long long);
16330 vector long long vec_vsrd (vector long long, vector unsigned long long);
16331 vector unsigned long long char vec_vsrd (vector unsigned long long,
16332 vector unsigned long long);
16334 vector long long vec_vsubudm (vector long long, vector long long);
16335 vector long long vec_vsubudm (vector bool long long, vector long long);
16336 vector long long vec_vsubudm (vector long long, vector bool long long);
16337 vector unsigned long long vec_vsubudm (vector unsigned long long,
16338 vector unsigned long long);
16339 vector unsigned long long vec_vsubudm (vector bool long long,
16340 vector unsigned long long);
16341 vector unsigned long long vec_vsubudm (vector unsigned long long,
16342 vector bool long long);
16344 vector long long vec_vupkhsw (vector int);
16345 vector unsigned long long vec_vupkhsw (vector unsigned int);
16347 vector long long vec_vupklsw (vector int);
16348 vector unsigned long long vec_vupklsw (vector int);
16351 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16352 instruction set is available, the following additional functions are
16353 available for 64-bit targets. New vector types
16354 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
16355 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
16358 The normal vector extract, and set operations work on
16359 @var{vector __int128_t} and @var{vector __uint128_t} types,
16360 but the index value must be 0.
16363 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
16364 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
16366 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
16367 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
16369 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
16370 vector __int128_t);
16371 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
16372 vector __uint128_t);
16374 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
16375 vector __int128_t);
16376 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
16377 vector __uint128_t);
16379 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
16380 vector __int128_t);
16381 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
16382 vector __uint128_t);
16384 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
16385 vector __int128_t);
16386 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
16387 vector __uint128_t);
16389 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
16390 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
16392 __int128_t vec_vsubuqm (__int128_t, __int128_t);
16393 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
16395 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
16396 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
16397 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
16398 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
16399 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
16400 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
16401 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
16402 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
16403 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
16404 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
16407 If the cryptographic instructions are enabled (@option{-mcrypto} or
16408 @option{-mcpu=power8}), the following builtins are enabled.
16411 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
16413 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
16414 vector unsigned long long);
16416 vector unsigned long long __builtin_crypto_vcipherlast
16417 (vector unsigned long long,
16418 vector unsigned long long);
16420 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
16421 vector unsigned long long);
16423 vector unsigned long long __builtin_crypto_vncipherlast
16424 (vector unsigned long long,
16425 vector unsigned long long);
16427 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
16428 vector unsigned char,
16429 vector unsigned char);
16431 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
16432 vector unsigned short,
16433 vector unsigned short);
16435 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
16436 vector unsigned int,
16437 vector unsigned int);
16439 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
16440 vector unsigned long long,
16441 vector unsigned long long);
16443 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
16444 vector unsigned char);
16446 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
16447 vector unsigned short);
16449 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
16450 vector unsigned int);
16452 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
16453 vector unsigned long long);
16455 vector unsigned long long __builtin_crypto_vshasigmad
16456 (vector unsigned long long, int, int);
16458 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
16462 The second argument to the @var{__builtin_crypto_vshasigmad} and
16463 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
16464 integer that is 0 or 1. The third argument to these builtin functions
16465 must be a constant integer in the range of 0 to 15.
16467 @node PowerPC Hardware Transactional Memory Built-in Functions
16468 @subsection PowerPC Hardware Transactional Memory Built-in Functions
16469 GCC provides two interfaces for accessing the Hardware Transactional
16470 Memory (HTM) instructions available on some of the PowerPC family
16471 of processors (eg, POWER8). The two interfaces come in a low level
16472 interface, consisting of built-in functions specific to PowerPC and a
16473 higher level interface consisting of inline functions that are common
16474 between PowerPC and S/390.
16476 @subsubsection PowerPC HTM Low Level Built-in Functions
16478 The following low level built-in functions are available with
16479 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16480 They all generate the machine instruction that is part of the name.
16482 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16483 the full 4-bit condition register value set by their associated hardware
16484 instruction. The header file @code{htmintrin.h} defines some macros that can
16485 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16486 returns a simple true or false value depending on whether a transaction was
16487 successfully started or not. The arguments of the builtins match exactly the
16488 type and order of the associated hardware instruction's operands, except for
16489 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16490 Refer to the ISA manual for a description of each instruction's operands.
16493 unsigned int __builtin_tbegin (unsigned int)
16494 unsigned int __builtin_tend (unsigned int)
16496 unsigned int __builtin_tabort (unsigned int)
16497 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16498 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16499 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16500 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16502 unsigned int __builtin_tcheck (void)
16503 unsigned int __builtin_treclaim (unsigned int)
16504 unsigned int __builtin_trechkpt (void)
16505 unsigned int __builtin_tsr (unsigned int)
16508 In addition to the above HTM built-ins, we have added built-ins for
16509 some common extended mnemonics of the HTM instructions:
16512 unsigned int __builtin_tendall (void)
16513 unsigned int __builtin_tresume (void)
16514 unsigned int __builtin_tsuspend (void)
16517 Note that the semantics of the above HTM builtins are required to mimic
16518 the locking semantics used for critical sections. Builtins that are used
16519 to create a new transaction or restart a suspended transaction must have
16520 lock acquisition like semantics while those builtins that end or suspend a
16521 transaction must have lock release like semantics. Specifically, this must
16522 mimic lock semantics as specified by C++11, for example: Lock acquisition is
16523 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
16524 that returns 0, and lock release is as-if an execution of
16525 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
16526 implicit implementation-defined lock used for all transactions. The HTM
16527 instructions associated with with the builtins inherently provide the
16528 correct acquisition and release hardware barriers required. However,
16529 the compiler must also be prohibited from moving loads and stores across
16530 the builtins in a way that would violate their semantics. This has been
16531 accomplished by adding memory barriers to the associated HTM instructions
16532 (which is a conservative approach to provide acquire and release semantics).
16533 Earlier versions of the compiler did not treat the HTM instructions as
16534 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
16535 be used to determine whether the current compiler treats HTM instructions
16536 as memory barriers or not. This allows the user to explicitly add memory
16537 barriers to their code when using an older version of the compiler.
16539 The following set of built-in functions are available to gain access
16540 to the HTM specific special purpose registers.
16543 unsigned long __builtin_get_texasr (void)
16544 unsigned long __builtin_get_texasru (void)
16545 unsigned long __builtin_get_tfhar (void)
16546 unsigned long __builtin_get_tfiar (void)
16548 void __builtin_set_texasr (unsigned long);
16549 void __builtin_set_texasru (unsigned long);
16550 void __builtin_set_tfhar (unsigned long);
16551 void __builtin_set_tfiar (unsigned long);
16554 Example usage of these low level built-in functions may look like:
16557 #include <htmintrin.h>
16559 int num_retries = 10;
16563 if (__builtin_tbegin (0))
16565 /* Transaction State Initiated. */
16566 if (is_locked (lock))
16567 __builtin_tabort (0);
16568 ... transaction code...
16569 __builtin_tend (0);
16574 /* Transaction State Failed. Use locks if the transaction
16575 failure is "persistent" or we've tried too many times. */
16576 if (num_retries-- <= 0
16577 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16579 acquire_lock (lock);
16580 ... non transactional fallback path...
16581 release_lock (lock);
16588 One final built-in function has been added that returns the value of
16589 the 2-bit Transaction State field of the Machine Status Register (MSR)
16590 as stored in @code{CR0}.
16593 unsigned long __builtin_ttest (void)
16596 This built-in can be used to determine the current transaction state
16597 using the following code example:
16600 #include <htmintrin.h>
16602 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16604 if (tx_state == _HTM_TRANSACTIONAL)
16606 /* Code to use in transactional state. */
16608 else if (tx_state == _HTM_NONTRANSACTIONAL)
16610 /* Code to use in non-transactional state. */
16612 else if (tx_state == _HTM_SUSPENDED)
16614 /* Code to use in transaction suspended state. */
16618 @subsubsection PowerPC HTM High Level Inline Functions
16620 The following high level HTM interface is made available by including
16621 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16622 where CPU is `power8' or later. This interface is common between PowerPC
16623 and S/390, allowing users to write one HTM source implementation that
16624 can be compiled and executed on either system.
16627 long __TM_simple_begin (void)
16628 long __TM_begin (void* const TM_buff)
16629 long __TM_end (void)
16630 void __TM_abort (void)
16631 void __TM_named_abort (unsigned char const code)
16632 void __TM_resume (void)
16633 void __TM_suspend (void)
16635 long __TM_is_user_abort (void* const TM_buff)
16636 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16637 long __TM_is_illegal (void* const TM_buff)
16638 long __TM_is_footprint_exceeded (void* const TM_buff)
16639 long __TM_nesting_depth (void* const TM_buff)
16640 long __TM_is_nested_too_deep(void* const TM_buff)
16641 long __TM_is_conflict(void* const TM_buff)
16642 long __TM_is_failure_persistent(void* const TM_buff)
16643 long __TM_failure_address(void* const TM_buff)
16644 long long __TM_failure_code(void* const TM_buff)
16647 Using these common set of HTM inline functions, we can create
16648 a more portable version of the HTM example in the previous
16649 section that will work on either PowerPC or S/390:
16652 #include <htmxlintrin.h>
16654 int num_retries = 10;
16655 TM_buff_type TM_buff;
16659 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16661 /* Transaction State Initiated. */
16662 if (is_locked (lock))
16664 ... transaction code...
16670 /* Transaction State Failed. Use locks if the transaction
16671 failure is "persistent" or we've tried too many times. */
16672 if (num_retries-- <= 0
16673 || __TM_is_failure_persistent (TM_buff))
16675 acquire_lock (lock);
16676 ... non transactional fallback path...
16677 release_lock (lock);
16684 @node RX Built-in Functions
16685 @subsection RX Built-in Functions
16686 GCC supports some of the RX instructions which cannot be expressed in
16687 the C programming language via the use of built-in functions. The
16688 following functions are supported:
16690 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16691 Generates the @code{brk} machine instruction.
16694 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16695 Generates the @code{clrpsw} machine instruction to clear the specified
16696 bit in the processor status word.
16699 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16700 Generates the @code{int} machine instruction to generate an interrupt
16701 with the specified value.
16704 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16705 Generates the @code{machi} machine instruction to add the result of
16706 multiplying the top 16 bits of the two arguments into the
16710 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16711 Generates the @code{maclo} machine instruction to add the result of
16712 multiplying the bottom 16 bits of the two arguments into the
16716 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16717 Generates the @code{mulhi} machine instruction to place the result of
16718 multiplying the top 16 bits of the two arguments into the
16722 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16723 Generates the @code{mullo} machine instruction to place the result of
16724 multiplying the bottom 16 bits of the two arguments into the
16728 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16729 Generates the @code{mvfachi} machine instruction to read the top
16730 32 bits of the accumulator.
16733 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16734 Generates the @code{mvfacmi} machine instruction to read the middle
16735 32 bits of the accumulator.
16738 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16739 Generates the @code{mvfc} machine instruction which reads the control
16740 register specified in its argument and returns its value.
16743 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16744 Generates the @code{mvtachi} machine instruction to set the top
16745 32 bits of the accumulator.
16748 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16749 Generates the @code{mvtaclo} machine instruction to set the bottom
16750 32 bits of the accumulator.
16753 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16754 Generates the @code{mvtc} machine instruction which sets control
16755 register number @code{reg} to @code{val}.
16758 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16759 Generates the @code{mvtipl} machine instruction set the interrupt
16763 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16764 Generates the @code{racw} machine instruction to round the accumulator
16765 according to the specified mode.
16768 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16769 Generates the @code{revw} machine instruction which swaps the bytes in
16770 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16771 and also bits 16--23 occupy bits 24--31 and vice versa.
16774 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16775 Generates the @code{rmpa} machine instruction which initiates a
16776 repeated multiply and accumulate sequence.
16779 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16780 Generates the @code{round} machine instruction which returns the
16781 floating-point argument rounded according to the current rounding mode
16782 set in the floating-point status word register.
16785 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16786 Generates the @code{sat} machine instruction which returns the
16787 saturated value of the argument.
16790 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16791 Generates the @code{setpsw} machine instruction to set the specified
16792 bit in the processor status word.
16795 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16796 Generates the @code{wait} machine instruction.
16799 @node S/390 System z Built-in Functions
16800 @subsection S/390 System z Built-in Functions
16801 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16802 Generates the @code{tbegin} machine instruction starting a
16803 non-constrained hardware transaction. If the parameter is non-NULL the
16804 memory area is used to store the transaction diagnostic buffer and
16805 will be passed as first operand to @code{tbegin}. This buffer can be
16806 defined using the @code{struct __htm_tdb} C struct defined in
16807 @code{htmintrin.h} and must reside on a double-word boundary. The
16808 second tbegin operand is set to @code{0xff0c}. This enables
16809 save/restore of all GPRs and disables aborts for FPR and AR
16810 manipulations inside the transaction body. The condition code set by
16811 the tbegin instruction is returned as integer value. The tbegin
16812 instruction by definition overwrites the content of all FPRs. The
16813 compiler will generate code which saves and restores the FPRs. For
16814 soft-float code it is recommended to used the @code{*_nofloat}
16815 variant. In order to prevent a TDB from being written it is required
16816 to pass a constant zero value as parameter. Passing a zero value
16817 through a variable is not sufficient. Although modifications of
16818 access registers inside the transaction will not trigger an
16819 transaction abort it is not supported to actually modify them. Access
16820 registers do not get saved when entering a transaction. They will have
16821 undefined state when reaching the abort code.
16824 Macros for the possible return codes of tbegin are defined in the
16825 @code{htmintrin.h} header file:
16828 @item _HTM_TBEGIN_STARTED
16829 @code{tbegin} has been executed as part of normal processing. The
16830 transaction body is supposed to be executed.
16831 @item _HTM_TBEGIN_INDETERMINATE
16832 The transaction was aborted due to an indeterminate condition which
16833 might be persistent.
16834 @item _HTM_TBEGIN_TRANSIENT
16835 The transaction aborted due to a transient failure. The transaction
16836 should be re-executed in that case.
16837 @item _HTM_TBEGIN_PERSISTENT
16838 The transaction aborted due to a persistent failure. Re-execution
16839 under same circumstances will not be productive.
16842 @defmac _HTM_FIRST_USER_ABORT_CODE
16843 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16844 specifies the first abort code which can be used for
16845 @code{__builtin_tabort}. Values below this threshold are reserved for
16849 @deftp {Data type} {struct __htm_tdb}
16850 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16851 the structure of the transaction diagnostic block as specified in the
16852 Principles of Operation manual chapter 5-91.
16855 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16856 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16857 Using this variant in code making use of FPRs will leave the FPRs in
16858 undefined state when entering the transaction abort handler code.
16861 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16862 In addition to @code{__builtin_tbegin} a loop for transient failures
16863 is generated. If tbegin returns a condition code of 2 the transaction
16864 will be retried as often as specified in the second argument. The
16865 perform processor assist instruction is used to tell the CPU about the
16866 number of fails so far.
16869 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16870 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16871 restores. Using this variant in code making use of FPRs will leave
16872 the FPRs in undefined state when entering the transaction abort
16876 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16877 Generates the @code{tbeginc} machine instruction starting a constrained
16878 hardware transaction. The second operand is set to @code{0xff08}.
16881 @deftypefn {Built-in Function} int __builtin_tend (void)
16882 Generates the @code{tend} machine instruction finishing a transaction
16883 and making the changes visible to other threads. The condition code
16884 generated by tend is returned as integer value.
16887 @deftypefn {Built-in Function} void __builtin_tabort (int)
16888 Generates the @code{tabort} machine instruction with the specified
16889 abort code. Abort codes from 0 through 255 are reserved and will
16890 result in an error message.
16893 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16894 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16895 integer parameter is loaded into rX and a value of zero is loaded into
16896 rY. The integer parameter specifies the number of times the
16897 transaction repeatedly aborted.
16900 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16901 Generates the @code{etnd} machine instruction. The current nesting
16902 depth is returned as integer value. For a nesting depth of 0 the code
16903 is not executed as part of an transaction.
16906 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16908 Generates the @code{ntstg} machine instruction. The second argument
16909 is written to the first arguments location. The store operation will
16910 not be rolled-back in case of an transaction abort.
16913 @node SH Built-in Functions
16914 @subsection SH Built-in Functions
16915 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16916 families of processors:
16918 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16919 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16920 used by system code that manages threads and execution contexts. The compiler
16921 normally does not generate code that modifies the contents of @samp{GBR} and
16922 thus the value is preserved across function calls. Changing the @samp{GBR}
16923 value in user code must be done with caution, since the compiler might use
16924 @samp{GBR} in order to access thread local variables.
16928 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16929 Returns the value that is currently set in the @samp{GBR} register.
16930 Memory loads and stores that use the thread pointer as a base address are
16931 turned into @samp{GBR} based displacement loads and stores, if possible.
