1 @c Copyright (C) 1988-2015 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 2
920 debug info format can represent this, so use of DWARF 2 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 2 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 2 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.
1478 @cindex @code{__seg_tls} x86 named address space
1479 Some operating systems define either the @code{%fs} or @code{%gs}
1480 segment as the thread-local storage base for each thread. Objects
1481 within this address space are accessed with the appropriate
1482 segment override prefix.
1484 The pointer located at address 0 within the segment contains the
1485 offset of the segment within the generic address space. Thus this
1486 address space is considered a subspace of the generic address space,
1487 and the known segment offset is applied when converting addresses
1488 to and from the generic address space.
1490 The preprocessor symbol @code{__SEG_TLS} is defined when this
1491 address space is supported.
1496 @section Arrays of Length Zero
1497 @cindex arrays of length zero
1498 @cindex zero-length arrays
1499 @cindex length-zero arrays
1500 @cindex flexible array members
1502 Zero-length arrays are allowed in GNU C@. They are very useful as the
1503 last element of a structure that is really a header for a variable-length
1512 struct line *thisline = (struct line *)
1513 malloc (sizeof (struct line) + this_length);
1514 thisline->length = this_length;
1517 In ISO C90, you would have to give @code{contents} a length of 1, which
1518 means either you waste space or complicate the argument to @code{malloc}.
1520 In ISO C99, you would use a @dfn{flexible array member}, which is
1521 slightly different in syntax and semantics:
1525 Flexible array members are written as @code{contents[]} without
1529 Flexible array members have incomplete type, and so the @code{sizeof}
1530 operator may not be applied. As a quirk of the original implementation
1531 of zero-length arrays, @code{sizeof} evaluates to zero.
1534 Flexible array members may only appear as the last member of a
1535 @code{struct} that is otherwise non-empty.
1538 A structure containing a flexible array member, or a union containing
1539 such a structure (possibly recursively), may not be a member of a
1540 structure or an element of an array. (However, these uses are
1541 permitted by GCC as extensions.)
1544 Non-empty initialization of zero-length
1545 arrays is treated like any case where there are more initializer
1546 elements than the array holds, in that a suitable warning about ``excess
1547 elements in array'' is given, and the excess elements (all of them, in
1548 this case) are ignored.
1550 GCC allows static initialization of flexible array members.
1551 This is equivalent to defining a new structure containing the original
1552 structure followed by an array of sufficient size to contain the data.
1553 E.g.@: in the following, @code{f1} is constructed as if it were declared
1559 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1562 struct f1 f1; int data[3];
1563 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1567 The convenience of this extension is that @code{f1} has the desired
1568 type, eliminating the need to consistently refer to @code{f2.f1}.
1570 This has symmetry with normal static arrays, in that an array of
1571 unknown size is also written with @code{[]}.
1573 Of course, this extension only makes sense if the extra data comes at
1574 the end of a top-level object, as otherwise we would be overwriting
1575 data at subsequent offsets. To avoid undue complication and confusion
1576 with initialization of deeply nested arrays, we simply disallow any
1577 non-empty initialization except when the structure is the top-level
1578 object. For example:
1581 struct foo @{ int x; int y[]; @};
1582 struct bar @{ struct foo z; @};
1584 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1585 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1586 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1587 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1590 @node Empty Structures
1591 @section Structures with No Members
1592 @cindex empty structures
1593 @cindex zero-size structures
1595 GCC permits a C structure to have no members:
1602 The structure has size zero. In C++, empty structures are part
1603 of the language. G++ treats empty structures as if they had a single
1604 member of type @code{char}.
1606 @node Variable Length
1607 @section Arrays of Variable Length
1608 @cindex variable-length arrays
1609 @cindex arrays of variable length
1612 Variable-length automatic arrays are allowed in ISO C99, and as an
1613 extension GCC accepts them in C90 mode and in C++. These arrays are
1614 declared like any other automatic arrays, but with a length that is not
1615 a constant expression. The storage is allocated at the point of
1616 declaration and deallocated when the block scope containing the declaration
1622 concat_fopen (char *s1, char *s2, char *mode)
1624 char str[strlen (s1) + strlen (s2) + 1];
1627 return fopen (str, mode);
1631 @cindex scope of a variable length array
1632 @cindex variable-length array scope
1633 @cindex deallocating variable length arrays
1634 Jumping or breaking out of the scope of the array name deallocates the
1635 storage. Jumping into the scope is not allowed; you get an error
1638 @cindex variable-length array in a structure
1639 As an extension, GCC accepts variable-length arrays as a member of
1640 a structure or a union. For example:
1646 struct S @{ int x[n]; @};
1650 @cindex @code{alloca} vs variable-length arrays
1651 You can use the function @code{alloca} to get an effect much like
1652 variable-length arrays. The function @code{alloca} is available in
1653 many other C implementations (but not in all). On the other hand,
1654 variable-length arrays are more elegant.
1656 There are other differences between these two methods. Space allocated
1657 with @code{alloca} exists until the containing @emph{function} returns.
1658 The space for a variable-length array is deallocated as soon as the array
1659 name's scope ends, unless you also use @code{alloca} in this scope.
1661 You can also use variable-length arrays as arguments to functions:
1665 tester (int len, char data[len][len])
1671 The length of an array is computed once when the storage is allocated
1672 and is remembered for the scope of the array in case you access it with
1675 If you want to pass the array first and the length afterward, you can
1676 use a forward declaration in the parameter list---another GNU extension.
1680 tester (int len; char data[len][len], int len)
1686 @cindex parameter forward declaration
1687 The @samp{int len} before the semicolon is a @dfn{parameter forward
1688 declaration}, and it serves the purpose of making the name @code{len}
1689 known when the declaration of @code{data} is parsed.
1691 You can write any number of such parameter forward declarations in the
1692 parameter list. They can be separated by commas or semicolons, but the
1693 last one must end with a semicolon, which is followed by the ``real''
1694 parameter declarations. Each forward declaration must match a ``real''
1695 declaration in parameter name and data type. ISO C99 does not support
1696 parameter forward declarations.
1698 @node Variadic Macros
1699 @section Macros with a Variable Number of Arguments.
1700 @cindex variable number of arguments
1701 @cindex macro with variable arguments
1702 @cindex rest argument (in macro)
1703 @cindex variadic macros
1705 In the ISO C standard of 1999, a macro can be declared to accept a
1706 variable number of arguments much as a function can. The syntax for
1707 defining the macro is similar to that of a function. Here is an
1711 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1715 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1716 such a macro, it represents the zero or more tokens until the closing
1717 parenthesis that ends the invocation, including any commas. This set of
1718 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1719 wherever it appears. See the CPP manual for more information.
1721 GCC has long supported variadic macros, and used a different syntax that
1722 allowed you to give a name to the variable arguments just like any other
1723 argument. Here is an example:
1726 #define debug(format, args...) fprintf (stderr, format, args)
1730 This is in all ways equivalent to the ISO C example above, but arguably
1731 more readable and descriptive.
1733 GNU CPP has two further variadic macro extensions, and permits them to
1734 be used with either of the above forms of macro definition.
1736 In standard C, you are not allowed to leave the variable argument out
1737 entirely; but you are allowed to pass an empty argument. For example,
1738 this invocation is invalid in ISO C, because there is no comma after
1745 GNU CPP permits you to completely omit the variable arguments in this
1746 way. In the above examples, the compiler would complain, though since
1747 the expansion of the macro still has the extra comma after the format
1750 To help solve this problem, CPP behaves specially for variable arguments
1751 used with the token paste operator, @samp{##}. If instead you write
1754 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1758 and if the variable arguments are omitted or empty, the @samp{##}
1759 operator causes the preprocessor to remove the comma before it. If you
1760 do provide some variable arguments in your macro invocation, GNU CPP
1761 does not complain about the paste operation and instead places the
1762 variable arguments after the comma. Just like any other pasted macro
1763 argument, these arguments are not macro expanded.
1765 @node Escaped Newlines
1766 @section Slightly Looser Rules for Escaped Newlines
1767 @cindex escaped newlines
1768 @cindex newlines (escaped)
1770 The preprocessor treatment of escaped newlines is more relaxed
1771 than that specified by the C90 standard, which requires the newline
1772 to immediately follow a backslash.
1773 GCC's implementation allows whitespace in the form
1774 of spaces, horizontal and vertical tabs, and form feeds between the
1775 backslash and the subsequent newline. The preprocessor issues a
1776 warning, but treats it as a valid escaped newline and combines the two
1777 lines to form a single logical line. This works within comments and
1778 tokens, as well as between tokens. Comments are @emph{not} treated as
1779 whitespace for the purposes of this relaxation, since they have not
1780 yet been replaced with spaces.
1783 @section Non-Lvalue Arrays May Have Subscripts
1784 @cindex subscripting
1785 @cindex arrays, non-lvalue
1787 @cindex subscripting and function values
1788 In ISO C99, arrays that are not lvalues still decay to pointers, and
1789 may be subscripted, although they may not be modified or used after
1790 the next sequence point and the unary @samp{&} operator may not be
1791 applied to them. As an extension, GNU C allows such arrays to be
1792 subscripted in C90 mode, though otherwise they do not decay to
1793 pointers outside C99 mode. For example,
1794 this is valid in GNU C though not valid in C90:
1798 struct foo @{int a[4];@};
1804 return f().a[index];
1810 @section Arithmetic on @code{void}- and Function-Pointers
1811 @cindex void pointers, arithmetic
1812 @cindex void, size of pointer to
1813 @cindex function pointers, arithmetic
1814 @cindex function, size of pointer to
1816 In GNU C, addition and subtraction operations are supported on pointers to
1817 @code{void} and on pointers to functions. This is done by treating the
1818 size of a @code{void} or of a function as 1.
1820 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1821 and on function types, and returns 1.
1823 @opindex Wpointer-arith
1824 The option @option{-Wpointer-arith} requests a warning if these extensions
1827 @node Pointers to Arrays
1828 @section Pointers to Arrays with Qualifiers Work as Expected
1829 @cindex pointers to arrays
1830 @cindex const qualifier
1832 In GNU C, pointers to arrays with qualifiers work similar to pointers
1833 to other qualified types. For example, a value of type @code{int (*)[5]}
1834 can be used to initialize a variable of type @code{const int (*)[5]}.