16939 int get_tcb_value (void)
16941 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16942 return ((my_tcb*)__builtin_thread_pointer ())->c;
16948 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16949 Returns the value that is currently set in the @samp{FPSCR} register.
16952 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16953 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16954 preserving the current values of the FR, SZ and PR bits.
16957 @node SPARC VIS Built-in Functions
16958 @subsection SPARC VIS Built-in Functions
16960 GCC supports SIMD operations on the SPARC using both the generic vector
16961 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16962 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16963 switch, the VIS extension is exposed as the following built-in functions:
16966 typedef int v1si __attribute__ ((vector_size (4)));
16967 typedef int v2si __attribute__ ((vector_size (8)));
16968 typedef short v4hi __attribute__ ((vector_size (8)));
16969 typedef short v2hi __attribute__ ((vector_size (4)));
16970 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16971 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16973 void __builtin_vis_write_gsr (int64_t);
16974 int64_t __builtin_vis_read_gsr (void);
16976 void * __builtin_vis_alignaddr (void *, long);
16977 void * __builtin_vis_alignaddrl (void *, long);
16978 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16979 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16980 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16981 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16983 v4hi __builtin_vis_fexpand (v4qi);
16985 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16986 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16987 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16988 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16989 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16990 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16991 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16993 v4qi __builtin_vis_fpack16 (v4hi);
16994 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16995 v2hi __builtin_vis_fpackfix (v2si);
16996 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16998 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
17000 long __builtin_vis_edge8 (void *, void *);
17001 long __builtin_vis_edge8l (void *, void *);
17002 long __builtin_vis_edge16 (void *, void *);
17003 long __builtin_vis_edge16l (void *, void *);
17004 long __builtin_vis_edge32 (void *, void *);
17005 long __builtin_vis_edge32l (void *, void *);
17007 long __builtin_vis_fcmple16 (v4hi, v4hi);
17008 long __builtin_vis_fcmple32 (v2si, v2si);
17009 long __builtin_vis_fcmpne16 (v4hi, v4hi);
17010 long __builtin_vis_fcmpne32 (v2si, v2si);
17011 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
17012 long __builtin_vis_fcmpgt32 (v2si, v2si);
17013 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
17014 long __builtin_vis_fcmpeq32 (v2si, v2si);
17016 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
17017 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
17018 v2si __builtin_vis_fpadd32 (v2si, v2si);
17019 v1si __builtin_vis_fpadd32s (v1si, v1si);
17020 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
17021 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
17022 v2si __builtin_vis_fpsub32 (v2si, v2si);
17023 v1si __builtin_vis_fpsub32s (v1si, v1si);
17025 long __builtin_vis_array8 (long, long);
17026 long __builtin_vis_array16 (long, long);
17027 long __builtin_vis_array32 (long, long);
17030 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
17031 functions also become available:
17034 long __builtin_vis_bmask (long, long);
17035 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
17036 v2si __builtin_vis_bshufflev2si (v2si, v2si);
17037 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
17038 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
17040 long __builtin_vis_edge8n (void *, void *);
17041 long __builtin_vis_edge8ln (void *, void *);
17042 long __builtin_vis_edge16n (void *, void *);
17043 long __builtin_vis_edge16ln (void *, void *);
17044 long __builtin_vis_edge32n (void *, void *);
17045 long __builtin_vis_edge32ln (void *, void *);
17048 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
17049 functions also become available:
17052 void __builtin_vis_cmask8 (long);
17053 void __builtin_vis_cmask16 (long);
17054 void __builtin_vis_cmask32 (long);
17056 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
17058 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
17059 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
17060 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
17061 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
17062 v2si __builtin_vis_fsll16 (v2si, v2si);
17063 v2si __builtin_vis_fslas16 (v2si, v2si);
17064 v2si __builtin_vis_fsrl16 (v2si, v2si);
17065 v2si __builtin_vis_fsra16 (v2si, v2si);
17067 long __builtin_vis_pdistn (v8qi, v8qi);
17069 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
17071 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
17072 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
17074 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
17075 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
17076 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
17077 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
17078 v2si __builtin_vis_fpadds32 (v2si, v2si);
17079 v1si __builtin_vis_fpadds32s (v1si, v1si);
17080 v2si __builtin_vis_fpsubs32 (v2si, v2si);
17081 v1si __builtin_vis_fpsubs32s (v1si, v1si);
17083 long __builtin_vis_fucmple8 (v8qi, v8qi);
17084 long __builtin_vis_fucmpne8 (v8qi, v8qi);
17085 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
17086 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
17088 float __builtin_vis_fhadds (float, float);
17089 double __builtin_vis_fhaddd (double, double);
17090 float __builtin_vis_fhsubs (float, float);
17091 double __builtin_vis_fhsubd (double, double);
17092 float __builtin_vis_fnhadds (float, float);
17093 double __builtin_vis_fnhaddd (double, double);
17095 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
17096 int64_t __builtin_vis_xmulx (int64_t, int64_t);
17097 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
17100 @node SPU Built-in Functions
17101 @subsection SPU Built-in Functions
17103 GCC provides extensions for the SPU processor as described in the
17104 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
17105 found at @uref{http://cell.scei.co.jp/} or
17106 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
17107 implementation differs in several ways.
17112 The optional extension of specifying vector constants in parentheses is
17116 A vector initializer requires no cast if the vector constant is of the
17117 same type as the variable it is initializing.
17120 If @code{signed} or @code{unsigned} is omitted, the signedness of the
17121 vector type is the default signedness of the base type. The default
17122 varies depending on the operating system, so a portable program should
17123 always specify the signedness.
17126 By default, the keyword @code{__vector} is added. The macro
17127 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
17131 GCC allows using a @code{typedef} name as the type specifier for a
17135 For C, overloaded functions are implemented with macros so the following
17139 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
17143 Since @code{spu_add} is a macro, the vector constant in the example
17144 is treated as four separate arguments. Wrap the entire argument in
17145 parentheses for this to work.
17148 The extended version of @code{__builtin_expect} is not supported.
17152 @emph{Note:} Only the interface described in the aforementioned
17153 specification is supported. Internally, GCC uses built-in functions to
17154 implement the required functionality, but these are not supported and
17155 are subject to change without notice.
17157 @node TI C6X Built-in Functions
17158 @subsection TI C6X Built-in Functions
17160 GCC provides intrinsics to access certain instructions of the TI C6X
17161 processors. These intrinsics, listed below, are available after
17162 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
17163 to C6X instructions.
17167 int _sadd (int, int)
17168 int _ssub (int, int)
17169 int _sadd2 (int, int)
17170 int _ssub2 (int, int)
17171 long long _mpy2 (int, int)
17172 long long _smpy2 (int, int)
17173 int _add4 (int, int)
17174 int _sub4 (int, int)
17175 int _saddu4 (int, int)
17177 int _smpy (int, int)
17178 int _smpyh (int, int)
17179 int _smpyhl (int, int)
17180 int _smpylh (int, int)
17182 int _sshl (int, int)
17183 int _subc (int, int)
17185 int _avg2 (int, int)
17186 int _avgu4 (int, int)
17188 int _clrr (int, int)
17189 int _extr (int, int)
17190 int _extru (int, int)
17196 @node TILE-Gx Built-in Functions
17197 @subsection TILE-Gx Built-in Functions
17199 GCC provides intrinsics to access every instruction of the TILE-Gx
17200 processor. The intrinsics are of the form:
17204 unsigned long long __insn_@var{op} (...)
17208 Where @var{op} is the name of the instruction. Refer to the ISA manual
17209 for the complete list of instructions.
17211 GCC also provides intrinsics to directly access the network registers.
17212 The intrinsics are:
17216 unsigned long long __tile_idn0_receive (void)
17217 unsigned long long __tile_idn1_receive (void)
17218 unsigned long long __tile_udn0_receive (void)
17219 unsigned long long __tile_udn1_receive (void)
17220 unsigned long long __tile_udn2_receive (void)
17221 unsigned long long __tile_udn3_receive (void)
17222 void __tile_idn_send (unsigned long long)
17223 void __tile_udn_send (unsigned long long)
17227 The intrinsic @code{void __tile_network_barrier (void)} is used to
17228 guarantee that no network operations before it are reordered with
17231 @node TILEPro Built-in Functions
17232 @subsection TILEPro Built-in Functions
17234 GCC provides intrinsics to access every instruction of the TILEPro
17235 processor. The intrinsics are of the form:
17239 unsigned __insn_@var{op} (...)
17244 where @var{op} is the name of the instruction. Refer to the ISA manual
17245 for the complete list of instructions.
17247 GCC also provides intrinsics to directly access the network registers.
17248 The intrinsics are:
17252 unsigned __tile_idn0_receive (void)
17253 unsigned __tile_idn1_receive (void)
17254 unsigned __tile_sn_receive (void)
17255 unsigned __tile_udn0_receive (void)
17256 unsigned __tile_udn1_receive (void)
17257 unsigned __tile_udn2_receive (void)
17258 unsigned __tile_udn3_receive (void)
17259 void __tile_idn_send (unsigned)
17260 void __tile_sn_send (unsigned)
17261 void __tile_udn_send (unsigned)
17265 The intrinsic @code{void __tile_network_barrier (void)} is used to
17266 guarantee that no network operations before it are reordered with
17269 @node x86 Built-in Functions
17270 @subsection x86 Built-in Functions
17272 These built-in functions are available for the x86-32 and x86-64 family
17273 of computers, depending on the command-line switches used.
17275 If you specify command-line switches such as @option{-msse},
17276 the compiler could use the extended instruction sets even if the built-ins
17277 are not used explicitly in the program. For this reason, applications
17278 that perform run-time CPU detection must compile separate files for each
17279 supported architecture, using the appropriate flags. In particular,
17280 the file containing the CPU detection code should be compiled without
17283 The following machine modes are available for use with MMX built-in functions
17284 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
17285 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
17286 vector of eight 8-bit integers. Some of the built-in functions operate on
17287 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
17289 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
17290 of two 32-bit floating-point values.
17292 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
17293 floating-point values. Some instructions use a vector of four 32-bit
17294 integers, these use @code{V4SI}. Finally, some instructions operate on an
17295 entire vector register, interpreting it as a 128-bit integer, these use mode
17298 In 64-bit mode, the x86-64 family of processors uses additional built-in
17299 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
17300 floating point and @code{TC} 128-bit complex floating-point values.
17302 The following floating-point built-in functions are available in 64-bit
17303 mode. All of them implement the function that is part of the name.
17306 __float128 __builtin_fabsq (__float128)
17307 __float128 __builtin_copysignq (__float128, __float128)
17310 The following built-in function is always available.
17313 @item void __builtin_ia32_pause (void)
17314 Generates the @code{pause} machine instruction with a compiler memory
17318 The following floating-point built-in functions are made available in the
17322 @item __float128 __builtin_infq (void)
17323 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
17324 @findex __builtin_infq
17326 @item __float128 __builtin_huge_valq (void)
17327 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
17328 @findex __builtin_huge_valq
17331 The following built-in functions are always available and can be used to
17332 check the target platform type.
17334 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
17335 This function runs the CPU detection code to check the type of CPU and the
17336 features supported. This built-in function needs to be invoked along with the built-in functions
17337 to check CPU type and features, @code{__builtin_cpu_is} and
17338 @code{__builtin_cpu_supports}, only when used in a function that is
17339 executed before any constructors are called. The CPU detection code is
17340 automatically executed in a very high priority constructor.
17342 For example, this function has to be used in @code{ifunc} resolvers that
17343 check for CPU type using the built-in functions @code{__builtin_cpu_is}
17344 and @code{__builtin_cpu_supports}, or in constructors on targets that
17345 don't support constructor priority.
17348 static void (*resolve_memcpy (void)) (void)
17350 // ifunc resolvers fire before constructors, explicitly call the init
17352 __builtin_cpu_init ();
17353 if (__builtin_cpu_supports ("ssse3"))
17354 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
17356 return default_memcpy;
17359 void *memcpy (void *, const void *, size_t)
17360 __attribute__ ((ifunc ("resolve_memcpy")));
17365 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
17366 This function returns a positive integer if the run-time CPU
17367 is of type @var{cpuname}
17368 and returns @code{0} otherwise. The following CPU names can be detected:
17384 Intel Core i7 Nehalem CPU.
17387 Intel Core i7 Westmere CPU.
17390 Intel Core i7 Sandy Bridge CPU.
17396 AMD Family 10h CPU.
17399 AMD Family 10h Barcelona CPU.
17402 AMD Family 10h Shanghai CPU.
17405 AMD Family 10h Istanbul CPU.
17408 AMD Family 14h CPU.
17411 AMD Family 15h CPU.
17414 AMD Family 15h Bulldozer version 1.
17417 AMD Family 15h Bulldozer version 2.
17420 AMD Family 15h Bulldozer version 3.
17423 AMD Family 15h Bulldozer version 4.
17426 AMD Family 16h CPU.
17429 AMD Family 17h CPU.
17432 Here is an example:
17434 if (__builtin_cpu_is ("corei7"))
17436 do_corei7 (); // Core i7 specific implementation.
17440 do_generic (); // Generic implementation.
17445 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
17446 This function returns a positive integer if the run-time CPU
17447 supports @var{feature}
17448 and returns @code{0} otherwise. The following features can be detected:
17456 POPCNT instruction.
17464 SSSE3 instructions.
17466 SSE4.1 instructions.
17468 SSE4.2 instructions.
17474 AVX512F instructions.
17477 Here is an example:
17479 if (__builtin_cpu_supports ("popcnt"))
17481 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
17485 count = generic_countbits (n); //generic implementation.
17491 The following built-in functions are made available by @option{-mmmx}.
17492 All of them generate the machine instruction that is part of the name.
17495 v8qi __builtin_ia32_paddb (v8qi, v8qi)
17496 v4hi __builtin_ia32_paddw (v4hi, v4hi)
17497 v2si __builtin_ia32_paddd (v2si, v2si)
17498 v8qi __builtin_ia32_psubb (v8qi, v8qi)
17499 v4hi __builtin_ia32_psubw (v4hi, v4hi)
17500 v2si __builtin_ia32_psubd (v2si, v2si)
17501 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17502 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17503 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17504 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17505 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17506 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17507 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17508 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17509 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17510 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17511 di __builtin_ia32_pand (di, di)
17512 di __builtin_ia32_pandn (di,di)
17513 di __builtin_ia32_por (di, di)
17514 di __builtin_ia32_pxor (di, di)
17515 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17516 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17517 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17518 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17519 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17520 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17521 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17522 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17523 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17524 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17525 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17526 v2si __builtin_ia32_punpckldq (v2si, v2si)
17527 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17528 v4hi __builtin_ia32_packssdw (v2si, v2si)
17529 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17531 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17532 v2si __builtin_ia32_pslld (v2si, v2si)
17533 v1di __builtin_ia32_psllq (v1di, v1di)
17534 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17535 v2si __builtin_ia32_psrld (v2si, v2si)
17536 v1di __builtin_ia32_psrlq (v1di, v1di)
17537 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17538 v2si __builtin_ia32_psrad (v2si, v2si)
17539 v4hi __builtin_ia32_psllwi (v4hi, int)
17540 v2si __builtin_ia32_pslldi (v2si, int)
17541 v1di __builtin_ia32_psllqi (v1di, int)
17542 v4hi __builtin_ia32_psrlwi (v4hi, int)
17543 v2si __builtin_ia32_psrldi (v2si, int)
17544 v1di __builtin_ia32_psrlqi (v1di, int)
17545 v4hi __builtin_ia32_psrawi (v4hi, int)
17546 v2si __builtin_ia32_psradi (v2si, int)
17550 The following built-in functions are made available either with
17551 @option{-msse}, or with a combination of @option{-m3dnow} and
17552 @option{-march=athlon}. All of them generate the machine
17553 instruction that is part of the name.
17556 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17557 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17558 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17559 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17560 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17561 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17562 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17563 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17564 int __builtin_ia32_pmovmskb (v8qi)
17565 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17566 void __builtin_ia32_movntq (di *, di)
17567 void __builtin_ia32_sfence (void)
17570 The following built-in functions are available when @option{-msse} is used.
17571 All of them generate the machine instruction that is part of the name.
17574 int __builtin_ia32_comieq (v4sf, v4sf)
17575 int __builtin_ia32_comineq (v4sf, v4sf)
17576 int __builtin_ia32_comilt (v4sf, v4sf)
17577 int __builtin_ia32_comile (v4sf, v4sf)
17578 int __builtin_ia32_comigt (v4sf, v4sf)
17579 int __builtin_ia32_comige (v4sf, v4sf)
17580 int __builtin_ia32_ucomieq (v4sf, v4sf)
17581 int __builtin_ia32_ucomineq (v4sf, v4sf)
17582 int __builtin_ia32_ucomilt (v4sf, v4sf)
17583 int __builtin_ia32_ucomile (v4sf, v4sf)
17584 int __builtin_ia32_ucomigt (v4sf, v4sf)
17585 int __builtin_ia32_ucomige (v4sf, v4sf)
17586 v4sf __builtin_ia32_addps (v4sf, v4sf)
17587 v4sf __builtin_ia32_subps (v4sf, v4sf)
17588 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17589 v4sf __builtin_ia32_divps (v4sf, v4sf)
17590 v4sf __builtin_ia32_addss (v4sf, v4sf)
17591 v4sf __builtin_ia32_subss (v4sf, v4sf)
17592 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17593 v4sf __builtin_ia32_divss (v4sf, v4sf)
17594 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17595 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17596 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17597 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17598 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17599 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17600 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17601 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17602 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17603 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17604 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17605 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17606 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17607 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17608 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17609 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17610 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17611 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17612 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17613 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17614 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17615 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17616 v4sf __builtin_ia32_minps (v4sf, v4sf)
17617 v4sf __builtin_ia32_minss (v4sf, v4sf)
17618 v4sf __builtin_ia32_andps (v4sf, v4sf)
17619 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17620 v4sf __builtin_ia32_orps (v4sf, v4sf)
17621 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17622 v4sf __builtin_ia32_movss (v4sf, v4sf)
17623 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17624 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17625 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17626 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17627 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17628 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17629 v2si __builtin_ia32_cvtps2pi (v4sf)
17630 int __builtin_ia32_cvtss2si (v4sf)
17631 v2si __builtin_ia32_cvttps2pi (v4sf)
17632 int __builtin_ia32_cvttss2si (v4sf)
17633 v4sf __builtin_ia32_rcpps (v4sf)
17634 v4sf __builtin_ia32_rsqrtps (v4sf)
17635 v4sf __builtin_ia32_sqrtps (v4sf)
17636 v4sf __builtin_ia32_rcpss (v4sf)
17637 v4sf __builtin_ia32_rsqrtss (v4sf)
17638 v4sf __builtin_ia32_sqrtss (v4sf)
17639 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17640 void __builtin_ia32_movntps (float *, v4sf)
17641 int __builtin_ia32_movmskps (v4sf)
17644 The following built-in functions are available when @option{-msse} is used.
17647 @item v4sf __builtin_ia32_loadups (float *)
17648 Generates the @code{movups} machine instruction as a load from memory.
17649 @item void __builtin_ia32_storeups (float *, v4sf)
17650 Generates the @code{movups} machine instruction as a store to memory.
17651 @item v4sf __builtin_ia32_loadss (float *)
17652 Generates the @code{movss} machine instruction as a load from memory.
17653 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17654 Generates the @code{movhps} machine instruction as a load from memory.
17655 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17656 Generates the @code{movlps} machine instruction as a load from memory
17657 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17658 Generates the @code{movhps} machine instruction as a store to memory.
17659 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17660 Generates the @code{movlps} machine instruction as a store to memory.
17663 The following built-in functions are available when @option{-msse2} is used.
17664 All of them generate the machine instruction that is part of the name.
17667 int __builtin_ia32_comisdeq (v2df, v2df)
17668 int __builtin_ia32_comisdlt (v2df, v2df)
17669 int __builtin_ia32_comisdle (v2df, v2df)
17670 int __builtin_ia32_comisdgt (v2df, v2df)
17671 int __builtin_ia32_comisdge (v2df, v2df)
17672 int __builtin_ia32_comisdneq (v2df, v2df)
17673 int __builtin_ia32_ucomisdeq (v2df, v2df)
17674 int __builtin_ia32_ucomisdlt (v2df, v2df)
17675 int __builtin_ia32_ucomisdle (v2df, v2df)
17676 int __builtin_ia32_ucomisdgt (v2df, v2df)
17677 int __builtin_ia32_ucomisdge (v2df, v2df)
17678 int __builtin_ia32_ucomisdneq (v2df, v2df)
17679 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17680 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17681 v2df __builtin_ia32_cmplepd (v2df, v2df)
17682 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17683 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17684 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17685 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17686 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17687 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17688 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17689 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17690 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17691 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17692 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17693 v2df __builtin_ia32_cmplesd (v2df, v2df)
17694 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17695 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17696 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17697 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17698 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17699 v2di __builtin_ia32_paddq (v2di, v2di)
17700 v2di __builtin_ia32_psubq (v2di, v2di)
17701 v2df __builtin_ia32_addpd (v2df, v2df)
17702 v2df __builtin_ia32_subpd (v2df, v2df)
17703 v2df __builtin_ia32_mulpd (v2df, v2df)
17704 v2df __builtin_ia32_divpd (v2df, v2df)
17705 v2df __builtin_ia32_addsd (v2df, v2df)
17706 v2df __builtin_ia32_subsd (v2df, v2df)
17707 v2df __builtin_ia32_mulsd (v2df, v2df)
17708 v2df __builtin_ia32_divsd (v2df, v2df)
17709 v2df __builtin_ia32_minpd (v2df, v2df)
17710 v2df __builtin_ia32_maxpd (v2df, v2df)
17711 v2df __builtin_ia32_minsd (v2df, v2df)
17712 v2df __builtin_ia32_maxsd (v2df, v2df)
17713 v2df __builtin_ia32_andpd (v2df, v2df)
17714 v2df __builtin_ia32_andnpd (v2df, v2df)
17715 v2df __builtin_ia32_orpd (v2df, v2df)
17716 v2df __builtin_ia32_xorpd (v2df, v2df)
17717 v2df __builtin_ia32_movsd (v2df, v2df)
17718 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17719 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17720 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17721 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17722 v4si __builtin_ia32_paddd128 (v4si, v4si)
17723 v2di __builtin_ia32_paddq128 (v2di, v2di)
17724 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17725 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17726 v4si __builtin_ia32_psubd128 (v4si, v4si)
17727 v2di __builtin_ia32_psubq128 (v2di, v2di)
17728 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17729 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17730 v2di __builtin_ia32_pand128 (v2di, v2di)
17731 v2di __builtin_ia32_pandn128 (v2di, v2di)
17732 v2di __builtin_ia32_por128 (v2di, v2di)
17733 v2di __builtin_ia32_pxor128 (v2di, v2di)
17734 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17735 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17736 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17737 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17738 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17739 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17740 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17741 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17742 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17743 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17744 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17745 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17746 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17747 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17748 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17749 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17750 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17751 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17752 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17753 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17754 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17755 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17756 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17757 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17758 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17759 v2df __builtin_ia32_loadupd (double *)
17760 void __builtin_ia32_storeupd (double *, v2df)
17761 v2df __builtin_ia32_loadhpd (v2df, double const *)
17762 v2df __builtin_ia32_loadlpd (v2df, double const *)
17763 int __builtin_ia32_movmskpd (v2df)
17764 int __builtin_ia32_pmovmskb128 (v16qi)
17765 void __builtin_ia32_movnti (int *, int)
17766 void __builtin_ia32_movnti64 (long long int *, long long int)
17767 void __builtin_ia32_movntpd (double *, v2df)
17768 void __builtin_ia32_movntdq (v2df *, v2df)
17769 v4si __builtin_ia32_pshufd (v4si, int)
17770 v8hi __builtin_ia32_pshuflw (v8hi, int)
17771 v8hi __builtin_ia32_pshufhw (v8hi, int)
17772 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17773 v2df __builtin_ia32_sqrtpd (v2df)
17774 v2df __builtin_ia32_sqrtsd (v2df)
17775 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17776 v2df __builtin_ia32_cvtdq2pd (v4si)
17777 v4sf __builtin_ia32_cvtdq2ps (v4si)
17778 v4si __builtin_ia32_cvtpd2dq (v2df)
17779 v2si __builtin_ia32_cvtpd2pi (v2df)
17780 v4sf __builtin_ia32_cvtpd2ps (v2df)
17781 v4si __builtin_ia32_cvttpd2dq (v2df)
17782 v2si __builtin_ia32_cvttpd2pi (v2df)
17783 v2df __builtin_ia32_cvtpi2pd (v2si)
17784 int __builtin_ia32_cvtsd2si (v2df)
17785 int __builtin_ia32_cvttsd2si (v2df)
17786 long long __builtin_ia32_cvtsd2si64 (v2df)
17787 long long __builtin_ia32_cvttsd2si64 (v2df)
17788 v4si __builtin_ia32_cvtps2dq (v4sf)
17789 v2df __builtin_ia32_cvtps2pd (v4sf)
17790 v4si __builtin_ia32_cvttps2dq (v4sf)
17791 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17792 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17793 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17794 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17795 void __builtin_ia32_clflush (const void *)
17796 void __builtin_ia32_lfence (void)
17797 void __builtin_ia32_mfence (void)
17798 v16qi __builtin_ia32_loaddqu (const char *)
17799 void __builtin_ia32_storedqu (char *, v16qi)
17800 v1di __builtin_ia32_pmuludq (v2si, v2si)
17801 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17802 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17803 v4si __builtin_ia32_pslld128 (v4si, v4si)
17804 v2di __builtin_ia32_psllq128 (v2di, v2di)
17805 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17806 v4si __builtin_ia32_psrld128 (v4si, v4si)
17807 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17808 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17809 v4si __builtin_ia32_psrad128 (v4si, v4si)
17810 v2di __builtin_ia32_pslldqi128 (v2di, int)
17811 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17812 v4si __builtin_ia32_pslldi128 (v4si, int)
17813 v2di __builtin_ia32_psllqi128 (v2di, int)
17814 v2di __builtin_ia32_psrldqi128 (v2di, int)
17815 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17816 v4si __builtin_ia32_psrldi128 (v4si, int)
17817 v2di __builtin_ia32_psrlqi128 (v2di, int)
17818 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17819 v4si __builtin_ia32_psradi128 (v4si, int)
17820 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17821 v2di __builtin_ia32_movq128 (v2di)
17824 The following built-in functions are available when @option{-msse3} is used.
17825 All of them generate the machine instruction that is part of the name.
17828 v2df __builtin_ia32_addsubpd (v2df, v2df)
17829 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17830 v2df __builtin_ia32_haddpd (v2df, v2df)
17831 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17832 v2df __builtin_ia32_hsubpd (v2df, v2df)
17833 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17834 v16qi __builtin_ia32_lddqu (char const *)
17835 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17836 v4sf __builtin_ia32_movshdup (v4sf)
17837 v4sf __builtin_ia32_movsldup (v4sf)
17838 void __builtin_ia32_mwait (unsigned int, unsigned int)
17841 The following built-in functions are available when @option{-mssse3} is used.
17842 All of them generate the machine instruction that is part of the name.
17845 v2si __builtin_ia32_phaddd (v2si, v2si)
17846 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17847 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17848 v2si __builtin_ia32_phsubd (v2si, v2si)
17849 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17850 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17851 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17852 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17853 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17854 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17855 v2si __builtin_ia32_psignd (v2si, v2si)
17856 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17857 v1di __builtin_ia32_palignr (v1di, v1di, int)
17858 v8qi __builtin_ia32_pabsb (v8qi)
17859 v2si __builtin_ia32_pabsd (v2si)
17860 v4hi __builtin_ia32_pabsw (v4hi)
17863 The following built-in functions are available when @option{-mssse3} is used.
17864 All of them generate the machine instruction that is part of the name.
17867 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17868 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17869 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17870 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17871 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17872 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17873 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17874 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17875 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17876 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17877 v4si __builtin_ia32_psignd128 (v4si, v4si)
17878 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17879 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17880 v16qi __builtin_ia32_pabsb128 (v16qi)
17881 v4si __builtin_ia32_pabsd128 (v4si)
17882 v8hi __builtin_ia32_pabsw128 (v8hi)
17885 The following built-in functions are available when @option{-msse4.1} is
17886 used. All of them generate the machine instruction that is part of the
17890 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17891 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17892 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17893 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17894 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17895 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17896 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17897 v2di __builtin_ia32_movntdqa (v2di *);
17898 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17899 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17900 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17901 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17902 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17903 v8hi __builtin_ia32_phminposuw128 (v8hi)
17904 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17905 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17906 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17907 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17908 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17909 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17910 v4si __builtin_ia32_pminud128 (v4si, v4si)
17911 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17912 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17913 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17914 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17915 v2di __builtin_ia32_pmovsxdq128 (v4si)
17916 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17917 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17918 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17919 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17920 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17921 v2di __builtin_ia32_pmovzxdq128 (v4si)
17922 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17923 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17924 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17925 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17926 int __builtin_ia32_ptestc128 (v2di, v2di)
17927 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17928 int __builtin_ia32_ptestz128 (v2di, v2di)
17929 v2df __builtin_ia32_roundpd (v2df, const int)
17930 v4sf __builtin_ia32_roundps (v4sf, const int)
17931 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17932 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17935 The following built-in functions are available when @option{-msse4.1} is
17939 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17940 Generates the @code{insertps} machine instruction.
17941 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17942 Generates the @code{pextrb} machine instruction.
17943 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17944 Generates the @code{pinsrb} machine instruction.
17945 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17946 Generates the @code{pinsrd} machine instruction.
17947 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17948 Generates the @code{pinsrq} machine instruction in 64bit mode.
17951 The following built-in functions are changed to generate new SSE4.1
17952 instructions when @option{-msse4.1} is used.
17955 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17956 Generates the @code{extractps} machine instruction.
17957 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17958 Generates the @code{pextrd} machine instruction.
17959 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17960 Generates the @code{pextrq} machine instruction in 64bit mode.
17963 The following built-in functions are available when @option{-msse4.2} is
17964 used. All of them generate the machine instruction that is part of the
17968 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17969 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17970 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17971 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17972 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17973 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17974 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17975 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17976 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17977 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17978 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17979 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17980 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17981 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17982 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17985 The following built-in functions are available when @option{-msse4.2} is
17989 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17990 Generates the @code{crc32b} machine instruction.
17991 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17992 Generates the @code{crc32w} machine instruction.
17993 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17994 Generates the @code{crc32l} machine instruction.
17995 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17996 Generates the @code{crc32q} machine instruction.
17999 The following built-in functions are changed to generate new SSE4.2
18000 instructions when @option{-msse4.2} is used.
18003 @item int __builtin_popcount (unsigned int)
18004 Generates the @code{popcntl} machine instruction.
18005 @item int __builtin_popcountl (unsigned long)
18006 Generates the @code{popcntl} or @code{popcntq} machine instruction,
18007 depending on the size of @code{unsigned long}.