1835 These types are incompatible in ISO C because the @code{const} qualifier
1836 is formally attached to the element type of the array and not the
1841 transpose (int N, int M, double out[M][N], const double in[N][M]);
1845 transpose(3, 2, y, x);
1849 @section Non-Constant Initializers
1850 @cindex initializers, non-constant
1851 @cindex non-constant initializers
1853 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1854 automatic variable are not required to be constant expressions in GNU C@.
1855 Here is an example of an initializer with run-time varying elements:
1858 foo (float f, float g)
1860 float beat_freqs[2] = @{ f-g, f+g @};
1865 @node Compound Literals
1866 @section Compound Literals
1867 @cindex constructor expressions
1868 @cindex initializations in expressions
1869 @cindex structures, constructor expression
1870 @cindex expressions, constructor
1871 @cindex compound literals
1872 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1874 ISO C99 supports compound literals. A compound literal looks like
1875 a cast containing an initializer. Its value is an object of the
1876 type specified in the cast, containing the elements specified in
1877 the initializer; it is an lvalue. As an extension, GCC supports
1878 compound literals in C90 mode and in C++, though the semantics are
1879 somewhat different in C++.
1881 Usually, the specified type is a structure. Assume that
1882 @code{struct foo} and @code{structure} are declared as shown:
1885 struct foo @{int a; char b[2];@} structure;
1889 Here is an example of constructing a @code{struct foo} with a compound literal:
1892 structure = ((struct foo) @{x + y, 'a', 0@});
1896 This is equivalent to writing the following:
1900 struct foo temp = @{x + y, 'a', 0@};
1905 You can also construct an array, though this is dangerous in C++, as
1906 explained below. If all the elements of the compound literal are
1907 (made up of) simple constant expressions, suitable for use in
1908 initializers of objects of static storage duration, then the compound
1909 literal can be coerced to a pointer to its first element and used in
1910 such an initializer, as shown here:
1913 char **foo = (char *[]) @{ "x", "y", "z" @};
1916 Compound literals for scalar types and union types are
1917 also allowed, but then the compound literal is equivalent
1920 As a GNU extension, GCC allows initialization of objects with static storage
1921 duration by compound literals (which is not possible in ISO C99, because
1922 the initializer is not a constant).
1923 It is handled as if the object is initialized only with the bracket
1924 enclosed list if the types of the compound literal and the object match.
1925 The initializer list of the compound literal must be constant.
1926 If the object being initialized has array type of unknown size, the size is
1927 determined by compound literal size.
1930 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1931 static int y[] = (int []) @{1, 2, 3@};
1932 static int z[] = (int [3]) @{1@};
1936 The above lines are equivalent to the following:
1938 static struct foo x = @{1, 'a', 'b'@};
1939 static int y[] = @{1, 2, 3@};
1940 static int z[] = @{1, 0, 0@};
1943 In C, a compound literal designates an unnamed object with static or
1944 automatic storage duration. In C++, a compound literal designates a
1945 temporary object, which only lives until the end of its
1946 full-expression. As a result, well-defined C code that takes the
1947 address of a subobject of a compound literal can be undefined in C++,
1948 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1949 For instance, if the array compound literal example above appeared
1950 inside a function, any subsequent use of @samp{foo} in C++ has
1951 undefined behavior because the lifetime of the array ends after the
1952 declaration of @samp{foo}.
1954 As an optimization, the C++ compiler sometimes gives array compound
1955 literals longer lifetimes: when the array either appears outside a
1956 function or has const-qualified type. If @samp{foo} and its
1957 initializer had elements of @samp{char *const} type rather than
1958 @samp{char *}, or if @samp{foo} were a global variable, the array
1959 would have static storage duration. But it is probably safest just to
1960 avoid the use of array compound literals in code compiled as C++.
1962 @node Designated Inits
1963 @section Designated Initializers
1964 @cindex initializers with labeled elements
1965 @cindex labeled elements in initializers
1966 @cindex case labels in initializers
1967 @cindex designated initializers
1969 Standard C90 requires the elements of an initializer to appear in a fixed
1970 order, the same as the order of the elements in the array or structure
1973 In ISO C99 you can give the elements in any order, specifying the array
1974 indices or structure field names they apply to, and GNU C allows this as
1975 an extension in C90 mode as well. This extension is not
1976 implemented in GNU C++.
1978 To specify an array index, write
1979 @samp{[@var{index}] =} before the element value. For example,
1982 int a[6] = @{ [4] = 29, [2] = 15 @};
1989 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1993 The index values must be constant expressions, even if the array being
1994 initialized is automatic.
1996 An alternative syntax for this that has been obsolete since GCC 2.5 but
1997 GCC still accepts is to write @samp{[@var{index}]} before the element
1998 value, with no @samp{=}.
2000 To initialize a range of elements to the same value, write
2001 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2002 extension. For example,
2005 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2009 If the value in it has side-effects, the side-effects happen only once,
2010 not for each initialized field by the range initializer.
2013 Note that the length of the array is the highest value specified
2016 In a structure initializer, specify the name of a field to initialize
2017 with @samp{.@var{fieldname} =} before the element value. For example,
2018 given the following structure,
2021 struct point @{ int x, y; @};
2025 the following initialization
2028 struct point p = @{ .y = yvalue, .x = xvalue @};
2035 struct point p = @{ xvalue, yvalue @};
2038 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2039 @samp{@var{fieldname}:}, as shown here:
2042 struct point p = @{ y: yvalue, x: xvalue @};
2045 Omitted field members are implicitly initialized the same as objects
2046 that have static storage duration.
2049 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2050 @dfn{designator}. You can also use a designator (or the obsolete colon
2051 syntax) when initializing a union, to specify which element of the union
2052 should be used. For example,
2055 union foo @{ int i; double d; @};
2057 union foo f = @{ .d = 4 @};
2061 converts 4 to a @code{double} to store it in the union using
2062 the second element. By contrast, casting 4 to type @code{union foo}
2063 stores it into the union as the integer @code{i}, since it is
2064 an integer. (@xref{Cast to Union}.)
2066 You can combine this technique of naming elements with ordinary C
2067 initialization of successive elements. Each initializer element that
2068 does not have a designator applies to the next consecutive element of the
2069 array or structure. For example,
2072 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2079 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2082 Labeling the elements of an array initializer is especially useful
2083 when the indices are characters or belong to an @code{enum} type.
2088 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2089 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2092 @cindex designator lists
2093 You can also write a series of @samp{.@var{fieldname}} and
2094 @samp{[@var{index}]} designators before an @samp{=} to specify a
2095 nested subobject to initialize; the list is taken relative to the
2096 subobject corresponding to the closest surrounding brace pair. For
2097 example, with the @samp{struct point} declaration above:
2100 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2104 If the same field is initialized multiple times, it has the value from
2105 the last initialization. If any such overridden initialization has
2106 side-effect, it is unspecified whether the side-effect happens or not.
2107 Currently, GCC discards them and issues a warning.
2110 @section Case Ranges
2112 @cindex ranges in case statements
2114 You can specify a range of consecutive values in a single @code{case} label,
2118 case @var{low} ... @var{high}:
2122 This has the same effect as the proper number of individual @code{case}
2123 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2125 This feature is especially useful for ranges of ASCII character codes:
2131 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2132 it may be parsed wrong when you use it with integer values. For example,
2147 @section Cast to a Union Type
2148 @cindex cast to a union
2149 @cindex union, casting to a
2151 A cast to union type is similar to other casts, except that the type
2152 specified is a union type. You can specify the type either with
2153 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2154 a constructor, not a cast, and hence does not yield an lvalue like
2155 normal casts. (@xref{Compound Literals}.)
2157 The types that may be cast to the union type are those of the members
2158 of the union. Thus, given the following union and variables:
2161 union foo @{ int i; double d; @};
2167 both @code{x} and @code{y} can be cast to type @code{union foo}.
2169 Using the cast as the right-hand side of an assignment to a variable of
2170 union type is equivalent to storing in a member of the union:
2175 u = (union foo) x @equiv{} u.i = x
2176 u = (union foo) y @equiv{} u.d = y
2179 You can also use the union cast as a function argument:
2182 void hack (union foo);
2184 hack ((union foo) x);
2187 @node Mixed Declarations
2188 @section Mixed Declarations and Code
2189 @cindex mixed declarations and code
2190 @cindex declarations, mixed with code
2191 @cindex code, mixed with declarations
2193 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2194 within compound statements. As an extension, GNU C also allows this in
2195 C90 mode. For example, you could do:
2204 Each identifier is visible from where it is declared until the end of
2205 the enclosing block.
2207 @node Function Attributes
2208 @section Declaring Attributes of Functions
2209 @cindex function attributes
2210 @cindex declaring attributes of functions
2211 @cindex @code{volatile} applied to function
2212 @cindex @code{const} applied to function
2214 In GNU C, you can use function attributes to declare certain things
2215 about functions called in your program which help the compiler
2216 optimize calls and check your code more carefully. For example, you
2217 can use attributes to declare that a function never returns
2218 (@code{noreturn}), returns a value depending only on its arguments
2219 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2221 You can also use attributes to control memory placement, code
2222 generation options or call/return conventions within the function
2223 being annotated. Many of these attributes are target-specific. For
2224 example, many targets support attributes for defining interrupt
2225 handler functions, which typically must follow special register usage
2226 and return conventions.
2228 Function attributes are introduced by the @code{__attribute__} keyword
2229 on a declaration, followed by an attribute specification inside double
2230 parentheses. You can specify multiple attributes in a declaration by
2231 separating them by commas within the double parentheses or by
2232 immediately following an attribute declaration with another attribute
2233 declaration. @xref{Attribute Syntax}, for the exact rules on
2234 attribute syntax and placement.
2236 GCC also supports attributes on
2237 variable declarations (@pxref{Variable Attributes}),
2238 labels (@pxref{Label Attributes}),
2239 enumerators (@pxref{Enumerator Attributes}),
2240 and types (@pxref{Type Attributes}).
2242 There is some overlap between the purposes of attributes and pragmas
2243 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2244 found convenient to use @code{__attribute__} to achieve a natural
2245 attachment of attributes to their corresponding declarations, whereas
2246 @code{#pragma} is of use for compatibility with other compilers
2247 or constructs that do not naturally form part of the grammar.