18008 @item int __builtin_popcountll (unsigned long long)
18009 Generates the @code{popcntq} machine instruction.
18012 The following built-in functions are available when @option{-mavx} is
18013 used. All of them generate the machine instruction that is part of the
18017 v4df __builtin_ia32_addpd256 (v4df,v4df)
18018 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
18019 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
18020 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
18021 v4df __builtin_ia32_andnpd256 (v4df,v4df)
18022 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
18023 v4df __builtin_ia32_andpd256 (v4df,v4df)
18024 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
18025 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
18026 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
18027 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
18028 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
18029 v2df __builtin_ia32_cmppd (v2df,v2df,int)
18030 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
18031 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
18032 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
18033 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
18034 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
18035 v4df __builtin_ia32_cvtdq2pd256 (v4si)
18036 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
18037 v4si __builtin_ia32_cvtpd2dq256 (v4df)
18038 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
18039 v8si __builtin_ia32_cvtps2dq256 (v8sf)
18040 v4df __builtin_ia32_cvtps2pd256 (v4sf)
18041 v4si __builtin_ia32_cvttpd2dq256 (v4df)
18042 v8si __builtin_ia32_cvttps2dq256 (v8sf)
18043 v4df __builtin_ia32_divpd256 (v4df,v4df)
18044 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
18045 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
18046 v4df __builtin_ia32_haddpd256 (v4df,v4df)
18047 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
18048 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
18049 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
18050 v32qi __builtin_ia32_lddqu256 (pcchar)
18051 v32qi __builtin_ia32_loaddqu256 (pcchar)
18052 v4df __builtin_ia32_loadupd256 (pcdouble)
18053 v8sf __builtin_ia32_loadups256 (pcfloat)
18054 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
18055 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
18056 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
18057 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
18058 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
18059 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
18060 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
18061 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
18062 v4df __builtin_ia32_maxpd256 (v4df,v4df)
18063 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
18064 v4df __builtin_ia32_minpd256 (v4df,v4df)
18065 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
18066 v4df __builtin_ia32_movddup256 (v4df)
18067 int __builtin_ia32_movmskpd256 (v4df)
18068 int __builtin_ia32_movmskps256 (v8sf)
18069 v8sf __builtin_ia32_movshdup256 (v8sf)
18070 v8sf __builtin_ia32_movsldup256 (v8sf)
18071 v4df __builtin_ia32_mulpd256 (v4df,v4df)
18072 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
18073 v4df __builtin_ia32_orpd256 (v4df,v4df)
18074 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
18075 v2df __builtin_ia32_pd_pd256 (v4df)
18076 v4df __builtin_ia32_pd256_pd (v2df)
18077 v4sf __builtin_ia32_ps_ps256 (v8sf)
18078 v8sf __builtin_ia32_ps256_ps (v4sf)
18079 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
18080 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
18081 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
18082 v8sf __builtin_ia32_rcpps256 (v8sf)
18083 v4df __builtin_ia32_roundpd256 (v4df,int)
18084 v8sf __builtin_ia32_roundps256 (v8sf,int)
18085 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
18086 v8sf __builtin_ia32_rsqrtps256 (v8sf)
18087 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
18088 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
18089 v4si __builtin_ia32_si_si256 (v8si)
18090 v8si __builtin_ia32_si256_si (v4si)
18091 v4df __builtin_ia32_sqrtpd256 (v4df)
18092 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
18093 v8sf __builtin_ia32_sqrtps256 (v8sf)
18094 void __builtin_ia32_storedqu256 (pchar,v32qi)
18095 void __builtin_ia32_storeupd256 (pdouble,v4df)
18096 void __builtin_ia32_storeups256 (pfloat,v8sf)
18097 v4df __builtin_ia32_subpd256 (v4df,v4df)
18098 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
18099 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
18100 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
18101 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
18102 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
18103 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
18104 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
18105 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
18106 v4sf __builtin_ia32_vbroadcastss (pcfloat)
18107 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
18108 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
18109 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
18110 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
18111 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
18112 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
18113 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
18114 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
18115 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
18116 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
18117 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
18118 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
18119 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
18120 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
18121 v2df __builtin_ia32_vpermilpd (v2df,int)
18122 v4df __builtin_ia32_vpermilpd256 (v4df,int)
18123 v4sf __builtin_ia32_vpermilps (v4sf,int)
18124 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
18125 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
18126 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
18127 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
18128 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
18129 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
18130 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
18131 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
18132 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
18133 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
18134 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
18135 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
18136 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
18137 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
18138 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
18139 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
18140 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
18141 void __builtin_ia32_vzeroall (void)
18142 void __builtin_ia32_vzeroupper (void)
18143 v4df __builtin_ia32_xorpd256 (v4df,v4df)
18144 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
18147 The following built-in functions are available when @option{-mavx2} is
18148 used. All of them generate the machine instruction that is part of the
18152 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
18153 v32qi __builtin_ia32_pabsb256 (v32qi)
18154 v16hi __builtin_ia32_pabsw256 (v16hi)
18155 v8si __builtin_ia32_pabsd256 (v8si)
18156 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
18157 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
18158 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
18159 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
18160 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
18161 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
18162 v8si __builtin_ia32_paddd256 (v8si,v8si)
18163 v4di __builtin_ia32_paddq256 (v4di,v4di)
18164 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
18165 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
18166 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
18167 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
18168 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
18169 v4di __builtin_ia32_andsi256 (v4di,v4di)
18170 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
18171 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
18172 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
18173 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
18174 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
18175 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
18176 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
18177 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
18178 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
18179 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
18180 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
18181 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
18182 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
18183 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
18184 v8si __builtin_ia32_phaddd256 (v8si,v8si)
18185 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
18186 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
18187 v8si __builtin_ia32_phsubd256 (v8si,v8si)
18188 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
18189 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
18190 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
18191 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
18192 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
18193 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
18194 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
18195 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
18196 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
18197 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
18198 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
18199 v8si __builtin_ia32_pminsd256 (v8si,v8si)
18200 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
18201 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
18202 v8si __builtin_ia32_pminud256 (v8si,v8si)
18203 int __builtin_ia32_pmovmskb256 (v32qi)
18204 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
18205 v8si __builtin_ia32_pmovsxbd256 (v16qi)
18206 v4di __builtin_ia32_pmovsxbq256 (v16qi)
18207 v8si __builtin_ia32_pmovsxwd256 (v8hi)
18208 v4di __builtin_ia32_pmovsxwq256 (v8hi)
18209 v4di __builtin_ia32_pmovsxdq256 (v4si)
18210 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
18211 v8si __builtin_ia32_pmovzxbd256 (v16qi)
18212 v4di __builtin_ia32_pmovzxbq256 (v16qi)
18213 v8si __builtin_ia32_pmovzxwd256 (v8hi)
18214 v4di __builtin_ia32_pmovzxwq256 (v8hi)
18215 v4di __builtin_ia32_pmovzxdq256 (v4si)
18216 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
18217 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
18218 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
18219 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
18220 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
18221 v8si __builtin_ia32_pmulld256 (v8si,v8si)
18222 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
18223 v4di __builtin_ia32_por256 (v4di,v4di)
18224 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
18225 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
18226 v8si __builtin_ia32_pshufd256 (v8si,int)
18227 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
18228 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
18229 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
18230 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
18231 v8si __builtin_ia32_psignd256 (v8si,v8si)
18232 v4di __builtin_ia32_pslldqi256 (v4di,int)
18233 v16hi __builtin_ia32_psllwi256 (16hi,int)
18234 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
18235 v8si __builtin_ia32_pslldi256 (v8si,int)
18236 v8si __builtin_ia32_pslld256(v8si,v4si)
18237 v4di __builtin_ia32_psllqi256 (v4di,int)
18238 v4di __builtin_ia32_psllq256(v4di,v2di)
18239 v16hi __builtin_ia32_psrawi256 (v16hi,int)
18240 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
18241 v8si __builtin_ia32_psradi256 (v8si,int)
18242 v8si __builtin_ia32_psrad256 (v8si,v4si)
18243 v4di __builtin_ia32_psrldqi256 (v4di, int)
18244 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
18245 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
18246 v8si __builtin_ia32_psrldi256 (v8si,int)
18247 v8si __builtin_ia32_psrld256 (v8si,v4si)
18248 v4di __builtin_ia32_psrlqi256 (v4di,int)
18249 v4di __builtin_ia32_psrlq256(v4di,v2di)
18250 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
18251 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
18252 v8si __builtin_ia32_psubd256 (v8si,v8si)
18253 v4di __builtin_ia32_psubq256 (v4di,v4di)
18254 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
18255 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
18256 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
18257 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
18258 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
18259 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
18260 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
18261 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
18262 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
18263 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
18264 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
18265 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
18266 v4di __builtin_ia32_pxor256 (v4di,v4di)
18267 v4di __builtin_ia32_movntdqa256 (pv4di)
18268 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
18269 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
18270 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
18271 v4di __builtin_ia32_vbroadcastsi256 (v2di)
18272 v4si __builtin_ia32_pblendd128 (v4si,v4si)
18273 v8si __builtin_ia32_pblendd256 (v8si,v8si)
18274 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
18275 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
18276 v8si __builtin_ia32_pbroadcastd256 (v4si)
18277 v4di __builtin_ia32_pbroadcastq256 (v2di)
18278 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
18279 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
18280 v4si __builtin_ia32_pbroadcastd128 (v4si)
18281 v2di __builtin_ia32_pbroadcastq128 (v2di)
18282 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
18283 v4df __builtin_ia32_permdf256 (v4df,int)
18284 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
18285 v4di __builtin_ia32_permdi256 (v4di,int)
18286 v4di __builtin_ia32_permti256 (v4di,v4di,int)
18287 v4di __builtin_ia32_extract128i256 (v4di,int)
18288 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
18289 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
18290 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
18291 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
18292 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
18293 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
18294 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
18295 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
18296 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
18297 v8si __builtin_ia32_psllv8si (v8si,v8si)
18298 v4si __builtin_ia32_psllv4si (v4si,v4si)
18299 v4di __builtin_ia32_psllv4di (v4di,v4di)
18300 v2di __builtin_ia32_psllv2di (v2di,v2di)
18301 v8si __builtin_ia32_psrav8si (v8si,v8si)
18302 v4si __builtin_ia32_psrav4si (v4si,v4si)
18303 v8si __builtin_ia32_psrlv8si (v8si,v8si)
18304 v4si __builtin_ia32_psrlv4si (v4si,v4si)
18305 v4di __builtin_ia32_psrlv4di (v4di,v4di)
18306 v2di __builtin_ia32_psrlv2di (v2di,v2di)
18307 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
18308 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
18309 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
18310 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
18311 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
18312 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
18313 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
18314 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
18315 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
18316 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
18317 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
18318 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
18319 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
18320 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
18321 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
18322 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
18325 The following built-in functions are available when @option{-maes} is
18326 used. All of them generate the machine instruction that is part of the
18330 v2di __builtin_ia32_aesenc128 (v2di, v2di)
18331 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
18332 v2di __builtin_ia32_aesdec128 (v2di, v2di)
18333 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
18334 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
18335 v2di __builtin_ia32_aesimc128 (v2di)
18338 The following built-in function is available when @option{-mpclmul} is
18342 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
18343 Generates the @code{pclmulqdq} machine instruction.
18346 The following built-in function is available when @option{-mfsgsbase} is
18347 used. All of them generate the machine instruction that is part of the
18351 unsigned int __builtin_ia32_rdfsbase32 (void)
18352 unsigned long long __builtin_ia32_rdfsbase64 (void)
18353 unsigned int __builtin_ia32_rdgsbase32 (void)
18354 unsigned long long __builtin_ia32_rdgsbase64 (void)
18355 void _writefsbase_u32 (unsigned int)
18356 void _writefsbase_u64 (unsigned long long)
18357 void _writegsbase_u32 (unsigned int)
18358 void _writegsbase_u64 (unsigned long long)
18361 The following built-in function is available when @option{-mrdrnd} is
18362 used. All of them generate the machine instruction that is part of the
18366 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
18367 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
18368 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
18371 The following built-in functions are available when @option{-msse4a} is used.
18372 All of them generate the machine instruction that is part of the name.
18375 void __builtin_ia32_movntsd (double *, v2df)
18376 void __builtin_ia32_movntss (float *, v4sf)
18377 v2di __builtin_ia32_extrq (v2di, v16qi)
18378 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
18379 v2di __builtin_ia32_insertq (v2di, v2di)
18380 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
18383 The following built-in functions are available when @option{-mxop} is used.
18385 v2df __builtin_ia32_vfrczpd (v2df)
18386 v4sf __builtin_ia32_vfrczps (v4sf)
18387 v2df __builtin_ia32_vfrczsd (v2df)
18388 v4sf __builtin_ia32_vfrczss (v4sf)
18389 v4df __builtin_ia32_vfrczpd256 (v4df)
18390 v8sf __builtin_ia32_vfrczps256 (v8sf)
18391 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
18392 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
18393 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
18394 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
18395 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
18396 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
18397 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
18398 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
18399 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
18400 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
18401 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
18402 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
18403 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
18404 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
18405 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18406 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
18407 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
18408 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
18409 v4si __builtin_ia32_vpcomequd (v4si, v4si)
18410 v2di __builtin_ia32_vpcomequq (v2di, v2di)
18411 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
18412 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18413 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
18414 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
18415 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
18416 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
18417 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
18418 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
18419 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
18420 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
18421 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
18422 v4si __builtin_ia32_vpcomged (v4si, v4si)
18423 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
18424 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
18425 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
18426 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
18427 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
18428 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
18429 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
18430 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
18431 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
18432 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
18433 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
18434 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
18435 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
18436 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
18437 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
18438 v4si __builtin_ia32_vpcomled (v4si, v4si)
18439 v2di __builtin_ia32_vpcomleq (v2di, v2di)
18440 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
18441 v4si __builtin_ia32_vpcomleud (v4si, v4si)
18442 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
18443 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
18444 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
18445 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
18446 v4si __builtin_ia32_vpcomltd (v4si, v4si)
18447 v2di __builtin_ia32_vpcomltq (v2di, v2di)
18448 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
18449 v4si __builtin_ia32_vpcomltud (v4si, v4si)
18450 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
18451 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
18452 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
18453 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
18454 v4si __builtin_ia32_vpcomned (v4si, v4si)
18455 v2di __builtin_ia32_vpcomneq (v2di, v2di)
18456 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
18457 v4si __builtin_ia32_vpcomneud (v4si, v4si)
18458 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
18459 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
18460 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
18461 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
18462 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
18463 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
18464 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
18465 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
18466 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
18467 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
18468 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
18469 v4si __builtin_ia32_vphaddbd (v16qi)
18470 v2di __builtin_ia32_vphaddbq (v16qi)
18471 v8hi __builtin_ia32_vphaddbw (v16qi)
18472 v2di __builtin_ia32_vphadddq (v4si)
18473 v4si __builtin_ia32_vphaddubd (v16qi)
18474 v2di __builtin_ia32_vphaddubq (v16qi)
18475 v8hi __builtin_ia32_vphaddubw (v16qi)
18476 v2di __builtin_ia32_vphaddudq (v4si)
18477 v4si __builtin_ia32_vphadduwd (v8hi)
18478 v2di __builtin_ia32_vphadduwq (v8hi)
18479 v4si __builtin_ia32_vphaddwd (v8hi)
18480 v2di __builtin_ia32_vphaddwq (v8hi)
18481 v8hi __builtin_ia32_vphsubbw (v16qi)
18482 v2di __builtin_ia32_vphsubdq (v4si)
18483 v4si __builtin_ia32_vphsubwd (v8hi)
18484 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
18485 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
18486 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
18487 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
18488 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
18489 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
18490 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
18491 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
18492 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
18493 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
18494 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
18495 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
18496 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
18497 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
18498 v4si __builtin_ia32_vprotd (v4si, v4si)
18499 v2di __builtin_ia32_vprotq (v2di, v2di)
18500 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18501 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18502 v4si __builtin_ia32_vpshad (v4si, v4si)
18503 v2di __builtin_ia32_vpshaq (v2di, v2di)
18504 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18505 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18506 v4si __builtin_ia32_vpshld (v4si, v4si)
18507 v2di __builtin_ia32_vpshlq (v2di, v2di)
18508 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18511 The following built-in functions are available when @option{-mfma4} is used.
18512 All of them generate the machine instruction that is part of the name.
18515 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18516 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18517 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18518 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18519 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18520 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18521 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18522 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18523 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18524 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18525 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18526 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18527 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18528 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18529 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18530 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18531 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18532 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18533 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18534 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18535 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18536 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18537 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18538 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18539 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18540 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18541 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18542 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18543 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18544 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18545 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18546 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18550 The following built-in functions are available when @option{-mlwp} is used.
18553 void __builtin_ia32_llwpcb16 (void *);
18554 void __builtin_ia32_llwpcb32 (void *);
18555 void __builtin_ia32_llwpcb64 (void *);
18556 void * __builtin_ia32_llwpcb16 (void);
18557 void * __builtin_ia32_llwpcb32 (void);
18558 void * __builtin_ia32_llwpcb64 (void);
18559 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18560 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18561 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18562 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18563 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18564 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18567 The following built-in functions are available when @option{-mbmi} is used.
18568 All of them generate the machine instruction that is part of the name.
18570 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18571 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18574 The following built-in functions are available when @option{-mbmi2} is used.
18575 All of them generate the machine instruction that is part of the name.
18577 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18578 unsigned int _pdep_u32 (unsigned int, unsigned int)
18579 unsigned int _pext_u32 (unsigned int, unsigned int)
18580 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18581 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18582 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18585 The following built-in functions are available when @option{-mlzcnt} is used.
18586 All of them generate the machine instruction that is part of the name.
18588 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18589 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18590 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18593 The following built-in functions are available when @option{-mfxsr} is used.
18594 All of them generate the machine instruction that is part of the name.
18596 void __builtin_ia32_fxsave (void *)
18597 void __builtin_ia32_fxrstor (void *)
18598 void __builtin_ia32_fxsave64 (void *)
18599 void __builtin_ia32_fxrstor64 (void *)
18602 The following built-in functions are available when @option{-mxsave} is used.
18603 All of them generate the machine instruction that is part of the name.
18605 void __builtin_ia32_xsave (void *, long long)
18606 void __builtin_ia32_xrstor (void *, long long)
18607 void __builtin_ia32_xsave64 (void *, long long)
18608 void __builtin_ia32_xrstor64 (void *, long long)
18611 The following built-in functions are available when @option{-mxsaveopt} is used.
18612 All of them generate the machine instruction that is part of the name.
18614 void __builtin_ia32_xsaveopt (void *, long long)
18615 void __builtin_ia32_xsaveopt64 (void *, long long)
18618 The following built-in functions are available when @option{-mtbm} is used.
18619 Both of them generate the immediate form of the bextr machine instruction.
18621 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18622 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18626 The following built-in functions are available when @option{-m3dnow} is used.