2249 In addition to the attributes documented here,
2250 GCC plugins may provide their own attributes.
2253 * Common Function Attributes::
2254 * AArch64 Function Attributes::
2255 * ARC Function Attributes::
2256 * ARM Function Attributes::
2257 * AVR Function Attributes::
2258 * Blackfin Function Attributes::
2259 * CR16 Function Attributes::
2260 * Epiphany Function Attributes::
2261 * H8/300 Function Attributes::
2262 * IA-64 Function Attributes::
2263 * M32C Function Attributes::
2264 * M32R/D Function Attributes::
2265 * m68k Function Attributes::
2266 * MCORE Function Attributes::
2267 * MeP Function Attributes::
2268 * MicroBlaze Function Attributes::
2269 * Microsoft Windows Function Attributes::
2270 * MIPS Function Attributes::
2271 * MSP430 Function Attributes::
2272 * NDS32 Function Attributes::
2273 * Nios II Function Attributes::
2274 * PowerPC Function Attributes::
2275 * RL78 Function Attributes::
2276 * RX Function Attributes::
2277 * S/390 Function Attributes::
2278 * SH Function Attributes::
2279 * SPU Function Attributes::
2280 * Symbian OS Function Attributes::
2281 * Visium Function Attributes::
2282 * x86 Function Attributes::
2283 * Xstormy16 Function Attributes::
2286 @node Common Function Attributes
2287 @subsection Common Function Attributes
2289 The following attributes are supported on most targets.
2292 @c Keep this table alphabetized by attribute name. Treat _ as space.
2294 @item alias ("@var{target}")
2295 @cindex @code{alias} function attribute
2296 The @code{alias} attribute causes the declaration to be emitted as an
2297 alias for another symbol, which must be specified. For instance,
2300 void __f () @{ /* @r{Do something.} */; @}
2301 void f () __attribute__ ((weak, alias ("__f")));
2305 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2306 mangled name for the target must be used. It is an error if @samp{__f}
2307 is not defined in the same translation unit.
2309 This attribute requires assembler and object file support,
2310 and may not be available on all targets.
2312 @item aligned (@var{alignment})
2313 @cindex @code{aligned} function attribute
2314 This attribute specifies a minimum alignment for the function,
2317 You cannot use this attribute to decrease the alignment of a function,
2318 only to increase it. However, when you explicitly specify a function
2319 alignment this overrides the effect of the
2320 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2323 Note that the effectiveness of @code{aligned} attributes may be
2324 limited by inherent limitations in your linker. On many systems, the
2325 linker is only able to arrange for functions to be aligned up to a
2326 certain maximum alignment. (For some linkers, the maximum supported
2327 alignment may be very very small.) See your linker documentation for
2328 further information.
2330 The @code{aligned} attribute can also be used for variables and fields
2331 (@pxref{Variable Attributes}.)
2334 @cindex @code{alloc_align} function attribute
2335 The @code{alloc_align} attribute is used to tell the compiler that the
2336 function return value points to memory, where the returned pointer minimum
2337 alignment is given by one of the functions parameters. GCC uses this
2338 information to improve pointer alignment analysis.
2340 The function parameter denoting the allocated alignment is specified by
2341 one integer argument, whose number is the argument of the attribute.
2342 Argument numbering starts at one.
2347 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2351 declares that @code{my_memalign} returns memory with minimum alignment
2352 given by parameter 1.
2355 @cindex @code{alloc_size} function attribute
2356 The @code{alloc_size} attribute is used to tell the compiler that the
2357 function return value points to memory, where the size is given by
2358 one or two of the functions parameters. GCC uses this
2359 information to improve the correctness of @code{__builtin_object_size}.
2361 The function parameter(s) denoting the allocated size are specified by
2362 one or two integer arguments supplied to the attribute. The allocated size
2363 is either the value of the single function argument specified or the product
2364 of the two function arguments specified. Argument numbering starts at
2370 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2371 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2375 declares that @code{my_calloc} returns memory of the size given by
2376 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2377 of the size given by parameter 2.
2380 @cindex @code{always_inline} function attribute
2381 Generally, functions are not inlined unless optimization is specified.
2382 For functions declared inline, this attribute inlines the function
2383 independent of any restrictions that otherwise apply to inlining.
2384 Failure to inline such a function is diagnosed as an error.
2385 Note that if such a function is called indirectly the compiler may
2386 or may not inline it depending on optimization level and a failure
2387 to inline an indirect call may or may not be diagnosed.
2390 @cindex @code{artificial} function attribute
2391 This attribute is useful for small inline wrappers that if possible
2392 should appear during debugging as a unit. Depending on the debug
2393 info format it either means marking the function as artificial
2394 or using the caller location for all instructions within the inlined
2397 @item assume_aligned
2398 @cindex @code{assume_aligned} function attribute
2399 The @code{assume_aligned} attribute is used to tell the compiler that the
2400 function return value points to memory, where the returned pointer minimum
2401 alignment is given by the first argument.
2402 If the attribute has two arguments, the second argument is misalignment offset.
2407 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2408 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2412 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2413 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2416 @item bnd_instrument
2417 @cindex @code{bnd_instrument} function attribute
2418 The @code{bnd_instrument} attribute on functions is used to inform the
2419 compiler that the function should be instrumented when compiled
2420 with the @option{-fchkp-instrument-marked-only} option.
2423 @cindex @code{bnd_legacy} function attribute
2424 @cindex Pointer Bounds Checker attributes
2425 The @code{bnd_legacy} attribute on functions is used to inform the
2426 compiler that the function should not be instrumented when compiled
2427 with the @option{-fcheck-pointer-bounds} option.
2430 @cindex @code{cold} function attribute
2431 The @code{cold} attribute on functions is used to inform the compiler that
2432 the function is unlikely to be executed. The function is optimized for
2433 size rather than speed and on many targets it is placed into a special
2434 subsection of the text section so all cold functions appear close together,
2435 improving code locality of non-cold parts of program. The paths leading
2436 to calls of cold functions within code are marked as unlikely by the branch
2437 prediction mechanism. It is thus useful to mark functions used to handle
2438 unlikely conditions, such as @code{perror}, as cold to improve optimization
2439 of hot functions that do call marked functions in rare occasions.
2441 When profile feedback is available, via @option{-fprofile-use}, cold functions
2442 are automatically detected and this attribute is ignored.
2445 @cindex @code{const} function attribute
2446 @cindex functions that have no side effects
2447 Many functions do not examine any values except their arguments, and
2448 have no effects except the return value. Basically this is just slightly
2449 more strict class than the @code{pure} attribute below, since function is not
2450 allowed to read global memory.
2452 @cindex pointer arguments
2453 Note that a function that has pointer arguments and examines the data
2454 pointed to must @emph{not} be declared @code{const}. Likewise, a
2455 function that calls a non-@code{const} function usually must not be
2456 @code{const}. It does not make sense for a @code{const} function to
2461 @itemx constructor (@var{priority})
2462 @itemx destructor (@var{priority})
2463 @cindex @code{constructor} function attribute
2464 @cindex @code{destructor} function attribute
2465 The @code{constructor} attribute causes the function to be called
2466 automatically before execution enters @code{main ()}. Similarly, the
2467 @code{destructor} attribute causes the function to be called
2468 automatically after @code{main ()} completes or @code{exit ()} is
2469 called. Functions with these attributes are useful for
2470 initializing data that is used implicitly during the execution of
2473 You may provide an optional integer priority to control the order in
2474 which constructor and destructor functions are run. A constructor
2475 with a smaller priority number runs before a constructor with a larger
2476 priority number; the opposite relationship holds for destructors. So,
2477 if you have a constructor that allocates a resource and a destructor
2478 that deallocates the same resource, both functions typically have the
2479 same priority. The priorities for constructor and destructor
2480 functions are the same as those specified for namespace-scope C++
2481 objects (@pxref{C++ Attributes}).
2483 These attributes are not currently implemented for Objective-C@.
2486 @itemx deprecated (@var{msg})
2487 @cindex @code{deprecated} function attribute
2488 The @code{deprecated} attribute results in a warning if the function
2489 is used anywhere in the source file. This is useful when identifying
2490 functions that are expected to be removed in a future version of a
2491 program. The warning also includes the location of the declaration
2492 of the deprecated function, to enable users to easily find further
2493 information about why the function is deprecated, or what they should
2494 do instead. Note that the warnings only occurs for uses:
2497 int old_fn () __attribute__ ((deprecated));
2499 int (*fn_ptr)() = old_fn;
2503 results in a warning on line 3 but not line 2. The optional @var{msg}
2504 argument, which must be a string, is printed in the warning if
2507 The @code{deprecated} attribute can also be used for variables and
2508 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2510 @item error ("@var{message}")
2511 @itemx warning ("@var{message}")
2512 @cindex @code{error} function attribute
2513 @cindex @code{warning} function attribute
2514 If the @code{error} or @code{warning} attribute
2515 is used on a function declaration and a call to such a function
2516 is not eliminated through dead code elimination or other optimizations,
2517 an error or warning (respectively) that includes @var{message} is diagnosed.
2519 for compile-time checking, especially together with @code{__builtin_constant_p}
2520 and inline functions where checking the inline function arguments is not
2521 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2523 While it is possible to leave the function undefined and thus invoke
2524 a link failure (to define the function with
2525 a message in @code{.gnu.warning*} section),
2526 when using these attributes the problem is diagnosed
2527 earlier and with exact location of the call even in presence of inline
2528 functions or when not emitting debugging information.
2530 @item externally_visible
2531 @cindex @code{externally_visible} function attribute
2532 This attribute, attached to a global variable or function, nullifies
2533 the effect of the @option{-fwhole-program} command-line option, so the
2534 object remains visible outside the current compilation unit.
2536 If @option{-fwhole-program} is used together with @option{-flto} and
2537 @command{gold} is used as the linker plugin,
2538 @code{externally_visible} attributes are automatically added to functions
2539 (not variable yet due to a current @command{gold} issue)
2540 that are accessed outside of LTO objects according to resolution file
2541 produced by @command{gold}.
2542 For other linkers that cannot generate resolution file,
2543 explicit @code{externally_visible} attributes are still necessary.
2546 @cindex @code{flatten} function attribute
2547 Generally, inlining into a function is limited. For a function marked with
2548 this attribute, every call inside this function is inlined, if possible.
2549 Whether the function itself is considered for inlining depends on its size and
2550 the current inlining parameters.
2552 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2553 @cindex @code{format} function attribute
2554 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2556 The @code{format} attribute specifies that a function takes @code{printf},
2557 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2558 should be type-checked against a format string. For example, the
2563 my_printf (void *my_object, const char *my_format, ...)
2564 __attribute__ ((format (printf, 2, 3)));
2568 causes the compiler to check the arguments in calls to @code{my_printf}
2569 for consistency with the @code{printf} style format string argument
2572 The parameter @var{archetype} determines how the format string is
2573 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2574 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2575 @code{strfmon}. (You can also use @code{__printf__},
2576 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2577 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2578 @code{ms_strftime} are also present.
2579 @var{archetype} values such as @code{printf} refer to the formats accepted
2580 by the system's C runtime library,
2581 while values prefixed with @samp{gnu_} always refer
2582 to the formats accepted by the GNU C Library. On Microsoft Windows
2583 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2584 @file{msvcrt.dll} library.
2585 The parameter @var{string-index}
2586 specifies which argument is the format string argument (starting
2587 from 1), while @var{first-to-check} is the number of the first
2588 argument to check against the format string. For functions
2589 where the arguments are not available to be checked (such as
2590 @code{vprintf}), specify the third parameter as zero. In this case the
2591 compiler only checks the format string for consistency. For
2592 @code{strftime} formats, the third parameter is required to be zero.
2593 Since non-static C++ methods have an implicit @code{this} argument, the
2594 arguments of such methods should be counted from two, not one, when
2595 giving values for @var{string-index} and @var{first-to-check}.
2597 In the example above, the format string (@code{my_format}) is the second
2598 argument of the function @code{my_print}, and the arguments to check
2599 start with the third argument, so the correct parameters for the format
2600 attribute are 2 and 3.
2602 @opindex ffreestanding
2603 @opindex fno-builtin
2604 The @code{format} attribute allows you to identify your own functions
2605 that take format strings as arguments, so that GCC can check the
2606 calls to these functions for errors. The compiler always (unless
2607 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2608 for the standard library functions @code{printf}, @code{fprintf},
2609 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2610 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2611 warnings are requested (using @option{-Wformat}), so there is no need to
2612 modify the header file @file{stdio.h}. In C99 mode, the functions
2613 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2614 @code{vsscanf} are also checked. Except in strictly conforming C
2615 standard modes, the X/Open function @code{strfmon} is also checked as
2616 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2617 @xref{C Dialect Options,,Options Controlling C Dialect}.
2619 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2620 recognized in the same context. Declarations including these format attributes
2621 are parsed for correct syntax, however the result of checking of such format
2622 strings is not yet defined, and is not carried out by this version of the
2625 The target may also provide additional types of format checks.
2626 @xref{Target Format Checks,,Format Checks Specific to Particular
2629 @item format_arg (@var{string-index})
2630 @cindex @code{format_arg} function attribute
2631 @opindex Wformat-nonliteral
2632 The @code{format_arg} attribute specifies that a function takes a format
2633 string for a @code{printf}, @code{scanf}, @code{strftime} or
2634 @code{strfmon} style function and modifies it (for example, to translate
2635 it into another language), so the result can be passed to a
2636 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2637 function (with the remaining arguments to the format function the same
2638 as they would have been for the unmodified string). For example, the
2643 my_dgettext (char *my_domain, const char *my_format)
2644 __attribute__ ((format_arg (2)));
2648 causes the compiler to check the arguments in calls to a @code{printf},
2649 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2650 format string argument is a call to the @code{my_dgettext} function, for
2651 consistency with the format string argument @code{my_format}. If the
2652 @code{format_arg} attribute had not been specified, all the compiler
2653 could tell in such calls to format functions would be that the format
2654 string argument is not constant; this would generate a warning when
2655 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2656 without the attribute.
2658 The parameter @var{string-index} specifies which argument is the format
2659 string argument (starting from one). Since non-static C++ methods have
2660 an implicit @code{this} argument, the arguments of such methods should
2661 be counted from two.
2663 The @code{format_arg} attribute allows you to identify your own
2664 functions that modify format strings, so that GCC can check the
2665 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2666 type function whose operands are a call to one of your own function.
2667 The compiler always treats @code{gettext}, @code{dgettext}, and
2668 @code{dcgettext} in this manner except when strict ISO C support is
2669 requested by @option{-ansi} or an appropriate @option{-std} option, or
2670 @option{-ffreestanding} or @option{-fno-builtin}
2671 is used. @xref{C Dialect Options,,Options
2672 Controlling C Dialect}.
2674 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2675 @code{NSString} reference for compatibility with the @code{format} attribute
2678 The target may also allow additional types in @code{format-arg} attributes.
2679 @xref{Target Format Checks,,Format Checks Specific to Particular
2683 @cindex @code{gnu_inline} function attribute
2684 This attribute should be used with a function that is also declared
2685 with the @code{inline} keyword. It directs GCC to treat the function
2686 as if it were defined in gnu90 mode even when compiling in C99 or
2689 If the function is declared @code{extern}, then this definition of the
2690 function is used only for inlining. In no case is the function
2691 compiled as a standalone function, not even if you take its address
2692 explicitly. Such an address becomes an external reference, as if you
2693 had only declared the function, and had not defined it. This has
2694 almost the effect of a macro. The way to use this is to put a
2695 function definition in a header file with this attribute, and put
2696 another copy of the function, without @code{extern}, in a library
2697 file. The definition in the header file causes most calls to the
2698 function to be inlined. If any uses of the function remain, they
2699 refer to the single copy in the library. Note that the two
2700 definitions of the functions need not be precisely the same, although
2701 if they do not have the same effect your program may behave oddly.
2703 In C, if the function is neither @code{extern} nor @code{static}, then
2704 the function is compiled as a standalone function, as well as being
2705 inlined where possible.
2707 This is how GCC traditionally handled functions declared
2708 @code{inline}. Since ISO C99 specifies a different semantics for
2709 @code{inline}, this function attribute is provided as a transition
2710 measure and as a useful feature in its own right. This attribute is
2711 available in GCC 4.1.3 and later. It is available if either of the
2712 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2713 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2714 Function is As Fast As a Macro}.
2716 In C++, this attribute does not depend on @code{extern} in any way,
2717 but it still requires the @code{inline} keyword to enable its special
2721 @cindex @code{hot} function attribute
2722 The @code{hot} attribute on a function is used to inform the compiler that
2723 the function is a hot spot of the compiled program. The function is
2724 optimized more aggressively and on many targets it is placed into a special
2725 subsection of the text section so all hot functions appear close together,
2728 When profile feedback is available, via @option{-fprofile-use}, hot functions
2729 are automatically detected and this attribute is ignored.
2731 @item ifunc ("@var{resolver}")
2732 @cindex @code{ifunc} function attribute
2733 @cindex indirect functions
2734 @cindex functions that are dynamically resolved
2735 The @code{ifunc} attribute is used to mark a function as an indirect
2736 function using the STT_GNU_IFUNC symbol type extension to the ELF
2737 standard. This allows the resolution of the symbol value to be
2738 determined dynamically at load time, and an optimized version of the
2739 routine can be selected for the particular processor or other system
2740 characteristics determined then. To use this attribute, first define
2741 the implementation functions available, and a resolver function that
2742 returns a pointer to the selected implementation function. The
2743 implementation functions' declarations must match the API of the
2744 function being implemented, the resolver's declaration is be a
2745 function returning pointer to void function returning void:
2748 void *my_memcpy (void *dst, const void *src, size_t len)
2753 static void (*resolve_memcpy (void)) (void)
2755 return my_memcpy; // we'll just always select this routine
2760 The exported header file declaring the function the user calls would
2764 extern void *memcpy (void *, const void *, size_t);
2768 allowing the user to call this as a regular function, unaware of the
2769 implementation. Finally, the indirect function needs to be defined in
2770 the same translation unit as the resolver function:
2773 void *memcpy (void *, const void *, size_t)
2774 __attribute__ ((ifunc ("resolve_memcpy")));
2777 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2778 and GNU C Library version 2.11.1 are required to use this feature.
2781 @itemx interrupt_handler
2782 Many GCC back ends support attributes to indicate that a function is
2783 an interrupt handler, which tells the compiler to generate function
2784 entry and exit sequences that differ from those from regular
2785 functions. The exact syntax and behavior are target-specific;
2786 refer to the following subsections for details.
2789 @cindex @code{leaf} function attribute
2790 Calls to external functions with this attribute must return to the current
2791 compilation unit only by return or by exception handling. In particular, leaf
2792 functions are not allowed to call callback function passed to it from the current
2793 compilation unit or directly call functions exported by the unit or longjmp
2794 into the unit. Leaf function might still call functions from other compilation
2795 units and thus they are not necessarily leaf in the sense that they contain no
2796 function calls at all.
2798 The attribute is intended for library functions to improve dataflow analysis.
2799 The compiler takes the hint that any data not escaping the current compilation unit can
2800 not be used or modified by the leaf function. For example, the @code{sin} function
2801 is a leaf function, but @code{qsort} is not.
2803 Note that leaf functions might invoke signals and signal handlers might be
2804 defined in the current compilation unit and use static variables. The only
2805 compliant way to write such a signal handler is to declare such variables
2808 The attribute has no effect on functions defined within the current compilation
2809 unit. This is to allow easy merging of multiple compilation units into one,
2810 for example, by using the link-time optimization. For this reason the
2811 attribute is not allowed on types to annotate indirect calls.
2815 @cindex @code{malloc} function attribute
2816 @cindex functions that behave like malloc
2817 This tells the compiler that a function is @code{malloc}-like, i.e.,
2818 that the pointer @var{P} returned by the function cannot alias any
2819 other pointer valid when the function returns, and moreover no
2820 pointers to valid objects occur in any storage addressed by @var{P}.
2822 Using this attribute can improve optimization. Functions like
2823 @code{malloc} and @code{calloc} have this property because they return
2824 a pointer to uninitialized or zeroed-out storage. However, functions
2825 like @code{realloc} do not have this property, as they can return a
2826 pointer to storage containing pointers.
2829 @cindex @code{no_icf} function attribute
2830 This function attribute prevents a functions from being merged with another
2831 semantically equivalent function.