18627 All of them generate the machine instruction that is part of the name.
18630 void __builtin_ia32_femms (void)
18631 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18632 v2si __builtin_ia32_pf2id (v2sf)
18633 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18634 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18635 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18636 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18637 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18638 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18639 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18640 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18641 v2sf __builtin_ia32_pfrcp (v2sf)
18642 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18643 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18644 v2sf __builtin_ia32_pfrsqrt (v2sf)
18645 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18646 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18647 v2sf __builtin_ia32_pi2fd (v2si)
18648 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18651 The following built-in functions are available when both @option{-m3dnow}
18652 and @option{-march=athlon} are used. All of them generate the machine
18653 instruction that is part of the name.
18656 v2si __builtin_ia32_pf2iw (v2sf)
18657 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18658 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18659 v2sf __builtin_ia32_pi2fw (v2si)
18660 v2sf __builtin_ia32_pswapdsf (v2sf)
18661 v2si __builtin_ia32_pswapdsi (v2si)
18664 The following built-in functions are available when @option{-mrtm} is used
18665 They are used for restricted transactional memory. These are the internal
18666 low level functions. Normally the functions in
18667 @ref{x86 transactional memory intrinsics} should be used instead.
18670 int __builtin_ia32_xbegin ()
18671 void __builtin_ia32_xend ()
18672 void __builtin_ia32_xabort (status)
18673 int __builtin_ia32_xtest ()
18676 The following built-in functions are available when @option{-mmwaitx} is used.
18677 All of them generate the machine instruction that is part of the name.
18679 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18680 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18683 The following built-in functions are available when @option{-mclzero} is used.
18684 All of them generate the machine instruction that is part of the name.
18686 void __builtin_i32_clzero (void *)
18689 The following built-in functions are available when @option{-mpku} is used.
18690 They generate reads and writes to PKRU.
18692 void __builtin_ia32_wrpkru (unsigned int)
18693 unsigned int __builtin_ia32_rdpkru ()
18696 @node x86 transactional memory intrinsics
18697 @subsection x86 Transactional Memory Intrinsics
18699 These hardware transactional memory intrinsics for x86 allow you to use
18700 memory transactions with RTM (Restricted Transactional Memory).
18701 This support is enabled with the @option{-mrtm} option.
18702 For using HLE (Hardware Lock Elision) see
18703 @ref{x86 specific memory model extensions for transactional memory} instead.
18705 A memory transaction commits all changes to memory in an atomic way,
18706 as visible to other threads. If the transaction fails it is rolled back
18707 and all side effects discarded.
18709 Generally there is no guarantee that a memory transaction ever succeeds
18710 and suitable fallback code always needs to be supplied.
18712 @deftypefn {RTM Function} {unsigned} _xbegin ()
18713 Start a RTM (Restricted Transactional Memory) transaction.
18714 Returns @code{_XBEGIN_STARTED} when the transaction
18715 started successfully (note this is not 0, so the constant has to be
18716 explicitly tested).
18718 If the transaction aborts, all side-effects
18719 are undone and an abort code encoded as a bit mask is returned.
18720 The following macros are defined:
18723 @item _XABORT_EXPLICIT
18724 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18725 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18726 @item _XABORT_RETRY
18727 Transaction retry is possible.
18728 @item _XABORT_CONFLICT
18729 Transaction abort due to a memory conflict with another thread.
18730 @item _XABORT_CAPACITY
18731 Transaction abort due to the transaction using too much memory.
18732 @item _XABORT_DEBUG
18733 Transaction abort due to a debug trap.
18734 @item _XABORT_NESTED
18735 Transaction abort in an inner nested transaction.
18738 There is no guarantee
18739 any transaction ever succeeds, so there always needs to be a valid
18743 @deftypefn {RTM Function} {void} _xend ()
18744 Commit the current transaction. When no transaction is active this faults.
18745 All memory side-effects of the transaction become visible
18746 to other threads in an atomic manner.
18749 @deftypefn {RTM Function} {int} _xtest ()
18750 Return a nonzero value if a transaction is currently active, otherwise 0.
18753 @deftypefn {RTM Function} {void} _xabort (status)
18754 Abort the current transaction. When no transaction is active this is a no-op.
18755 The @var{status} is an 8-bit constant; its value is encoded in the return
18756 value from @code{_xbegin}.
18759 Here is an example showing handling for @code{_XABORT_RETRY}
18760 and a fallback path for other failures:
18763 #include <immintrin.h>
18765 int n_tries, max_tries;
18766 unsigned status = _XABORT_EXPLICIT;
18769 for (n_tries = 0; n_tries < max_tries; n_tries++)
18771 status = _xbegin ();
18772 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18775 if (status == _XBEGIN_STARTED)
18777 ... transaction code...
18782 ... non-transactional fallback path...
18787 Note that, in most cases, the transactional and non-transactional code
18788 must synchronize together to ensure consistency.
18790 @node Target Format Checks
18791 @section Format Checks Specific to Particular Target Machines
18793 For some target machines, GCC supports additional options to the
18795 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18798 * Solaris Format Checks::
18799 * Darwin Format Checks::
18802 @node Solaris Format Checks
18803 @subsection Solaris Format Checks
18805 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18806 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18807 conversions, and the two-argument @code{%b} conversion for displaying
18808 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18810 @node Darwin Format Checks
18811 @subsection Darwin Format Checks
18813 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18814 attribute context. Declarations made with such attribution are parsed for correct syntax
18815 and format argument types. However, parsing of the format string itself is currently undefined
18816 and is not carried out by this version of the compiler.
18818 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18819 also be used as format arguments. Note that the relevant headers are only likely to be
18820 available on Darwin (OSX) installations. On such installations, the XCode and system
18821 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18822 associated functions.
18825 @section Pragmas Accepted by GCC
18827 @cindex @code{#pragma}
18829 GCC supports several types of pragmas, primarily in order to compile
18830 code originally written for other compilers. Note that in general
18831 we do not recommend the use of pragmas; @xref{Function Attributes},
18832 for further explanation.
18835 * AArch64 Pragmas::
18839 * RS/6000 and PowerPC Pragmas::
18842 * Solaris Pragmas::
18843 * Symbol-Renaming Pragmas::
18844 * Structure-Layout Pragmas::
18846 * Diagnostic Pragmas::
18847 * Visibility Pragmas::
18848 * Push/Pop Macro Pragmas::
18849 * Function Specific Option Pragmas::
18850 * Loop-Specific Pragmas::
18853 @node AArch64 Pragmas
18854 @subsection AArch64 Pragmas
18856 The pragmas defined by the AArch64 target correspond to the AArch64
18857 target function attributes. They can be specified as below:
18859 #pragma GCC target("string")
18862 where @code{@var{string}} can be any string accepted as an AArch64 target
18863 attribute. @xref{AArch64 Function Attributes}, for more details
18864 on the permissible values of @code{string}.
18867 @subsection ARM Pragmas
18869 The ARM target defines pragmas for controlling the default addition of
18870 @code{long_call} and @code{short_call} attributes to functions.
18871 @xref{Function Attributes}, for information about the effects of these
18876 @cindex pragma, long_calls
18877 Set all subsequent functions to have the @code{long_call} attribute.
18879 @item no_long_calls
18880 @cindex pragma, no_long_calls
18881 Set all subsequent functions to have the @code{short_call} attribute.
18883 @item long_calls_off
18884 @cindex pragma, long_calls_off
18885 Do not affect the @code{long_call} or @code{short_call} attributes of
18886 subsequent functions.
18890 @subsection M32C Pragmas
18893 @item GCC memregs @var{number}
18894 @cindex pragma, memregs
18895 Overrides the command-line option @code{-memregs=} for the current
18896 file. Use with care! This pragma must be before any function in the
18897 file, and mixing different memregs values in different objects may
18898 make them incompatible. This pragma is useful when a
18899 performance-critical function uses a memreg for temporary values,
18900 as it may allow you to reduce the number of memregs used.
18902 @item ADDRESS @var{name} @var{address}
18903 @cindex pragma, address
18904 For any declared symbols matching @var{name}, this does three things
18905 to that symbol: it forces the symbol to be located at the given
18906 address (a number), it forces the symbol to be volatile, and it
18907 changes the symbol's scope to be static. This pragma exists for
18908 compatibility with other compilers, but note that the common
18909 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18913 #pragma ADDRESS port3 0x103
18920 @subsection MeP Pragmas
18924 @item custom io_volatile (on|off)
18925 @cindex pragma, custom io_volatile
18926 Overrides the command-line option @code{-mio-volatile} for the current
18927 file. Note that for compatibility with future GCC releases, this
18928 option should only be used once before any @code{io} variables in each
18931 @item GCC coprocessor available @var{registers}
18932 @cindex pragma, coprocessor available
18933 Specifies which coprocessor registers are available to the register
18934 allocator. @var{registers} may be a single register, register range
18935 separated by ellipses, or comma-separated list of those. Example:
18938 #pragma GCC coprocessor available $c0...$c10, $c28
18941 @item GCC coprocessor call_saved @var{registers}
18942 @cindex pragma, coprocessor call_saved
18943 Specifies which coprocessor registers are to be saved and restored by
18944 any function using them. @var{registers} may be a single register,
18945 register range separated by ellipses, or comma-separated list of
18949 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18952 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18953 @cindex pragma, coprocessor subclass
18954 Creates and defines a register class. These register classes can be
18955 used by inline @code{asm} constructs. @var{registers} may be a single
18956 register, register range separated by ellipses, or comma-separated
18957 list of those. Example:
18960 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18962 asm ("cpfoo %0" : "=B" (x));
18965 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18966 @cindex pragma, disinterrupt
18967 For the named functions, the compiler adds code to disable interrupts
18968 for the duration of those functions. If any functions so named
18969 are not encountered in the source, a warning is emitted that the pragma is
18970 not used. Examples:
18973 #pragma disinterrupt foo
18974 #pragma disinterrupt bar, grill
18975 int foo () @{ @dots{} @}
18978 @item GCC call @var{name} , @var{name} @dots{}
18979 @cindex pragma, call
18980 For the named functions, the compiler always uses a register-indirect
18981 call model when calling the named functions. Examples:
18990 @node RS/6000 and PowerPC Pragmas
18991 @subsection RS/6000 and PowerPC Pragmas
18993 The RS/6000 and PowerPC targets define one pragma for controlling
18994 whether or not the @code{longcall} attribute is added to function
18995 declarations by default. This pragma overrides the @option{-mlongcall}
18996 option, but not the @code{longcall} and @code{shortcall} attributes.
18997 @xref{RS/6000 and PowerPC Options}, for more information about when long
18998 calls are and are not necessary.
19002 @cindex pragma, longcall
19003 Apply the @code{longcall} attribute to all subsequent function
19007 Do not apply the @code{longcall} attribute to subsequent function
19011 @c Describe h8300 pragmas here.
19012 @c Describe sh pragmas here.
19013 @c Describe v850 pragmas here.
19015 @node S/390 Pragmas
19016 @subsection S/390 Pragmas
19018 The pragmas defined by the S/390 target correspond to the S/390
19019 target function attributes and some the additional options:
19026 Note that options of the pragma, unlike options of the target
19027 attribute, do change the value of preprocessor macros like
19028 @code{__VEC__}. They can be specified as below:
19031 #pragma GCC target("string[,string]...")
19032 #pragma GCC target("string"[,"string"]...)
19035 @node Darwin Pragmas
19036 @subsection Darwin Pragmas
19038 The following pragmas are available for all architectures running the
19039 Darwin operating system. These are useful for compatibility with other
19043 @item mark @var{tokens}@dots{}
19044 @cindex pragma, mark
19045 This pragma is accepted, but has no effect.
19047 @item options align=@var{alignment}
19048 @cindex pragma, options align
19049 This pragma sets the alignment of fields in structures. The values of
19050 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
19051 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
19052 properly; to restore the previous setting, use @code{reset} for the
19055 @item segment @var{tokens}@dots{}
19056 @cindex pragma, segment
19057 This pragma is accepted, but has no effect.
19059 @item unused (@var{var} [, @var{var}]@dots{})
19060 @cindex pragma, unused
19061 This pragma declares variables to be possibly unused. GCC does not
19062 produce warnings for the listed variables. The effect is similar to
19063 that of the @code{unused} attribute, except that this pragma may appear
19064 anywhere within the variables' scopes.
19067 @node Solaris Pragmas
19068 @subsection Solaris Pragmas
19070 The Solaris target supports @code{#pragma redefine_extname}
19071 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
19072 @code{#pragma} directives for compatibility with the system compiler.
19075 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
19076 @cindex pragma, align
19078 Increase the minimum alignment of each @var{variable} to @var{alignment}.
19079 This is the same as GCC's @code{aligned} attribute @pxref{Variable
19080 Attributes}). Macro expansion occurs on the arguments to this pragma
19081 when compiling C and Objective-C@. It does not currently occur when
19082 compiling C++, but this is a bug which may be fixed in a future
19085 @item fini (@var{function} [, @var{function}]...)
19086 @cindex pragma, fini
19088 This pragma causes each listed @var{function} to be called after
19089 main, or during shared module unloading, by adding a call to the
19090 @code{.fini} section.
19092 @item init (@var{function} [, @var{function}]...)
19093 @cindex pragma, init
19095 This pragma causes each listed @var{function} to be called during
19096 initialization (before @code{main}) or during shared module loading, by
19097 adding a call to the @code{.init} section.
19101 @node Symbol-Renaming Pragmas
19102 @subsection Symbol-Renaming Pragmas
19104 GCC supports a @code{#pragma} directive that changes the name used in
19105 assembly for a given declaration. While this pragma is supported on all
19106 platforms, it is intended primarily to provide compatibility with the
19107 Solaris system headers. This effect can also be achieved using the asm
19108 labels extension (@pxref{Asm Labels}).
19111 @item redefine_extname @var{oldname} @var{newname}
19112 @cindex pragma, redefine_extname
19114 This pragma gives the C function @var{oldname} the assembly symbol
19115 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
19116 is defined if this pragma is available (currently on all platforms).
19119 This pragma and the asm labels extension interact in a complicated
19120 manner. Here are some corner cases you may want to be aware of:
19123 @item This pragma silently applies only to declarations with external
19124 linkage. Asm labels do not have this restriction.
19126 @item In C++, this pragma silently applies only to declarations with
19127 ``C'' linkage. Again, asm labels do not have this restriction.
19129 @item If either of the ways of changing the assembly name of a
19130 declaration are applied to a declaration whose assembly name has
19131 already been determined (either by a previous use of one of these
19132 features, or because the compiler needed the assembly name in order to
19133 generate code), and the new name is different, a warning issues and
19134 the name does not change.
19136 @item The @var{oldname} used by @code{#pragma redefine_extname} is
19137 always the C-language name.
19140 @node Structure-Layout Pragmas
19141 @subsection Structure-Layout Pragmas
19143 For compatibility with Microsoft Windows compilers, GCC supports a
19144 set of @code{#pragma} directives that change the maximum alignment of
19145 members of structures (other than zero-width bit-fields), unions, and
19146 classes subsequently defined. The @var{n} value below always is required
19147 to be a small power of two and specifies the new alignment in bytes.
19150 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
19151 @item @code{#pragma pack()} sets the alignment to the one that was in
19152 effect when compilation started (see also command-line option
19153 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
19154 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
19155 setting on an internal stack and then optionally sets the new alignment.
19156 @item @code{#pragma pack(pop)} restores the alignment setting to the one
19157 saved at the top of the internal stack (and removes that stack entry).
19158 Note that @code{#pragma pack([@var{n}])} does not influence this internal
19159 stack; thus it is possible to have @code{#pragma pack(push)} followed by
19160 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
19161 @code{#pragma pack(pop)}.
19164 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
19165 directive which lays out structures and unions subsequently defined as the
19166 documented @code{__attribute__ ((ms_struct))}.
19169 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
19170 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
19171 @item @code{#pragma ms_struct reset} goes back to the default layout.
19174 Most targets also support the @code{#pragma scalar_storage_order} directive
19175 which lays out structures and unions subsequently defined as the documented
19176 @code{__attribute__ ((scalar_storage_order))}.
19179 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
19180 of the scalar fields to big-endian.
19181 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
19182 of the scalar fields to little-endian.
19183 @item @code{#pragma scalar_storage_order default} goes back to the endianness
19184 that was in effect when compilation started (see also command-line option
19185 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
19189 @subsection Weak Pragmas
19191 For compatibility with SVR4, GCC supports a set of @code{#pragma}
19192 directives for declaring symbols to be weak, and defining weak
19196 @item #pragma weak @var{symbol}
19197 @cindex pragma, weak
19198 This pragma declares @var{symbol} to be weak, as if the declaration
19199 had the attribute of the same name. The pragma may appear before
19200 or after the declaration of @var{symbol}. It is not an error for
19201 @var{symbol} to never be defined at all.