2833 @item no_instrument_function
2834 @cindex @code{no_instrument_function} function attribute
2835 @opindex finstrument-functions
2836 If @option{-finstrument-functions} is given, profiling function calls are
2837 generated at entry and exit of most user-compiled functions.
2838 Functions with this attribute are not so instrumented.
2841 @cindex @code{no_reorder} function attribute
2842 Do not reorder functions or variables marked @code{no_reorder}
2843 against each other or top level assembler statements the executable.
2844 The actual order in the program will depend on the linker command
2845 line. Static variables marked like this are also not removed.
2846 This has a similar effect
2847 as the @option{-fno-toplevel-reorder} option, but only applies to the
2850 @item no_sanitize_address
2851 @itemx no_address_safety_analysis
2852 @cindex @code{no_sanitize_address} function attribute
2853 The @code{no_sanitize_address} attribute on functions is used
2854 to inform the compiler that it should not instrument memory accesses
2855 in the function when compiling with the @option{-fsanitize=address} option.
2856 The @code{no_address_safety_analysis} is a deprecated alias of the
2857 @code{no_sanitize_address} attribute, new code should use
2858 @code{no_sanitize_address}.
2860 @item no_sanitize_thread
2861 @cindex @code{no_sanitize_thread} function attribute
2862 The @code{no_sanitize_thread} attribute on functions is used
2863 to inform the compiler that it should not instrument memory accesses
2864 in the function when compiling with the @option{-fsanitize=thread} option.
2866 @item no_sanitize_undefined
2867 @cindex @code{no_sanitize_undefined} function attribute
2868 The @code{no_sanitize_undefined} attribute on functions is used
2869 to inform the compiler that it should not check for undefined behavior
2870 in the function when compiling with the @option{-fsanitize=undefined} option.
2872 @item no_split_stack
2873 @cindex @code{no_split_stack} function attribute
2874 @opindex fsplit-stack
2875 If @option{-fsplit-stack} is given, functions have a small
2876 prologue which decides whether to split the stack. Functions with the
2877 @code{no_split_stack} attribute do not have that prologue, and thus
2878 may run with only a small amount of stack space available.
2881 @cindex @code{noclone} function attribute
2882 This function attribute prevents a function from being considered for
2883 cloning---a mechanism that produces specialized copies of functions
2884 and which is (currently) performed by interprocedural constant
2888 @cindex @code{noinline} function attribute
2889 This function attribute prevents a function from being considered for
2891 @c Don't enumerate the optimizations by name here; we try to be
2892 @c future-compatible with this mechanism.
2893 If the function does not have side-effects, there are optimizations
2894 other than inlining that cause function calls to be optimized away,
2895 although the function call is live. To keep such calls from being
2902 (@pxref{Extended Asm}) in the called function, to serve as a special
2905 @item nonnull (@var{arg-index}, @dots{})
2906 @cindex @code{nonnull} function attribute
2907 @cindex functions with non-null pointer arguments
2908 The @code{nonnull} attribute specifies that some function parameters should
2909 be non-null pointers. For instance, the declaration:
2913 my_memcpy (void *dest, const void *src, size_t len)
2914 __attribute__((nonnull (1, 2)));
2918 causes the compiler to check that, in calls to @code{my_memcpy},
2919 arguments @var{dest} and @var{src} are non-null. If the compiler
2920 determines that a null pointer is passed in an argument slot marked
2921 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2922 is issued. The compiler may also choose to make optimizations based
2923 on the knowledge that certain function arguments will never be null.
2925 If no argument index list is given to the @code{nonnull} attribute,
2926 all pointer arguments are marked as non-null. To illustrate, the
2927 following declaration is equivalent to the previous example:
2931 my_memcpy (void *dest, const void *src, size_t len)
2932 __attribute__((nonnull));
2936 @cindex @code{noreturn} function attribute
2937 @cindex functions that never return
2938 A few standard library functions, such as @code{abort} and @code{exit},
2939 cannot return. GCC knows this automatically. Some programs define
2940 their own functions that never return. You can declare them
2941 @code{noreturn} to tell the compiler this fact. For example,
2945 void fatal () __attribute__ ((noreturn));
2948 fatal (/* @r{@dots{}} */)
2950 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2956 The @code{noreturn} keyword tells the compiler to assume that
2957 @code{fatal} cannot return. It can then optimize without regard to what
2958 would happen if @code{fatal} ever did return. This makes slightly
2959 better code. More importantly, it helps avoid spurious warnings of
2960 uninitialized variables.
2962 The @code{noreturn} keyword does not affect the exceptional path when that
2963 applies: a @code{noreturn}-marked function may still return to the caller
2964 by throwing an exception or calling @code{longjmp}.
2966 Do not assume that registers saved by the calling function are
2967 restored before calling the @code{noreturn} function.
2969 It does not make sense for a @code{noreturn} function to have a return
2970 type other than @code{void}.
2973 @cindex @code{nothrow} function attribute
2974 The @code{nothrow} attribute is used to inform the compiler that a
2975 function cannot throw an exception. For example, most functions in
2976 the standard C library can be guaranteed not to throw an exception
2977 with the notable exceptions of @code{qsort} and @code{bsearch} that
2978 take function pointer arguments.
2981 @cindex @code{noplt} function attribute
2982 The @code{noplt} attribute is the counterpart to option @option{-fno-plt} and
2983 does not use PLT for calls to functions marked with this attribute in position
2988 /* Externally defined function foo. */
2989 int foo () __attribute__ ((noplt));
2992 main (/* @r{@dots{}} */)
3001 The @code{noplt} attribute on function foo tells the compiler to assume that
3002 the function foo is externally defined and the call to foo must avoid the PLT
3003 in position independent code.
3005 Additionally, a few targets also convert calls to those functions that are
3006 marked to not use the PLT to use the GOT instead for non-position independent
3010 @cindex @code{optimize} function attribute
3011 The @code{optimize} attribute is used to specify that a function is to
3012 be compiled with different optimization options than specified on the
3013 command line. Arguments can either be numbers or strings. Numbers
3014 are assumed to be an optimization level. Strings that begin with
3015 @code{O} are assumed to be an optimization option, while other options
3016 are assumed to be used with a @code{-f} prefix. You can also use the
3017 @samp{#pragma GCC optimize} pragma to set the optimization options
3018 that affect more than one function.
3019 @xref{Function Specific Option Pragmas}, for details about the
3020 @samp{#pragma GCC optimize} pragma.
3022 This can be used for instance to have frequently-executed functions
3023 compiled with more aggressive optimization options that produce faster
3024 and larger code, while other functions can be compiled with less
3028 @cindex @code{pure} function attribute
3029 @cindex functions that have no side effects
3030 Many functions have no effects except the return value and their
3031 return value depends only on the parameters and/or global variables.
3032 Such a function can be subject
3033 to common subexpression elimination and loop optimization just as an
3034 arithmetic operator would be. These functions should be declared
3035 with the attribute @code{pure}. For example,
3038 int square (int) __attribute__ ((pure));
3042 says that the hypothetical function @code{square} is safe to call
3043 fewer times than the program says.
3045 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3046 Interesting non-pure functions are functions with infinite loops or those
3047 depending on volatile memory or other system resource, that may change between
3048 two consecutive calls (such as @code{feof} in a multithreading environment).
3050 @item returns_nonnull
3051 @cindex @code{returns_nonnull} function attribute
3052 The @code{returns_nonnull} attribute specifies that the function
3053 return value should be a non-null pointer. For instance, the declaration:
3057 mymalloc (size_t len) __attribute__((returns_nonnull));
3061 lets the compiler optimize callers based on the knowledge
3062 that the return value will never be null.
3065 @cindex @code{returns_twice} function attribute
3066 @cindex functions that return more than once
3067 The @code{returns_twice} attribute tells the compiler that a function may
3068 return more than one time. The compiler ensures that all registers
3069 are dead before calling such a function and emits a warning about
3070 the variables that may be clobbered after the second return from the
3071 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3072 The @code{longjmp}-like counterpart of such function, if any, might need
3073 to be marked with the @code{noreturn} attribute.
3075 @item section ("@var{section-name}")
3076 @cindex @code{section} function attribute
3077 @cindex functions in arbitrary sections
3078 Normally, the compiler places the code it generates in the @code{text} section.
3079 Sometimes, however, you need additional sections, or you need certain
3080 particular functions to appear in special sections. The @code{section}
3081 attribute specifies that a function lives in a particular section.
3082 For example, the declaration:
3085 extern void foobar (void) __attribute__ ((section ("bar")));
3089 puts the function @code{foobar} in the @code{bar} section.
3091 Some file formats do not support arbitrary sections so the @code{section}
3092 attribute is not available on all platforms.
3093 If you need to map the entire contents of a module to a particular
3094 section, consider using the facilities of the linker instead.
3097 @cindex @code{sentinel} function attribute
3098 This function attribute ensures that a parameter in a function call is
3099 an explicit @code{NULL}. The attribute is only valid on variadic
3100 functions. By default, the sentinel is located at position zero, the
3101 last parameter of the function call. If an optional integer position
3102 argument P is supplied to the attribute, the sentinel must be located at
3103 position P counting backwards from the end of the argument list.
3106 __attribute__ ((sentinel))
3108 __attribute__ ((sentinel(0)))
3111 The attribute is automatically set with a position of 0 for the built-in
3112 functions @code{execl} and @code{execlp}. The built-in function
3113 @code{execle} has the attribute set with a position of 1.
3115 A valid @code{NULL} in this context is defined as zero with any pointer
3116 type. If your system defines the @code{NULL} macro with an integer type
3117 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3118 with a copy that redefines NULL appropriately.
3120 The warnings for missing or incorrect sentinels are enabled with
3124 @cindex @code{stack_protect} function attribute
3125 This function attribute make a stack protection of the function if
3126 flags @option{fstack-protector} or @option{fstack-protector-strong}
3127 or @option{fstack-protector-explicit} are set.
3129 @item target_clones (@var{options})
3130 @cindex @code{target_clones} function attribute
3131 The @code{target_clones} attribute is used to specify that a function is to
3132 be cloned into multiple versions compiled with different target options
3133 than specified on the command line. The supported options and restrictions
3134 are the same as for @code{target} attribute.
3136 For instance on an x86, you could compile a function with
3137 @code{target_clones("sse4.1,avx")}. It will create 2 function clones,
3138 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3139 At the function call it will create resolver @code{ifunc}, that will
3140 dynamically call a clone suitable for current architecture.
3143 @itemx simd("@var{mask}")
3144 @cindex @code{simd} function attribute.
3145 This attribute enables creation of one or more function versions that
3146 can process multiple arguments using SIMD instructions from a
3147 single invocation. Specifying this attribute allows compiler to
3148 assume that such versions are available at link time (provided
3149 in the same or another translation unit). Generated versions are
3150 target dependent and described in corresponding Vector ABI document. For
3151 x86_64 target this document can be found
3152 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3153 The attribute should not be used together with Cilk Plus @code{vector}
3154 attribute on the same function.
3155 If the attribute is specified and @code{#pragma omp declare simd}
3156 present on a declaration and @code{-fopenmp} or @code{-fopenmp-simd}
3157 switch is specified, then the attribute is ignored.
3158 The optional argument @var{mask} may have "notinbranch" or "inbranch"
3159 value and instructs the compiler to generate non-masked or masked
3160 clones correspondingly. By default, all clones are generated.
3162 @item target (@var{options})
3163 @cindex @code{target} function attribute
3164 Multiple target back ends implement the @code{target} attribute
3165 to specify that a function is to
3166 be compiled with different target options than specified on the
3167 command line. This can be used for instance to have functions
3168 compiled with a different ISA (instruction set architecture) than the
3169 default. You can also use the @samp{#pragma GCC target} pragma to set
3170 more than one function to be compiled with specific target options.
3171 @xref{Function Specific Option Pragmas}, for details about the
3172 @samp{#pragma GCC target} pragma.
3174 For instance, on an x86, you could declare one function with the
3175 @code{target("sse4.1,arch=core2")} attribute and another with
3176 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3177 compiling the first function with @option{-msse4.1} and
3178 @option{-march=core2} options, and the second function with
3179 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3180 to make sure that a function is only invoked on a machine that
3181 supports the particular ISA it is compiled for (for example by using
3182 @code{cpuid} on x86 to determine what feature bits and architecture
3186 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3187 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3190 You can either use multiple
3191 strings separated by commas to specify multiple options,
3192 or separate the options with a comma (@samp{,}) within a single string.
3194 The options supported are specific to each target; refer to @ref{x86
3195 Function Attributes}, @ref{PowerPC Function Attributes},
3196 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3200 @cindex @code{unused} function attribute
3201 This attribute, attached to a function, means that the function is meant
3202 to be possibly unused. GCC does not produce a warning for this
3206 @cindex @code{used} function attribute
3207 This attribute, attached to a function, means that code must be emitted
3208 for the function even if it appears that the function is not referenced.
3209 This is useful, for example, when the function is referenced only in
3212 When applied to a member function of a C++ class template, the
3213 attribute also means that the function is instantiated if the
3214 class itself is instantiated.
3216 @item visibility ("@var{visibility_type}")
3217 @cindex @code{visibility} function attribute
3218 This attribute affects the linkage of the declaration to which it is attached.
3219 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3220 (@pxref{Common Type Attributes}) as well as functions.
3222 There are four supported @var{visibility_type} values: default,
3223 hidden, protected or internal visibility.
3226 void __attribute__ ((visibility ("protected")))
3227 f () @{ /* @r{Do something.} */; @}
3228 int i __attribute__ ((visibility ("hidden")));
3231 The possible values of @var{visibility_type} correspond to the
3232 visibility settings in the ELF gABI.
3235 @c keep this list of visibilities in alphabetical order.
3238 Default visibility is the normal case for the object file format.
3239 This value is available for the visibility attribute to override other
3240 options that may change the assumed visibility of entities.
3242 On ELF, default visibility means that the declaration is visible to other
3243 modules and, in shared libraries, means that the declared entity may be
3246 On Darwin, default visibility means that the declaration is visible to
3249 Default visibility corresponds to ``external linkage'' in the language.
3252 Hidden visibility indicates that the entity declared has a new
3253 form of linkage, which we call ``hidden linkage''. Two
3254 declarations of an object with hidden linkage refer to the same object
3255 if they are in the same shared object.
3258 Internal visibility is like hidden visibility, but with additional
3259 processor specific semantics. Unless otherwise specified by the
3260 psABI, GCC defines internal visibility to mean that a function is
3261 @emph{never} called from another module. Compare this with hidden
3262 functions which, while they cannot be referenced directly by other
3263 modules, can be referenced indirectly via function pointers. By
3264 indicating that a function cannot be called from outside the module,
3265 GCC may for instance omit the load of a PIC register since it is known
3266 that the calling function loaded the correct value.
3269 Protected visibility is like default visibility except that it
3270 indicates that references within the defining module bind to the
3271 definition in that module. That is, the declared entity cannot be
3272 overridden by another module.
3276 All visibilities are supported on many, but not all, ELF targets
3277 (supported when the assembler supports the @samp{.visibility}
3278 pseudo-op). Default visibility is supported everywhere. Hidden
3279 visibility is supported on Darwin targets.
3281 The visibility attribute should be applied only to declarations that
3282 would otherwise have external linkage. The attribute should be applied
3283 consistently, so that the same entity should not be declared with
3284 different settings of the attribute.
3286 In C++, the visibility attribute applies to types as well as functions
3287 and objects, because in C++ types have linkage. A class must not have
3288 greater visibility than its non-static data member types and bases,
3289 and class members default to the visibility of their class. Also, a
3290 declaration without explicit visibility is limited to the visibility
3293 In C++, you can mark member functions and static member variables of a
3294 class with the visibility attribute. This is useful if you know a
3295 particular method or static member variable should only be used from
3296 one shared object; then you can mark it hidden while the rest of the
3297 class has default visibility. Care must be taken to avoid breaking
3298 the One Definition Rule; for example, it is usually not useful to mark
3299 an inline method as hidden without marking the whole class as hidden.
3301 A C++ namespace declaration can also have the visibility attribute.
3304 namespace nspace1 __attribute__ ((visibility ("protected")))
3305 @{ /* @r{Do something.} */; @}
3308 This attribute applies only to the particular namespace body, not to
3309 other definitions of the same namespace; it is equivalent to using
3310 @samp{#pragma GCC visibility} before and after the namespace
3311 definition (@pxref{Visibility Pragmas}).
3313 In C++, if a template argument has limited visibility, this
3314 restriction is implicitly propagated to the template instantiation.
3315 Otherwise, template instantiations and specializations default to the
3316 visibility of their template.
3318 If both the template and enclosing class have explicit visibility, the
3319 visibility from the template is used.
3321 @item warn_unused_result
3322 @cindex @code{warn_unused_result} function attribute
3323 The @code{warn_unused_result} attribute causes a warning to be emitted
3324 if a caller of the function with this attribute does not use its
3325 return value. This is useful for functions where not checking
3326 the result is either a security problem or always a bug, such as
3330 int fn () __attribute__ ((warn_unused_result));
3333 if (fn () < 0) return -1;
3340 results in warning on line 5.
3343 @cindex @code{weak} function attribute
3344 The @code{weak} attribute causes the declaration to be emitted as a weak
3345 symbol rather than a global. This is primarily useful in defining
3346 library functions that can be overridden in user code, though it can
3347 also be used with non-function declarations. Weak symbols are supported
3348 for ELF targets, and also for a.out targets when using the GNU assembler
3352 @itemx weakref ("@var{target}")
3353 @cindex @code{weakref} function attribute
3354 The @code{weakref} attribute marks a declaration as a weak reference.
3355 Without arguments, it should be accompanied by an @code{alias} attribute
3356 naming the target symbol. Optionally, the @var{target} may be given as
3357 an argument to @code{weakref} itself. In either case, @code{weakref}
3358 implicitly marks the declaration as @code{weak}. Without a
3359 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3360 @code{weakref} is equivalent to @code{weak}.
3363 static int x() __attribute__ ((weakref ("y")));
3364 /* is equivalent to... */
3365 static int x() __attribute__ ((weak, weakref, alias ("y")));
3367 static int x() __attribute__ ((weakref));
3368 static int x() __attribute__ ((alias ("y")));
3371 A weak reference is an alias that does not by itself require a
3372 definition to be given for the target symbol. If the target symbol is
3373 only referenced through weak references, then it becomes a @code{weak}
3374 undefined symbol. If it is directly referenced, however, then such
3375 strong references prevail, and a definition is required for the
3376 symbol, not necessarily in the same translation unit.
3378 The effect is equivalent to moving all references to the alias to a
3379 separate translation unit, renaming the alias to the aliased symbol,
3380 declaring it as weak, compiling the two separate translation units and
3381 performing a reloadable link on them.
3383 At present, a declaration to which @code{weakref} is attached can
3384 only be @code{static}.
3389 @cindex lower memory region on the MSP430
3390 @cindex upper memory region on the MSP430
3391 @cindex either memory region on the MSP430
3392 On the MSP430 target these attributes can be used to specify whether
3393 the function or variable should be placed into low memory, high
3394 memory, or the placement should be left to the linker to decide. The
3395 attributes are only significant if compiling for the MSP430X
3398 The attributes work in conjunction with a linker script that has been
3399 augmented to specify where to place sections with a @code{.lower} and
3400 a @code{.upper} prefix. So for example as well as placing the
3401 @code{.data} section the script would also specify the placement of a
3402 @code{.lower.data} and a @code{.upper.data} section. The intention
3403 being that @code{lower} sections are placed into a small but easier to
3404 access memory region and the upper sections are placed into a larger, but
3405 slower to access region.
3407 The @code{either} attribute is special. It tells the linker to place
3408 the object into the corresponding @code{lower} section if there is
3409 room for it. If there is insufficient room then the object is placed
3410 into the corresponding @code{upper} section instead. Note - the
3411 placement algorithm is not very sophisticated. It will not attempt to
3412 find an optimal packing of the @code{lower} sections. It just makes
3413 one pass over the objects and does the best that it can. Using the
3414 @option{-ffunction-sections} and @option{-fdata-sections} command line
3415 options can help the packing however, since they produce smaller,
3416 easier to pack regions.
3419 On the MSP430 a function can be given the @code{reentant} attribute.
3420 This makes the function disable interrupts upon entry and enable
3421 interrupts upon exit. Reentrant functions cannot be @code{naked}.