19203 @item #pragma weak @var{symbol1} = @var{symbol2}
19204 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
19205 It is an error if @var{symbol2} is not defined in the current
19209 @node Diagnostic Pragmas
19210 @subsection Diagnostic Pragmas
19212 GCC allows the user to selectively enable or disable certain types of
19213 diagnostics, and change the kind of the diagnostic. For example, a
19214 project's policy might require that all sources compile with
19215 @option{-Werror} but certain files might have exceptions allowing
19216 specific types of warnings. Or, a project might selectively enable
19217 diagnostics and treat them as errors depending on which preprocessor
19218 macros are defined.
19221 @item #pragma GCC diagnostic @var{kind} @var{option}
19222 @cindex pragma, diagnostic
19224 Modifies the disposition of a diagnostic. Note that not all
19225 diagnostics are modifiable; at the moment only warnings (normally
19226 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
19227 Use @option{-fdiagnostics-show-option} to determine which diagnostics
19228 are controllable and which option controls them.
19230 @var{kind} is @samp{error} to treat this diagnostic as an error,
19231 @samp{warning} to treat it like a warning (even if @option{-Werror} is
19232 in effect), or @samp{ignored} if the diagnostic is to be ignored.
19233 @var{option} is a double quoted string that matches the command-line
19237 #pragma GCC diagnostic warning "-Wformat"
19238 #pragma GCC diagnostic error "-Wformat"
19239 #pragma GCC diagnostic ignored "-Wformat"
19242 Note that these pragmas override any command-line options. GCC keeps
19243 track of the location of each pragma, and issues diagnostics according
19244 to the state as of that point in the source file. Thus, pragmas occurring
19245 after a line do not affect diagnostics caused by that line.
19247 @item #pragma GCC diagnostic push
19248 @itemx #pragma GCC diagnostic pop
19250 Causes GCC to remember the state of the diagnostics as of each
19251 @code{push}, and restore to that point at each @code{pop}. If a
19252 @code{pop} has no matching @code{push}, the command-line options are
19256 #pragma GCC diagnostic error "-Wuninitialized"
19257 foo(a); /* error is given for this one */
19258 #pragma GCC diagnostic push
19259 #pragma GCC diagnostic ignored "-Wuninitialized"
19260 foo(b); /* no diagnostic for this one */
19261 #pragma GCC diagnostic pop
19262 foo(c); /* error is given for this one */
19263 #pragma GCC diagnostic pop
19264 foo(d); /* depends on command-line options */
19269 GCC also offers a simple mechanism for printing messages during
19273 @item #pragma message @var{string}
19274 @cindex pragma, diagnostic
19276 Prints @var{string} as a compiler message on compilation. The message
19277 is informational only, and is neither a compilation warning nor an error.
19280 #pragma message "Compiling " __FILE__ "..."
19283 @var{string} may be parenthesized, and is printed with location
19284 information. For example,
19287 #define DO_PRAGMA(x) _Pragma (#x)
19288 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
19290 TODO(Remember to fix this)
19294 prints @samp{/tmp/file.c:4: note: #pragma message:
19295 TODO - Remember to fix this}.
19299 @node Visibility Pragmas
19300 @subsection Visibility Pragmas
19303 @item #pragma GCC visibility push(@var{visibility})
19304 @itemx #pragma GCC visibility pop
19305 @cindex pragma, visibility
19307 This pragma allows the user to set the visibility for multiple
19308 declarations without having to give each a visibility attribute
19309 (@pxref{Function Attributes}).
19311 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
19312 declarations. Class members and template specializations are not
19313 affected; if you want to override the visibility for a particular
19314 member or instantiation, you must use an attribute.
19319 @node Push/Pop Macro Pragmas
19320 @subsection Push/Pop Macro Pragmas
19322 For compatibility with Microsoft Windows compilers, GCC supports
19323 @samp{#pragma push_macro(@var{"macro_name"})}
19324 and @samp{#pragma pop_macro(@var{"macro_name"})}.
19327 @item #pragma push_macro(@var{"macro_name"})
19328 @cindex pragma, push_macro
19329 This pragma saves the value of the macro named as @var{macro_name} to
19330 the top of the stack for this macro.
19332 @item #pragma pop_macro(@var{"macro_name"})
19333 @cindex pragma, pop_macro
19334 This pragma sets the value of the macro named as @var{macro_name} to
19335 the value on top of the stack for this macro. If the stack for
19336 @var{macro_name} is empty, the value of the macro remains unchanged.
19343 #pragma push_macro("X")
19346 #pragma pop_macro("X")
19351 In this example, the definition of X as 1 is saved by @code{#pragma
19352 push_macro} and restored by @code{#pragma pop_macro}.
19354 @node Function Specific Option Pragmas
19355 @subsection Function Specific Option Pragmas
19358 @item #pragma GCC target (@var{"string"}...)
19359 @cindex pragma GCC target
19361 This pragma allows you to set target specific options for functions
19362 defined later in the source file. One or more strings can be
19363 specified. Each function that is defined after this point is as
19364 if @code{attribute((target("STRING")))} was specified for that
19365 function. The parenthesis around the options is optional.
19366 @xref{Function Attributes}, for more information about the
19367 @code{target} attribute and the attribute syntax.
19369 The @code{#pragma GCC target} pragma is presently implemented for
19370 x86, PowerPC, and Nios II targets only.
19374 @item #pragma GCC optimize (@var{"string"}...)
19375 @cindex pragma GCC optimize
19377 This pragma allows you to set global optimization options for functions
19378 defined later in the source file. One or more strings can be
19379 specified. Each function that is defined after this point is as
19380 if @code{attribute((optimize("STRING")))} was specified for that
19381 function. The parenthesis around the options is optional.
19382 @xref{Function Attributes}, for more information about the
19383 @code{optimize} attribute and the attribute syntax.
19387 @item #pragma GCC push_options
19388 @itemx #pragma GCC pop_options
19389 @cindex pragma GCC push_options
19390 @cindex pragma GCC pop_options
19392 These pragmas maintain a stack of the current target and optimization
19393 options. It is intended for include files where you temporarily want
19394 to switch to using a different @samp{#pragma GCC target} or
19395 @samp{#pragma GCC optimize} and then to pop back to the previous
19400 @item #pragma GCC reset_options
19401 @cindex pragma GCC reset_options
19403 This pragma clears the current @code{#pragma GCC target} and
19404 @code{#pragma GCC optimize} to use the default switches as specified
19405 on the command line.
19408 @node Loop-Specific Pragmas
19409 @subsection Loop-Specific Pragmas
19412 @item #pragma GCC ivdep
19413 @cindex pragma GCC ivdep
19416 With this pragma, the programmer asserts that there are no loop-carried
19417 dependencies which would prevent consecutive iterations of
19418 the following loop from executing concurrently with SIMD
19419 (single instruction multiple data) instructions.
19421 For example, the compiler can only unconditionally vectorize the following
19422 loop with the pragma:
19425 void foo (int n, int *a, int *b, int *c)
19429 for (i = 0; i < n; ++i)
19430 a[i] = b[i] + c[i];
19435 In this example, using the @code{restrict} qualifier had the same
19436 effect. In the following example, that would not be possible. Assume
19437 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
19438 that it can unconditionally vectorize the following loop:
19441 void ignore_vec_dep (int *a, int k, int c, int m)
19444 for (int i = 0; i < m; i++)
19445 a[i] = a[i + k] * c;
19450 @node Unnamed Fields
19451 @section Unnamed Structure and Union Fields
19452 @cindex @code{struct}
19453 @cindex @code{union}
19455 As permitted by ISO C11 and for compatibility with other compilers,
19456 GCC allows you to define
19457 a structure or union that contains, as fields, structures and unions
19458 without names. For example:
19472 In this example, you are able to access members of the unnamed
19473 union with code like @samp{foo.b}. Note that only unnamed structs and
19474 unions are allowed, you may not have, for example, an unnamed
19477 You must never create such structures that cause ambiguous field definitions.
19478 For example, in this structure:
19490 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
19491 The compiler gives errors for such constructs.
19493 @opindex fms-extensions
19494 Unless @option{-fms-extensions} is used, the unnamed field must be a
19495 structure or union definition without a tag (for example, @samp{struct
19496 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
19497 also be a definition with a tag such as @samp{struct foo @{ int a;
19498 @};}, a reference to a previously defined structure or union such as
19499 @samp{struct foo;}, or a reference to a @code{typedef} name for a
19500 previously defined structure or union type.
19502 @opindex fplan9-extensions
19503 The option @option{-fplan9-extensions} enables
19504 @option{-fms-extensions} as well as two other extensions. First, a
19505 pointer to a structure is automatically converted to a pointer to an
19506 anonymous field for assignments and function calls. For example:
19509 struct s1 @{ int a; @};
19510 struct s2 @{ struct s1; @};
19511 extern void f1 (struct s1 *);
19512 void f2 (struct s2 *p) @{ f1 (p); @}
19516 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
19517 converted into a pointer to the anonymous field.
19519 Second, when the type of an anonymous field is a @code{typedef} for a
19520 @code{struct} or @code{union}, code may refer to the field using the
19521 name of the @code{typedef}.
19524 typedef struct @{ int a; @} s1;
19525 struct s2 @{ s1; @};
19526 s1 f1 (struct s2 *p) @{ return p->s1; @}
19529 These usages are only permitted when they are not ambiguous.
19532 @section Thread-Local Storage
19533 @cindex Thread-Local Storage
19534 @cindex @acronym{TLS}
19535 @cindex @code{__thread}
19537 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
19538 are allocated such that there is one instance of the variable per extant
19539 thread. The runtime model GCC uses to implement this originates
19540 in the IA-64 processor-specific ABI, but has since been migrated
19541 to other processors as well. It requires significant support from
19542 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
19543 system libraries (@file{libc.so} and @file{libpthread.so}), so it
19544 is not available everywhere.
19546 At the user level, the extension is visible with a new storage
19547 class keyword: @code{__thread}. For example:
19551 extern __thread struct state s;
19552 static __thread char *p;
19555 The @code{__thread} specifier may be used alone, with the @code{extern}
19556 or @code{static} specifiers, but with no other storage class specifier.
19557 When used with @code{extern} or @code{static}, @code{__thread} must appear
19558 immediately after the other storage class specifier.
19560 The @code{__thread} specifier may be applied to any global, file-scoped
19561 static, function-scoped static, or static data member of a class. It may
19562 not be applied to block-scoped automatic or non-static data member.
19564 When the address-of operator is applied to a thread-local variable, it is
19565 evaluated at run time and returns the address of the current thread's
19566 instance of that variable. An address so obtained may be used by any
19567 thread. When a thread terminates, any pointers to thread-local variables
19568 in that thread become invalid.
19570 No static initialization may refer to the address of a thread-local variable.
19572 In C++, if an initializer is present for a thread-local variable, it must
19573 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19576 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19577 ELF Handling For Thread-Local Storage} for a detailed explanation of
19578 the four thread-local storage addressing models, and how the runtime
19579 is expected to function.
19582 * C99 Thread-Local Edits::
19583 * C++98 Thread-Local Edits::
19586 @node C99 Thread-Local Edits
19587 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19589 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19590 that document the exact semantics of the language extension.
19594 @cite{5.1.2 Execution environments}
19596 Add new text after paragraph 1
19599 Within either execution environment, a @dfn{thread} is a flow of
19600 control within a program. It is implementation defined whether
19601 or not there may be more than one thread associated with a program.
19602 It is implementation defined how threads beyond the first are
19603 created, the name and type of the function called at thread
19604 startup, and how threads may be terminated. However, objects
19605 with thread storage duration shall be initialized before thread
19610 @cite{6.2.4 Storage durations of objects}
19612 Add new text before paragraph 3
19615 An object whose identifier is declared with the storage-class
19616 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19617 Its lifetime is the entire execution of the thread, and its
19618 stored value is initialized only once, prior to thread startup.
19622 @cite{6.4.1 Keywords}
19624 Add @code{__thread}.
19627 @cite{6.7.1 Storage-class specifiers}
19629 Add @code{__thread} to the list of storage class specifiers in
19632 Change paragraph 2 to
19635 With the exception of @code{__thread}, at most one storage-class
19636 specifier may be given [@dots{}]. The @code{__thread} specifier may
19637 be used alone, or immediately following @code{extern} or
19641 Add new text after paragraph 6
19644 The declaration of an identifier for a variable that has
19645 block scope that specifies @code{__thread} shall also
19646 specify either @code{extern} or @code{static}.
19648 The @code{__thread} specifier shall be used only with
19653 @node C++98 Thread-Local Edits
19654 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19656 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19657 that document the exact semantics of the language extension.
19661 @b{[intro.execution]}
19663 New text after paragraph 4
19666 A @dfn{thread} is a flow of control within the abstract machine.
19667 It is implementation defined whether or not there may be more than
19671 New text after paragraph 7
19674 It is unspecified whether additional action must be taken to
19675 ensure when and whether side effects are visible to other threads.
19681 Add @code{__thread}.
19684 @b{[basic.start.main]}
19686 Add after paragraph 5
19689 The thread that begins execution at the @code{main} function is called
19690 the @dfn{main thread}. It is implementation defined how functions
19691 beginning threads other than the main thread are designated or typed.
19692 A function so designated, as well as the @code{main} function, is called
19693 a @dfn{thread startup function}. It is implementation defined what
19694 happens if a thread startup function returns. It is implementation
19695 defined what happens to other threads when any thread calls @code{exit}.
19699 @b{[basic.start.init]}
19701 Add after paragraph 4
19704 The storage for an object of thread storage duration shall be
19705 statically initialized before the first statement of the thread startup
19706 function. An object of thread storage duration shall not require
19707 dynamic initialization.
19711 @b{[basic.start.term]}
19713 Add after paragraph 3
19716 The type of an object with thread storage duration shall not have a
19717 non-trivial destructor, nor shall it be an array type whose elements
19718 (directly or indirectly) have non-trivial destructors.
19724 Add ``thread storage duration'' to the list in paragraph 1.
19729 Thread, static, and automatic storage durations are associated with
19730 objects introduced by declarations [@dots{}].
19733 Add @code{__thread} to the list of specifiers in paragraph 3.
19736 @b{[basic.stc.thread]}
19738 New section before @b{[basic.stc.static]}
19741 The keyword @code{__thread} applied to a non-local object gives the
19742 object thread storage duration.
19744 A local variable or class data member declared both @code{static}
19745 and @code{__thread} gives the variable or member thread storage
19750 @b{[basic.stc.static]}
19755 All objects that have neither thread storage duration, dynamic
19756 storage duration nor are local [@dots{}].
19762 Add @code{__thread} to the list in paragraph 1.
19767 With the exception of @code{__thread}, at most one
19768 @var{storage-class-specifier} shall appear in a given
19769 @var{decl-specifier-seq}. The @code{__thread} specifier may
19770 be used alone, or immediately following the @code{extern} or
19771 @code{static} specifiers. [@dots{}]
19774 Add after paragraph 5
19777 The @code{__thread} specifier can be applied only to the names of objects
19778 and to anonymous unions.
19784 Add after paragraph 6
19787 Non-@code{static} members shall not be @code{__thread}.
19791 @node Binary constants
19792 @section Binary Constants using the @samp{0b} Prefix
19793 @cindex Binary constants using the @samp{0b} prefix
19795 Integer constants can be written as binary constants, consisting of a
19796 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19797 @samp{0B}. This is particularly useful in environments that operate a
19798 lot on the bit level (like microcontrollers).
19800 The following statements are identical:
19809 The type of these constants follows the same rules as for octal or
19810 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19813 @node C++ Extensions
19814 @chapter Extensions to the C++ Language
19815 @cindex extensions, C++ language
19816 @cindex C++ language extensions
19818 The GNU compiler provides these extensions to the C++ language (and you
19819 can also use most of the C language extensions in your C++ programs). If you
19820 want to write code that checks whether these features are available, you can
19821 test for the GNU compiler the same way as for C programs: check for a
19822 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19823 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19824 Predefined Macros,cpp,The GNU C Preprocessor}).
19827 * C++ Volatiles:: What constitutes an access to a volatile object.
19828 * Restricted Pointers:: C99 restricted pointers and references.
19829 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19830 * C++ Interface:: You can use a single C++ header file for both
19831 declarations and definitions.
19832 * Template Instantiation:: Methods for ensuring that exactly one copy of
19833 each needed template instantiation is emitted.
19834 * Bound member functions:: You can extract a function pointer to the
19835 method denoted by a @samp{->*} or @samp{.*} expression.
19836 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19837 * Function Multiversioning:: Declaring multiple function versions.
19838 * Namespace Association:: Strong using-directives for namespace association.
19839 * Type Traits:: Compiler support for type traits.
19840 * C++ Concepts:: Improved support for generic programming.
19841 * Java Exceptions:: Tweaking exception handling to work with Java.
19842 * Deprecated Features:: Things will disappear from G++.
19843 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19846 @node C++ Volatiles
19847 @section When is a Volatile C++ Object Accessed?
19848 @cindex accessing volatiles
19849 @cindex volatile read
19850 @cindex volatile write
19851 @cindex volatile access
19853 The C++ standard differs from the C standard in its treatment of
19854 volatile objects. It fails to specify what constitutes a volatile
19855 access, except to say that C++ should behave in a similar manner to C
19856 with respect to volatiles, where possible. However, the different
19857 lvalueness of expressions between C and C++ complicate the behavior.
19858 G++ behaves the same as GCC for volatile access, @xref{C
19859 Extensions,,Volatiles}, for a description of GCC's behavior.
19861 The C and C++ language specifications differ when an object is
19862 accessed in a void context:
19865 volatile int *src = @var{somevalue};
19869 The C++ standard specifies that such expressions do not undergo lvalue
19870 to rvalue conversion, and that the type of the dereferenced object may
19871 be incomplete. The C++ standard does not specify explicitly that it
19872 is lvalue to rvalue conversion that is responsible for causing an
19873 access. There is reason to believe that it is, because otherwise
19874 certain simple expressions become undefined. However, because it
19875 would surprise most programmers, G++ treats dereferencing a pointer to
19876 volatile object of complete type as GCC would do for an equivalent
19877 type in C@. When the object has incomplete type, G++ issues a
19878 warning; if you wish to force an error, you must force a conversion to
19879 rvalue with, for instance, a static cast.
19881 When using a reference to volatile, G++ does not treat equivalent
19882 expressions as accesses to volatiles, but instead issues a warning that
19883 no volatile is accessed. The rationale for this is that otherwise it
19884 becomes difficult to determine where volatile access occur, and not
19885 possible to ignore the return value from functions returning volatile
19886 references. Again, if you wish to force a read, cast the reference to
19889 G++ implements the same behavior as GCC does when assigning to a
19890 volatile object---there is no reread of the assigned-to object, the
19891 assigned rvalue is reused. Note that in C++ assignment expressions
19892 are lvalues, and if used as an lvalue, the volatile object is
19893 referred to. For instance, @var{vref} refers to @var{vobj}, as
19894 expected, in the following example:
19898 volatile int &vref = vobj = @var{something};
19901 @node Restricted Pointers
19902 @section Restricting Pointer Aliasing
19903 @cindex restricted pointers
19904 @cindex restricted references
19905 @cindex restricted this pointer
19907 As with the C front end, G++ understands the C99 feature of restricted pointers,
19908 specified with the @code{__restrict__}, or @code{__restrict} type
19909 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19910 language flag, @code{restrict} is not a keyword in C++.
19912 In addition to allowing restricted pointers, you can specify restricted
19913 references, which indicate that the reference is not aliased in the local
19917 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19924 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19925 @var{rref} refers to a (different) unaliased integer.
19927 You may also specify whether a member function's @var{this} pointer is
19928 unaliased by using @code{__restrict__} as a member function qualifier.
19931 void T::fn () __restrict__
19938 Within the body of @code{T::fn}, @var{this} has the effective
19939 definition @code{T *__restrict__ const this}. Notice that the
19940 interpretation of a @code{__restrict__} member function qualifier is
19941 different to that of @code{const} or @code{volatile} qualifier, in that it
19942 is applied to the pointer rather than the object. This is consistent with
19943 other compilers that implement restricted pointers.
19945 As with all outermost parameter qualifiers, @code{__restrict__} is
19946 ignored in function definition matching. This means you only need to
19947 specify @code{__restrict__} in a function definition, rather than
19948 in a function prototype as well.
19950 @node Vague Linkage
19951 @section Vague Linkage
19952 @cindex vague linkage
19954 There are several constructs in C++ that require space in the object
19955 file but are not clearly tied to a single translation unit. We say that
19956 these constructs have ``vague linkage''. Typically such constructs are
19957 emitted wherever they are needed, though sometimes we can be more
19961 @item Inline Functions
19962 Inline functions are typically defined in a header file which can be
19963 included in many different compilations. Hopefully they can usually be
19964 inlined, but sometimes an out-of-line copy is necessary, if the address
19965 of the function is taken or if inlining fails. In general, we emit an
19966 out-of-line copy in all translation units where one is needed. As an
19967 exception, we only emit inline virtual functions with the vtable, since
19968 it always requires a copy.
19970 Local static variables and string constants used in an inline function
19971 are also considered to have vague linkage, since they must be shared
19972 between all inlined and out-of-line instances of the function.
19976 C++ virtual functions are implemented in most compilers using a lookup
19977 table, known as a vtable. The vtable contains pointers to the virtual
19978 functions provided by a class, and each object of the class contains a
19979 pointer to its vtable (or vtables, in some multiple-inheritance
19980 situations). If the class declares any non-inline, non-pure virtual
19981 functions, the first one is chosen as the ``key method'' for the class,
19982 and the vtable is only emitted in the translation unit where the key
19985 @emph{Note:} If the chosen key method is later defined as inline, the
19986 vtable is still emitted in every translation unit that defines it.
19987 Make sure that any inline virtuals are declared inline in the class
19988 body, even if they are not defined there.
19990 @item @code{type_info} objects
19991 @cindex @code{type_info}
19993 C++ requires information about types to be written out in order to
19994 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19995 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19996 object is written out along with the vtable so that @samp{dynamic_cast}
19997 can determine the dynamic type of a class object at run time. For all
19998 other types, we write out the @samp{type_info} object when it is used: when
19999 applying @samp{typeid} to an expression, throwing an object, or
20000 referring to a type in a catch clause or exception specification.
20002 @item Template Instantiations
20003 Most everything in this section also applies to template instantiations,
20004 but there are other options as well.
20005 @xref{Template Instantiation,,Where's the Template?}.
20009 When used with GNU ld version 2.8 or later on an ELF system such as
20010 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
20011 these constructs will be discarded at link time. This is known as
20014 On targets that don't support COMDAT, but do support weak symbols, GCC
20015 uses them. This way one copy overrides all the others, but
20016 the unused copies still take up space in the executable.
20018 For targets that do not support either COMDAT or weak symbols,
20019 most entities with vague linkage are emitted as local symbols to
20020 avoid duplicate definition errors from the linker. This does not happen
20021 for local statics in inlines, however, as having multiple copies
20022 almost certainly breaks things.
20024 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
20025 another way to control placement of these constructs.
20027 @node C++ Interface
20028 @section C++ Interface and Implementation Pragmas
20030 @cindex interface and implementation headers, C++
20031 @cindex C++ interface and implementation headers
20032 @cindex pragmas, interface and implementation
20034 @code{#pragma interface} and @code{#pragma implementation} provide the
20035 user with a way of explicitly directing the compiler to emit entities
20036 with vague linkage (and debugging information) in a particular
20039 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
20040 by COMDAT support and the ``key method'' heuristic
20041 mentioned in @ref{Vague Linkage}. Using them can actually cause your
20042 program to grow due to unnecessary out-of-line copies of inline
20046 @item #pragma interface
20047 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
20048 @kindex #pragma interface
20049 Use this directive in @emph{header files} that define object classes, to save
20050 space in most of the object files that use those classes. Normally,
20051 local copies of certain information (backup copies of inline member
20052 functions, debugging information, and the internal tables that implement
20053 virtual functions) must be kept in each object file that includes class
20054 definitions. You can use this pragma to avoid such duplication. When a
20055 header file containing @samp{#pragma interface} is included in a
20056 compilation, this auxiliary information is not generated (unless
20057 the main input source file itself uses @samp{#pragma implementation}).
20058 Instead, the object files contain references to be resolved at link
20061 The second form of this directive is useful for the case where you have
20062 multiple headers with the same name in different directories. If you
20063 use this form, you must specify the same string to @samp{#pragma
20066 @item #pragma implementation
20067 @itemx #pragma implementation "@var{objects}.h"
20068 @kindex #pragma implementation
20069 Use this pragma in a @emph{main input file}, when you want full output from
20070 included header files to be generated (and made globally visible). The
20071 included header file, in turn, should use @samp{#pragma interface}.
20072 Backup copies of inline member functions, debugging information, and the
20073 internal tables used to implement virtual functions are all generated in
20074 implementation files.
20076 @cindex implied @code{#pragma implementation}
20077 @cindex @code{#pragma implementation}, implied
20078 @cindex naming convention, implementation headers
20079 If you use @samp{#pragma implementation} with no argument, it applies to
20080 an include file with the same basename@footnote{A file's @dfn{basename}
20081 is the name stripped of all leading path information and of trailing
20082 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
20083 file. For example, in @file{allclass.cc}, giving just
20084 @samp{#pragma implementation}
20085 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
20087 Use the string argument if you want a single implementation file to
20088 include code from multiple header files. (You must also use
20089 @samp{#include} to include the header file; @samp{#pragma
20090 implementation} only specifies how to use the file---it doesn't actually
20093 There is no way to split up the contents of a single header file into
20094 multiple implementation files.
20097 @cindex inlining and C++ pragmas
20098 @cindex C++ pragmas, effect on inlining
20099 @cindex pragmas in C++, effect on inlining
20100 @samp{#pragma implementation} and @samp{#pragma interface} also have an
20101 effect on function inlining.
20103 If you define a class in a header file marked with @samp{#pragma
20104 interface}, the effect on an inline function defined in that class is
20105 similar to an explicit @code{extern} declaration---the compiler emits
20106 no code at all to define an independent version of the function. Its
20107 definition is used only for inlining with its callers.
20109 @opindex fno-implement-inlines
20110 Conversely, when you include the same header file in a main source file
20111 that declares it as @samp{#pragma implementation}, the compiler emits
20112 code for the function itself; this defines a version of the function
20113 that can be found via pointers (or by callers compiled without
20114 inlining). If all calls to the function can be inlined, you can avoid
20115 emitting the function by compiling with @option{-fno-implement-inlines}.
20116 If any calls are not inlined, you will get linker errors.
20118 @node Template Instantiation
20119 @section Where's the Template?
20120 @cindex template instantiation
20122 C++ templates were the first language feature to require more
20123 intelligence from the environment than was traditionally found on a UNIX
20124 system. Somehow the compiler and linker have to make sure that each
20125 template instance occurs exactly once in the executable if it is needed,
20126 and not at all otherwise. There are two basic approaches to this
20127 problem, which are referred to as the Borland model and the Cfront model.
20130 @item Borland model
20131 Borland C++ solved the template instantiation problem by adding the code
20132 equivalent of common blocks to their linker; the compiler emits template
20133 instances in each translation unit that uses them, and the linker
20134 collapses them together. The advantage of this model is that the linker
20135 only has to consider the object files themselves; there is no external
20136 complexity to worry about. The disadvantage is that compilation time
20137 is increased because the template code is being compiled repeatedly.
20138 Code written for this model tends to include definitions of all
20139 templates in the header file, since they must be seen to be
20143 The AT&T C++ translator, Cfront, solved the template instantiation
20144 problem by creating the notion of a template repository, an
20145 automatically maintained place where template instances are stored. A
20146 more modern version of the repository works as follows: As individual
20147 object files are built, the compiler places any template definitions and
20148 instantiations encountered in the repository. At link time, the link
20149 wrapper adds in the objects in the repository and compiles any needed
20150 instances that were not previously emitted. The advantages of this
20151 model are more optimal compilation speed and the ability to use the
20152 system linker; to implement the Borland model a compiler vendor also
20153 needs to replace the linker. The disadvantages are vastly increased
20154 complexity, and thus potential for error; for some code this can be
20155 just as transparent, but in practice it can been very difficult to build
20156 multiple programs in one directory and one program in multiple
20157 directories. Code written for this model tends to separate definitions
20158 of non-inline member templates into a separate file, which should be
20159 compiled separately.
20162 G++ implements the Borland model on targets where the linker supports it,
20163 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
20164 Otherwise G++ implements neither automatic model.
20166 You have the following options for dealing with template instantiations:
20170 Do nothing. Code written for the Borland model works fine, but
20171 each translation unit contains instances of each of the templates it
20172 uses. The duplicate instances will be discarded by the linker, but in
20173 a large program, this can lead to an unacceptable amount of code
20174 duplication in object files or shared libraries.
20176 Duplicate instances of a template can be avoided by defining an explicit
20177 instantiation in one object file, and preventing the compiler from doing
20178 implicit instantiations in any other object files by using an explicit
20179 instantiation declaration, using the @code{extern template} syntax:
20182 extern template int max (int, int);
20185 This syntax is defined in the C++ 2011 standard, but has been supported by
20186 G++ and other compilers since well before 2011.
20188 Explicit instantiations can be used for the largest or most frequently
20189 duplicated instances, without having to know exactly which other instances
20190 are used in the rest of the program. You can scatter the explicit
20191 instantiations throughout your program, perhaps putting them in the
20192 translation units where the instances are used or the translation units
20193 that define the templates themselves; you can put all of the explicit
20194 instantiations you need into one big file; or you can create small files
20201 template class Foo<int>;
20202 template ostream& operator <<
20203 (ostream&, const Foo<int>&);
20207 for each of the instances you need, and create a template instantiation
20208 library from those.
20210 This is the simplest option, but also offers flexibility and
20211 fine-grained control when necessary. It is also the most portable
20212 alternative and programs using this approach will work with most modern
20217 Compile your template-using code with @option{-frepo}. The compiler
20218 generates files with the extension @samp{.rpo} listing all of the
20219 template instantiations used in the corresponding object files that
20220 could be instantiated there; the link wrapper, @samp{collect2},
20221 then updates the @samp{.rpo} files to tell the compiler where to place
20222 those instantiations and rebuild any affected object files. The
20223 link-time overhead is negligible after the first pass, as the compiler
20224 continues to place the instantiations in the same files.
20226 This can be a suitable option for application code written for the Borland
20227 model, as it usually just works. Code written for the Cfront model
20228 needs to be modified so that the template definitions are available at
20229 one or more points of instantiation; usually this is as simple as adding
20230 @code{#include <tmethods.cc>} to the end of each template header.
20232 For library code, if you want the library to provide all of the template
20233 instantiations it needs, just try to link all of its object files
20234 together; the link will fail, but cause the instantiations to be
20235 generated as a side effect. Be warned, however, that this may cause
20236 conflicts if multiple libraries try to provide the same instantiations.
20237 For greater control, use explicit instantiation as described in the next
20241 @opindex fno-implicit-templates
20242 Compile your code with @option{-fno-implicit-templates} to disable the
20243 implicit generation of template instances, and explicitly instantiate
20244 all the ones you use. This approach requires more knowledge of exactly
20245 which instances you need than do the others, but it's less
20246 mysterious and allows greater control if you want to ensure that only
20247 the intended instances are used.
20249 If you are using Cfront-model code, you can probably get away with not
20250 using @option{-fno-implicit-templates} when compiling files that don't
20251 @samp{#include} the member template definitions.
20253 If you use one big file to do the instantiations, you may want to
20254 compile it without @option{-fno-implicit-templates} so you get all of the
20255 instances required by your explicit instantiations (but not by any
20256 other files) without having to specify them as well.
20258 In addition to forward declaration of explicit instantiations
20259 (with @code{extern}), G++ has extended the template instantiation
20260 syntax to support instantiation of the compiler support data for a
20261 template class (i.e.@: the vtable) without instantiating any of its
20262 members (with @code{inline}), and instantiation of only the static data
20263 members of a template class, without the support data or member
20264 functions (with @code{static}):
20267 inline template class Foo<int>;
20268 static template class Foo<int>;
20272 @node Bound member functions
20273 @section Extracting the Function Pointer from a Bound Pointer to Member Function
20275 @cindex pointer to member function
20276 @cindex bound pointer to member function
20278 In C++, pointer to member functions (PMFs) are implemented using a wide
20279 pointer of sorts to handle all the possible call mechanisms; the PMF
20280 needs to store information about how to adjust the @samp{this} pointer,
20281 and if the function pointed to is virtual, where to find the vtable, and
20282 where in the vtable to look for the member function. If you are using
20283 PMFs in an inner loop, you should really reconsider that decision. If
20284 that is not an option, you can extract the pointer to the function that
20285 would be called for a given object/PMF pair and call it directly inside
20286 the inner loop, to save a bit of time.
20288 Note that you still pay the penalty for the call through a
20289 function pointer; on most modern architectures, such a call defeats the
20290 branch prediction features of the CPU@. This is also true of normal
20291 virtual function calls.
20293 The syntax for this extension is
20297 extern int (A::*fp)();
20298 typedef int (*fptr)(A *);
20300 fptr p = (fptr)(a.*fp);
20303 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
20304 no object is needed to obtain the address of the function. They can be
20305 converted to function pointers directly:
20308 fptr p1 = (fptr)(&A::foo);
20311 @opindex Wno-pmf-conversions
20312 You must specify @option{-Wno-pmf-conversions} to use this extension.
20314 @node C++ Attributes
20315 @section C++-Specific Variable, Function, and Type Attributes
20317 Some attributes only make sense for C++ programs.
20320 @item abi_tag ("@var{tag}", ...)