3424 On the MSP430 a function can be given the @code{critical} attribute.
3425 This makes the function disable interrupts upon entry and restore the
3426 previous interrupt enabled/disabled state upon exit. A function
3427 cannot have both the @code{reentrant} and @code{critical} attributes.
3428 Critical functions cannot be @code{naked}.
3431 On the MSP430 a function can be given the @code{wakeup} attribute.
3432 Such a function must also have the @code{interrupt} attribute. When a
3433 function with the @code{wakeup} attribute exists the processor will be
3434 woken up from any low-power state in which it may be residing.
3438 @c This is the end of the target-independent attribute table
3440 @node AArch64 Function Attributes
3441 @subsection AArch64 Function Attributes
3443 The following target-specific function attributes are available for the
3444 AArch64 target. For the most part, these options mirror the behavior of
3445 similar command-line options (@pxref{AArch64 Options}), but on a
3449 @item general-regs-only
3450 @cindex @code{general-regs-only} function attribute, AArch64
3451 Indicates that no floating-point or Advanced SIMD registers should be
3452 used when generating code for this function. If the function explicitly
3453 uses floating-point code, then the compiler gives an error. This is
3454 the same behavior as that of the command-line option
3455 @option{-mgeneral-regs-only}.
3457 @item fix-cortex-a53-835769
3458 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3459 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3460 applied to this function. To explicitly disable the workaround for this
3461 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3462 This corresponds to the behavior of the command line options
3463 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3466 @cindex @code{cmodel=} function attribute, AArch64
3467 Indicates that code should be generated for a particular code model for
3468 this function. The behavior and permissible arguments are the same as
3469 for the command line option @option{-mcmodel=}.
3472 @cindex @code{strict-align} function attribute, AArch64
3473 Indicates that the compiler should not assume that unaligned memory references
3474 are handled by the system. The behavior is the same as for the command-line
3475 option @option{-mstrict-align}.
3477 @item omit-leaf-frame-pointer
3478 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3479 Indicates that the frame pointer should be omitted for a leaf function call.
3480 To keep the frame pointer, the inverse attribute
3481 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3482 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3483 and @option{-mno-omit-leaf-frame-pointer}.
3486 @cindex @code{tls-dialect=} function attribute, AArch64
3487 Specifies the TLS dialect to use for this function. The behavior and
3488 permissible arguments are the same as for the command-line option
3489 @option{-mtls-dialect=}.
3492 @cindex @code{arch=} function attribute, AArch64
3493 Specifies the architecture version and architectural extensions to use
3494 for this function. The behavior and permissible arguments are the same as
3495 for the @option{-march=} command-line option.
3498 @cindex @code{tune=} function attribute, AArch64
3499 Specifies the core for which to tune the performance of this function.
3500 The behavior and permissible arguments are the same as for the @option{-mtune=}
3501 command-line option.
3504 @cindex @code{cpu=} function attribute, AArch64
3505 Specifies the core for which to tune the performance of this function and also
3506 whose architectural features to use. The behavior and valid arguments are the
3507 same as for the @option{-mcpu=} command-line option.
3511 The above target attributes can be specified as follows:
3514 __attribute__((target("@var{attr-string}")))
3522 where @code{@var{attr-string}} is one of the attribute strings specified above.
3524 Additionally, the architectural extension string may be specified on its
3525 own. This can be used to turn on and off particular architectural extensions
3526 without having to specify a particular architecture version or core. Example:
3529 __attribute__((target("+crc+nocrypto")))
3537 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3538 extension and disables the @code{crypto} extension for the function @code{foo}
3539 without modifying an existing @option{-march=} or @option{-mcpu} option.
3541 Multiple target function attributes can be specified by separating them with
3542 a comma. For example:
3544 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3552 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3553 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3555 @subsubsection Inlining rules
3556 Specifying target attributes on individual functions or performing link-time
3557 optimization across translation units compiled with different target options
3558 can affect function inlining rules:
3560 In particular, a caller function can inline a callee function only if the
3561 architectural features available to the callee are a subset of the features
3562 available to the caller.
3563 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3564 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3565 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3566 because the all the architectural features that function @code{bar} requires
3567 are available to function @code{foo}. Conversely, function @code{bar} cannot
3568 inline function @code{foo}.
3570 Additionally inlining a function compiled with @option{-mstrict-align} into a
3571 function compiled without @code{-mstrict-align} is not allowed.
3572 However, inlining a function compiled without @option{-mstrict-align} into a
3573 function compiled with @option{-mstrict-align} is allowed.
3575 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3576 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3577 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3578 architectural feature rules specified above.
3580 @node ARC Function Attributes
3581 @subsection ARC Function Attributes
3583 These function attributes are supported by the ARC back end:
3587 @cindex @code{interrupt} function attribute, ARC
3588 Use this attribute to indicate
3589 that the specified function is an interrupt handler. The compiler generates
3590 function entry and exit sequences suitable for use in an interrupt handler
3591 when this attribute is present.
3593 On the ARC, you must specify the kind of interrupt to be handled
3594 in a parameter to the interrupt attribute like this:
3597 void f () __attribute__ ((interrupt ("ilink1")));
3600 Permissible values for this parameter are: @w{@code{ilink1}} and
3606 @cindex @code{long_call} function attribute, ARC
3607 @cindex @code{medium_call} function attribute, ARC
3608 @cindex @code{short_call} function attribute, ARC
3609 @cindex indirect calls, ARC
3610 These attributes specify how a particular function is called.
3611 These attributes override the
3612 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3613 command-line switches and @code{#pragma long_calls} settings.
3615 For ARC, a function marked with the @code{long_call} attribute is
3616 always called using register-indirect jump-and-link instructions,
3617 thereby enabling the called function to be placed anywhere within the
3618 32-bit address space. A function marked with the @code{medium_call}
3619 attribute will always be close enough to be called with an unconditional
3620 branch-and-link instruction, which has a 25-bit offset from
3621 the call site. A function marked with the @code{short_call}
3622 attribute will always be close enough to be called with a conditional
3623 branch-and-link instruction, which has a 21-bit offset from
3627 @node ARM Function Attributes
3628 @subsection ARM Function Attributes
3630 These function attributes are supported for ARM targets:
3634 @cindex @code{interrupt} function attribute, ARM
3635 Use this attribute to indicate
3636 that the specified function is an interrupt handler. The compiler generates
3637 function entry and exit sequences suitable for use in an interrupt handler
3638 when this attribute is present.
3640 You can specify the kind of interrupt to be handled by
3641 adding an optional parameter to the interrupt attribute like this:
3644 void f () __attribute__ ((interrupt ("IRQ")));
3648 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3649 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3651 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3652 may be called with a word-aligned stack pointer.
3655 @cindex @code{isr} function attribute, ARM
3656 Use this attribute on ARM to write Interrupt Service Routines. This is an
3657 alias to the @code{interrupt} attribute above.
3661 @cindex @code{long_call} function attribute, ARM
3662 @cindex @code{short_call} function attribute, ARM
3663 @cindex indirect calls, ARM
3664 These attributes specify how a particular function is called.
3665 These attributes override the
3666 @option{-mlong-calls} (@pxref{ARM Options})
3667 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3668 @code{long_call} attribute indicates that the function might be far
3669 away from the call site and require a different (more expensive)
3670 calling sequence. The @code{short_call} attribute always places
3671 the offset to the function from the call site into the @samp{BL}
3672 instruction directly.
3675 @cindex @code{naked} function attribute, ARM
3676 This attribute allows the compiler to construct the
3677 requisite function declaration, while allowing the body of the
3678 function to be assembly code. The specified function will not have
3679 prologue/epilogue sequences generated by the compiler. Only basic
3680 @code{asm} statements can safely be included in naked functions
3681 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3682 basic @code{asm} and C code may appear to work, they cannot be
3683 depended upon to work reliably and are not supported.
3686 @cindex @code{pcs} function attribute, ARM
3688 The @code{pcs} attribute can be used to control the calling convention
3689 used for a function on ARM. The attribute takes an argument that specifies
3690 the calling convention to use.
3692 When compiling using the AAPCS ABI (or a variant of it) then valid
3693 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3694 order to use a variant other than @code{"aapcs"} then the compiler must
3695 be permitted to use the appropriate co-processor registers (i.e., the
3696 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3700 /* Argument passed in r0, and result returned in r0+r1. */
3701 double f2d (float) __attribute__((pcs("aapcs")));
3704 Variadic functions always use the @code{"aapcs"} calling convention and
3705 the compiler rejects attempts to specify an alternative.
3707 @item target (@var{options})
3708 @cindex @code{target} function attribute
3709 As discussed in @ref{Common Function Attributes}, this attribute
3710 allows specification of target-specific compilation options.
3712 On ARM, the following options are allowed:
3716 @cindex @code{target("thumb")} function attribute, ARM
3717 Force code generation in the Thumb (T16/T32) ISA, depending on the
3721 @cindex @code{target("arm")} function attribute, ARM
3722 Force code generation in the ARM (A32) ISA.
3724 Functions from different modes can be inlined in the caller's mode.
3727 @cindex @code{target("fpu=")} function attribute, ARM
3728 Specifies the fpu for which to tune the performance of this function.
3729 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3730 command-line option.
3736 @node AVR Function Attributes
3737 @subsection AVR Function Attributes
3739 These function attributes are supported by the AVR back end:
3743 @cindex @code{interrupt} function attribute, AVR
3744 Use this attribute to indicate
3745 that the specified function is an interrupt handler. The compiler generates
3746 function entry and exit sequences suitable for use in an interrupt handler
3747 when this attribute is present.
3749 On the AVR, the hardware globally disables interrupts when an
3750 interrupt is executed. The first instruction of an interrupt handler
3751 declared with this attribute is a @code{SEI} instruction to
3752 re-enable interrupts. See also the @code{signal} function attribute
3753 that does not insert a @code{SEI} instruction. If both @code{signal} and
3754 @code{interrupt} are specified for the same function, @code{signal}
3755 is silently ignored.
3758 @cindex @code{naked} function attribute, AVR
3759 This attribute allows the compiler to construct the
3760 requisite function declaration, while allowing the body of the
3761 function to be assembly code. The specified function will not have
3762 prologue/epilogue sequences generated by the compiler. Only basic
3763 @code{asm} statements can safely be included in naked functions
3764 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3765 basic @code{asm} and C code may appear to work, they cannot be
3766 depended upon to work reliably and are not supported.
3770 @cindex @code{OS_main} function attribute, AVR
3771 @cindex @code{OS_task} function attribute, AVR
3772 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3773 do not save/restore any call-saved register in their prologue/epilogue.
3775 The @code{OS_main} attribute can be used when there @emph{is
3776 guarantee} that interrupts are disabled at the time when the function
3777 is entered. This saves resources when the stack pointer has to be
3778 changed to set up a frame for local variables.
3780 The @code{OS_task} attribute can be used when there is @emph{no
3781 guarantee} that interrupts are disabled at that time when the function
3782 is entered like for, e@.g@. task functions in a multi-threading operating
3783 system. In that case, changing the stack pointer register is
3784 guarded by save/clear/restore of the global interrupt enable flag.
3786 The differences to the @code{naked} function attribute are:
3788 @item @code{naked} functions do not have a return instruction whereas
3789 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3790 @code{RETI} return instruction.
3791 @item @code{naked} functions do not set up a frame for local variables
3792 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3797 @cindex @code{signal} function attribute, AVR
3798 Use this attribute on the AVR to indicate that the specified
3799 function is an interrupt handler. The compiler generates function
3800 entry and exit sequences suitable for use in an interrupt handler when this
3801 attribute is present.
3803 See also the @code{interrupt} function attribute.
3805 The AVR hardware globally disables interrupts when an interrupt is executed.
3806 Interrupt handler functions defined with the @code{signal} attribute
3807 do not re-enable interrupts. It is save to enable interrupts in a
3808 @code{signal} handler. This ``save'' only applies to the code
3809 generated by the compiler and not to the IRQ layout of the
3810 application which is responsibility of the application.
3812 If both @code{signal} and @code{interrupt} are specified for the same
3813 function, @code{signal} is silently ignored.
3816 @node Blackfin Function Attributes
3817 @subsection Blackfin Function Attributes
3819 These function attributes are supported by the Blackfin back end:
3823 @item exception_handler
3824 @cindex @code{exception_handler} function attribute
3825 @cindex exception handler functions, Blackfin
3826 Use this attribute on the Blackfin to indicate that the specified function
3827 is an exception handler. The compiler generates function entry and
3828 exit sequences suitable for use in an exception handler when this
3829 attribute is present.
3831 @item interrupt_handler
3832 @cindex @code{interrupt_handler} function attribute, Blackfin
3833 Use this attribute to
3834 indicate that the specified function is an interrupt handler. The compiler
3835 generates function entry and exit sequences suitable for use in an
3836 interrupt handler when this attribute is present.
3839 @cindex @code{kspisusp} function attribute, Blackfin
3840 @cindex User stack pointer in interrupts on the Blackfin
3841 When used together with @code{interrupt_handler}, @code{exception_handler}
3842 or @code{nmi_handler}, code is generated to load the stack pointer
3843 from the USP register in the function prologue.
3846 @cindex @code{l1_text} function attribute, Blackfin
3847 This attribute specifies a function to be placed into L1 Instruction
3848 SRAM@. The function is put into a specific section named @code{.l1.text}.
3849 With @option{-mfdpic}, function calls with a such function as the callee
3850 or caller uses inlined PLT.
3853 @cindex @code{l2} function attribute, Blackfin
3854 This attribute specifies a function to be placed into L2
3855 SRAM. The function is put into a specific section named
3856 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3861 @cindex indirect calls, Blackfin
3862 @cindex @code{longcall} function attribute, Blackfin
3863 @cindex @code{shortcall} function attribute, Blackfin
3864 The @code{longcall} attribute
3865 indicates that the function might be far away from the call site and
3866 require a different (more expensive) calling sequence. The
3867 @code{shortcall} attribute indicates that the function is always close
3868 enough for the shorter calling sequence to be used. These attributes
3869 override the @option{-mlongcall} switch.
3872 @cindex @code{nesting} function attribute, Blackfin
3873 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3874 Use this attribute together with @code{interrupt_handler},
3875 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3876 entry code should enable nested interrupts or exceptions.
3879 @cindex @code{nmi_handler} function attribute, Blackfin
3880 @cindex NMI handler functions on the Blackfin processor
3881 Use this attribute on the Blackfin to indicate that the specified function
3882 is an NMI handler. The compiler generates function entry and
3883 exit sequences suitable for use in an NMI handler when this
3884 attribute is present.
3887 @cindex @code{saveall} function attribute, Blackfin
3888 @cindex save all registers on the Blackfin
3889 Use this attribute to indicate that
3890 all registers except the stack pointer should be saved in the prologue
3891 regardless of whether they are used or not.
3894 @node CR16 Function Attributes
3895 @subsection CR16 Function Attributes
3897 These function attributes are supported by the CR16 back end:
3901 @cindex @code{interrupt} function attribute, CR16
3902 Use this attribute to indicate
3903 that the specified function is an interrupt handler. The compiler generates
3904 function entry and exit sequences suitable for use in an interrupt handler
3905 when this attribute is present.
3908 @node Epiphany Function Attributes
3909 @subsection Epiphany Function Attributes
3911 These function attributes are supported by the Epiphany back end:
3915 @cindex @code{disinterrupt} function attribute, Epiphany
3916 This attribute causes the compiler to emit
3917 instructions to disable interrupts for the duration of the given
3920 @item forwarder_section
3921 @cindex @code{forwarder_section} function attribute, Epiphany
3922 This attribute modifies the behavior of an interrupt handler.
3923 The interrupt handler may be in external memory which cannot be
3924 reached by a branch instruction, so generate a local memory trampoline
3925 to transfer control. The single parameter identifies the section where
3926 the trampoline is placed.
3929 @cindex @code{interrupt} function attribute, Epiphany
3930 Use this attribute to indicate
3931 that the specified function is an interrupt handler. The compiler generates
3932 function entry and exit sequences suitable for use in an interrupt handler
3933 when this attribute is present. It may also generate
3934 a special section with code to initialize the interrupt vector table.
3936 On Epiphany targets one or more optional parameters can be added like this:
3939 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3942 Permissible values for these parameters are: @w{@code{reset}},
3943 @w{@code{software_exception}}, @w{@code{page_miss}},
3944 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3945 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3946 Multiple parameters indicate that multiple entries in the interrupt
3947 vector table should be initialized for this function, i.e.@: for each
3948 parameter @w{@var{name}}, a jump to the function is emitted in
3949 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3950 entirely, in which case no interrupt vector table entry is provided.
3952 Note that interrupts are enabled inside the function
3953 unless the @code{disinterrupt} attribute is also specified.
3955 The following examples are all valid uses of these attributes on
3958 void __attribute__ ((interrupt)) universal_handler ();
3959 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3960 void __attribute__ ((interrupt ("dma0, dma1")))
3961 universal_dma_handler ();
3962 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3963 fast_timer_handler ();
3964 void __attribute__ ((interrupt ("dma0, dma1"),
3965 forwarder_section ("tramp")))
3966 external_dma_handler ();
3971 @cindex @code{long_call} function attribute, Epiphany
3972 @cindex @code{short_call} function attribute, Epiphany
3973 @cindex indirect calls, Epiphany
3974 These attributes specify how a particular function is called.
3975 These attributes override the
3976 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3977 command-line switch and @code{#pragma long_calls} settings.
3981 @node H8/300 Function Attributes
3982 @subsection H8/300 Function Attributes
3984 These function attributes are available for H8/300 targets:
3987 @item function_vector
3988 @cindex @code{function_vector} function attribute, H8/300
3989 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3990 that the specified function should be called through the function vector.
3991 Calling a function through the function vector reduces code size; however,
3992 the function vector has a limited size (maximum 128 entries on the H8/300
3993 and 64 entries on the H8/300H and H8S)
3994 and shares space with the interrupt vector.
3996 @item interrupt_handler
3997 @cindex @code{interrupt_handler} function attribute, H8/300
3998 Use this attribute on the H8/300, H8/300H, and H8S to
3999 indicate that the specified function is an interrupt handler. The compiler
4000 generates function entry and exit sequences suitable for use in an
4001 interrupt handler when this attribute is present.
4004 @cindex @code{saveall} function attribute, H8/300
4005 @cindex save all registers on the H8/300, H8/300H, and H8S
4006 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
4007 all registers except the stack pointer should be saved in the prologue
4008 regardless of whether they are used or not.
4011 @node IA-64 Function Attributes
4012 @subsection IA-64 Function Attributes
4014 These function attributes are supported on IA-64 targets:
4017 @item syscall_linkage
4018 @cindex @code{syscall_linkage} function attribute, IA-64
4019 This attribute is used to modify the IA-64 calling convention by marking
4020 all input registers as live at all function exits. This makes it possible
4021 to restart a system call after an interrupt without having to save/restore
4022 the input registers. This also prevents kernel data from leaking into
4026 @cindex @code{version_id} function attribute, IA-64
4027 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4028 symbol to contain a version string, thus allowing for function level
4029 versioning. HP-UX system header files may use function level versioning
4030 for some system calls.
4033 extern int foo () __attribute__((version_id ("20040821")));
4037 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4040 @node M32C Function Attributes
4041 @subsection M32C Function Attributes
4043 These function attributes are supported by the M32C back end:
4047 @cindex @code{bank_switch} function attribute, M32C
4048 When added to an interrupt handler with the M32C port, causes the
4049 prologue and epilogue to use bank switching to preserve the registers
4050 rather than saving them on the stack.
4052 @item fast_interrupt
4053 @cindex @code{fast_interrupt} function attribute, M32C
4054 Use this attribute on the M32C port to indicate that the specified
4055 function is a fast interrupt handler. This is just like the
4056 @code{interrupt} attribute, except that @code{freit} is used to return
4057 instead of @code{reit}.
4059 @item function_vector
4060 @cindex @code{function_vector} function attribute, M16C/M32C
4061 On M16C/M32C targets, the @code{function_vector} attribute declares a
4062 special page subroutine call function. Use of this attribute reduces
4063 the code size by 2 bytes for each call generated to the
4064 subroutine. The argument to the attribute is the vector number entry
4065 from the special page vector table which contains the 16 low-order
4066 bits of the subroutine's entry address. Each vector table has special
4067 page number (18 to 255) that is used in @code{jsrs} instructions.