20321 @cindex @code{abi_tag} function attribute
20322 @cindex @code{abi_tag} variable attribute
20323 @cindex @code{abi_tag} type attribute
20324 The @code{abi_tag} attribute can be applied to a function, variable, or class
20325 declaration. It modifies the mangled name of the entity to
20326 incorporate the tag name, in order to distinguish the function or
20327 class from an earlier version with a different ABI; perhaps the class
20328 has changed size, or the function has a different return type that is
20329 not encoded in the mangled name.
20331 The attribute can also be applied to an inline namespace, but does not
20332 affect the mangled name of the namespace; in this case it is only used
20333 for @option{-Wabi-tag} warnings and automatic tagging of functions and
20334 variables. Tagging inline namespaces is generally preferable to
20335 tagging individual declarations, but the latter is sometimes
20336 necessary, such as when only certain members of a class need to be
20339 The argument can be a list of strings of arbitrary length. The
20340 strings are sorted on output, so the order of the list is
20343 A redeclaration of an entity must not add new ABI tags,
20344 since doing so would change the mangled name.
20346 The ABI tags apply to a name, so all instantiations and
20347 specializations of a template have the same tags. The attribute will
20348 be ignored if applied to an explicit specialization or instantiation.
20350 The @option{-Wabi-tag} flag enables a warning about a class which does
20351 not have all the ABI tags used by its subobjects and virtual functions; for users with code
20352 that needs to coexist with an earlier ABI, using this option can help
20353 to find all affected types that need to be tagged.
20355 When a type involving an ABI tag is used as the type of a variable or
20356 return type of a function where that tag is not already present in the
20357 signature of the function, the tag is automatically applied to the
20358 variable or function. @option{-Wabi-tag} also warns about this
20359 situation; this warning can be avoided by explicitly tagging the
20360 variable or function or moving it into a tagged inline namespace.
20362 @item init_priority (@var{priority})
20363 @cindex @code{init_priority} variable attribute
20365 In Standard C++, objects defined at namespace scope are guaranteed to be
20366 initialized in an order in strict accordance with that of their definitions
20367 @emph{in a given translation unit}. No guarantee is made for initializations
20368 across translation units. However, GNU C++ allows users to control the
20369 order of initialization of objects defined at namespace scope with the
20370 @code{init_priority} attribute by specifying a relative @var{priority},
20371 a constant integral expression currently bounded between 101 and 65535
20372 inclusive. Lower numbers indicate a higher priority.
20374 In the following example, @code{A} would normally be created before
20375 @code{B}, but the @code{init_priority} attribute reverses that order:
20378 Some_Class A __attribute__ ((init_priority (2000)));
20379 Some_Class B __attribute__ ((init_priority (543)));
20383 Note that the particular values of @var{priority} do not matter; only their
20386 @item java_interface
20387 @cindex @code{java_interface} type attribute
20389 This type attribute informs C++ that the class is a Java interface. It may
20390 only be applied to classes declared within an @code{extern "Java"} block.
20391 Calls to methods declared in this interface are dispatched using GCJ's
20392 interface table mechanism, instead of regular virtual table dispatch.
20395 @cindex @code{warn_unused} type attribute
20397 For C++ types with non-trivial constructors and/or destructors it is
20398 impossible for the compiler to determine whether a variable of this
20399 type is truly unused if it is not referenced. This type attribute
20400 informs the compiler that variables of this type should be warned
20401 about if they appear to be unused, just like variables of fundamental
20404 This attribute is appropriate for types which just represent a value,
20405 such as @code{std::string}; it is not appropriate for types which
20406 control a resource, such as @code{std::lock_guard}.
20408 This attribute is also accepted in C, but it is unnecessary because C
20409 does not have constructors or destructors.
20413 See also @ref{Namespace Association}.
20415 @node Function Multiversioning
20416 @section Function Multiversioning
20417 @cindex function versions
20419 With the GNU C++ front end, for x86 targets, you may specify multiple
20420 versions of a function, where each function is specialized for a
20421 specific target feature. At runtime, the appropriate version of the
20422 function is automatically executed depending on the characteristics of
20423 the execution platform. Here is an example.
20426 __attribute__ ((target ("default")))
20429 // The default version of foo.
20433 __attribute__ ((target ("sse4.2")))
20436 // foo version for SSE4.2
20440 __attribute__ ((target ("arch=atom")))
20443 // foo version for the Intel ATOM processor
20447 __attribute__ ((target ("arch=amdfam10")))
20450 // foo version for the AMD Family 0x10 processors.
20457 assert ((*p) () == foo ());
20462 In the above example, four versions of function foo are created. The
20463 first version of foo with the target attribute "default" is the default
20464 version. This version gets executed when no other target specific
20465 version qualifies for execution on a particular platform. A new version
20466 of foo is created by using the same function signature but with a
20467 different target string. Function foo is called or a pointer to it is
20468 taken just like a regular function. GCC takes care of doing the
20469 dispatching to call the right version at runtime. Refer to the
20470 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
20471 Function Multiversioning} for more details.
20473 @node Namespace Association
20474 @section Namespace Association
20476 @strong{Caution:} The semantics of this extension are equivalent
20477 to C++ 2011 inline namespaces. Users should use inline namespaces
20478 instead as this extension will be removed in future versions of G++.
20480 A using-directive with @code{__attribute ((strong))} is stronger
20481 than a normal using-directive in two ways:
20485 Templates from the used namespace can be specialized and explicitly
20486 instantiated as though they were members of the using namespace.
20489 The using namespace is considered an associated namespace of all
20490 templates in the used namespace for purposes of argument-dependent
20494 The used namespace must be nested within the using namespace so that
20495 normal unqualified lookup works properly.
20497 This is useful for composing a namespace transparently from
20498 implementation namespaces. For example:
20503 template <class T> struct A @{ @};
20505 using namespace debug __attribute ((__strong__));
20506 template <> struct A<int> @{ @}; // @r{OK to specialize}
20508 template <class T> void f (A<T>);
20513 f (std::A<float>()); // @r{lookup finds} std::f
20519 @section Type Traits
20521 The C++ front end implements syntactic extensions that allow
20522 compile-time determination of
20523 various characteristics of a type (or of a
20527 @item __has_nothrow_assign (type)
20528 If @code{type} is const qualified or is a reference type then the trait is
20529 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
20530 is true, else if @code{type} is a cv class or union type with copy assignment
20531 operators that are known not to throw an exception then the trait is true,
20532 else it is false. Requires: @code{type} shall be a complete type,
20533 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20535 @item __has_nothrow_copy (type)
20536 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
20537 @code{type} is a cv class or union type with copy constructors that
20538 are known not to throw an exception then the trait is true, else it is false.
20539 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
20540 @code{void}, or an array of unknown bound.
20542 @item __has_nothrow_constructor (type)
20543 If @code{__has_trivial_constructor (type)} is true then the trait is
20544 true, else if @code{type} is a cv class or union type (or array
20545 thereof) with a default constructor that is known not to throw an
20546 exception then the trait is true, else it is false. Requires:
20547 @code{type} shall be a complete type, (possibly cv-qualified)
20548 @code{void}, or an array of unknown bound.
20550 @item __has_trivial_assign (type)
20551 If @code{type} is const qualified or is a reference type then the trait is
20552 false. Otherwise if @code{__is_pod (type)} is true then the trait is
20553 true, else if @code{type} is a cv class or union type with a trivial
20554 copy assignment ([class.copy]) then the trait is true, else it is
20555 false. Requires: @code{type} shall be a complete type, (possibly
20556 cv-qualified) @code{void}, or an array of unknown bound.
20558 @item __has_trivial_copy (type)
20559 If @code{__is_pod (type)} is true or @code{type} is a reference type
20560 then the trait is true, else if @code{type} is a cv class or union type
20561 with a trivial copy constructor ([class.copy]) then the trait
20562 is true, else it is false. Requires: @code{type} shall be a complete
20563 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20565 @item __has_trivial_constructor (type)
20566 If @code{__is_pod (type)} is true then the trait is true, else if
20567 @code{type} is a cv class or union type (or array thereof) with a
20568 trivial default constructor ([class.ctor]) then the trait is true,
20569 else it is false. Requires: @code{type} shall be a complete
20570 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20572 @item __has_trivial_destructor (type)
20573 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20574 the trait is true, else if @code{type} is a cv class or union type (or
20575 array thereof) with a trivial destructor ([class.dtor]) then the trait
20576 is true, else it is false. Requires: @code{type} shall be a complete
20577 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20579 @item __has_virtual_destructor (type)
20580 If @code{type} is a class type with a virtual destructor
20581 ([class.dtor]) then the trait is true, else it is false. Requires:
20582 @code{type} shall be a complete type, (possibly cv-qualified)
20583 @code{void}, or an array of unknown bound.
20585 @item __is_abstract (type)
20586 If @code{type} is an abstract class ([class.abstract]) then the trait
20587 is true, else it is false. Requires: @code{type} shall be a complete
20588 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20590 @item __is_base_of (base_type, derived_type)
20591 If @code{base_type} is a base class of @code{derived_type}
20592 ([class.derived]) then the trait is true, otherwise it is false.
20593 Top-level cv qualifications of @code{base_type} and
20594 @code{derived_type} are ignored. For the purposes of this trait, a
20595 class type is considered is own base. Requires: if @code{__is_class
20596 (base_type)} and @code{__is_class (derived_type)} are true and
20597 @code{base_type} and @code{derived_type} are not the same type
20598 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20599 type. A diagnostic is produced if this requirement is not met.
20601 @item __is_class (type)
20602 If @code{type} is a cv class type, and not a union type
20603 ([basic.compound]) the trait is true, else it is false.
20605 @item __is_empty (type)
20606 If @code{__is_class (type)} is false then the trait is false.
20607 Otherwise @code{type} is considered empty if and only if: @code{type}
20608 has no non-static data members, or all non-static data members, if
20609 any, are bit-fields of length 0, and @code{type} has no virtual
20610 members, and @code{type} has no virtual base classes, and @code{type}
20611 has no base classes @code{base_type} for which
20612 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20613 be a complete type, (possibly cv-qualified) @code{void}, or an array
20616 @item __is_enum (type)
20617 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20618 true, else it is false.
20620 @item __is_literal_type (type)
20621 If @code{type} is a literal type ([basic.types]) the trait is
20622 true, else it is false. Requires: @code{type} shall be a complete type,
20623 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20625 @item __is_pod (type)
20626 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20627 else it is false. Requires: @code{type} shall be a complete type,
20628 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20630 @item __is_polymorphic (type)
20631 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20632 is true, else it is false. Requires: @code{type} shall be a complete
20633 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20635 @item __is_standard_layout (type)
20636 If @code{type} is a standard-layout type ([basic.types]) the trait is
20637 true, else it is false. Requires: @code{type} shall be a complete
20638 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20640 @item __is_trivial (type)
20641 If @code{type} is a trivial type ([basic.types]) the trait is
20642 true, else it is false. Requires: @code{type} shall be a complete
20643 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20645 @item __is_union (type)
20646 If @code{type} is a cv union type ([basic.compound]) the trait is
20647 true, else it is false.
20649 @item __underlying_type (type)
20650 The underlying type of @code{type}. Requires: @code{type} shall be
20651 an enumeration type ([dcl.enum]).
20657 @section C++ Concepts
20659 C++ concepts provide much-improved support for generic programming. In
20660 particular, they allow the specification of constraints on template arguments.
20661 The constraints are used to extend the usual overloading and partial
20662 specialization capabilities of the language, allowing generic data structures
20663 and algorithms to be ``refined'' based on their properties rather than their
20666 The following keywords are reserved for concepts.
20670 States an expression as an assumption, and if possible, verifies that the
20671 assumption is valid. For example, @code{assume(n > 0)}.
20674 Introduces an axiom definition. Axioms introduce requirements on values.
20677 Introduces a universally quantified object in an axiom. For example,
20678 @code{forall (int n) n + 0 == n}).
20681 Introduces a concept definition. Concepts are sets of syntactic and semantic
20682 requirements on types and their values.
20685 Introduces constraints on template arguments or requirements for a member
20686 function of a class template.
20690 The front end also exposes a number of internal mechanism that can be used
20691 to simplify the writing of type traits. Note that some of these traits are
20692 likely to be removed in the future.
20695 @item __is_same (type1, type2)
20696 A binary type trait: true whenever the type arguments are the same.
20701 @node Java Exceptions
20702 @section Java Exceptions
20704 The Java language uses a slightly different exception handling model
20705 from C++. Normally, GNU C++ automatically detects when you are
20706 writing C++ code that uses Java exceptions, and handle them
20707 appropriately. However, if C++ code only needs to execute destructors
20708 when Java exceptions are thrown through it, GCC guesses incorrectly.
20709 Sample problematic code is:
20712 struct S @{ ~S(); @};
20713 extern void bar(); // @r{is written in Java, and may throw exceptions}
20722 The usual effect of an incorrect guess is a link failure, complaining of
20723 a missing routine called @samp{__gxx_personality_v0}.
20725 You can inform the compiler that Java exceptions are to be used in a
20726 translation unit, irrespective of what it might think, by writing
20727 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20728 @samp{#pragma} must appear before any functions that throw or catch
20729 exceptions, or run destructors when exceptions are thrown through them.
20731 You cannot mix Java and C++ exceptions in the same translation unit. It
20732 is believed to be safe to throw a C++ exception from one file through
20733 another file compiled for the Java exception model, or vice versa, but
20734 there may be bugs in this area.
20736 @node Deprecated Features
20737 @section Deprecated Features
20739 In the past, the GNU C++ compiler was extended to experiment with new
20740 features, at a time when the C++ language was still evolving. Now that
20741 the C++ standard is complete, some of those features are superseded by
20742 superior alternatives. Using the old features might cause a warning in
20743 some cases that the feature will be dropped in the future. In other
20744 cases, the feature might be gone already.
20746 While the list below is not exhaustive, it documents some of the options
20747 that are now deprecated:
20750 @item -fexternal-templates
20751 @itemx -falt-external-templates
20752 These are two of the many ways for G++ to implement template
20753 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20754 defines how template definitions have to be organized across
20755 implementation units. G++ has an implicit instantiation mechanism that
20756 should work just fine for standard-conforming code.
20758 @item -fstrict-prototype
20759 @itemx -fno-strict-prototype
20760 Previously it was possible to use an empty prototype parameter list to
20761 indicate an unspecified number of parameters (like C), rather than no
20762 parameters, as C++ demands. This feature has been removed, except where
20763 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20766 G++ allows a virtual function returning @samp{void *} to be overridden
20767 by one returning a different pointer type. This extension to the
20768 covariant return type rules is now deprecated and will be removed from a
20771 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20772 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20773 and are now removed from G++. Code using these operators should be
20774 modified to use @code{std::min} and @code{std::max} instead.
20776 The named return value extension has been deprecated, and is now
20779 The use of initializer lists with new expressions has been deprecated,
20780 and is now removed from G++.
20782 Floating and complex non-type template parameters have been deprecated,
20783 and are now removed from G++.
20785 The implicit typename extension has been deprecated and is now
20788 The use of default arguments in function pointers, function typedefs
20789 and other places where they are not permitted by the standard is
20790 deprecated and will be removed from a future version of G++.
20792 G++ allows floating-point literals to appear in integral constant expressions,
20793 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20794 This extension is deprecated and will be removed from a future version.
20796 G++ allows static data members of const floating-point type to be declared
20797 with an initializer in a class definition. The standard only allows
20798 initializers for static members of const integral types and const
20799 enumeration types so this extension has been deprecated and will be removed
20800 from a future version.
20802 @node Backwards Compatibility
20803 @section Backwards Compatibility
20804 @cindex Backwards Compatibility
20805 @cindex ARM [Annotated C++ Reference Manual]
20807 Now that there is a definitive ISO standard C++, G++ has a specification
20808 to adhere to. The C++ language evolved over time, and features that
20809 used to be acceptable in previous drafts of the standard, such as the ARM
20810 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20811 compilation of C++ written to such drafts, G++ contains some backwards
20812 compatibilities. @emph{All such backwards compatibility features are
20813 liable to disappear in future versions of G++.} They should be considered
20814 deprecated. @xref{Deprecated Features}.
20818 If a variable is declared at for scope, it used to remain in scope until
20819 the end of the scope that contained the for statement (rather than just
20820 within the for scope). G++ retains this, but issues a warning, if such a
20821 variable is accessed outside the for scope.
20823 @item Implicit C language
20824 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20825 scope to set the language. On such systems, all header files are
20826 implicitly scoped inside a C language scope. Also, an empty prototype
20827 @code{()} is treated as an unspecified number of arguments, rather
20828 than no arguments, as C++ demands.
20831 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20832 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr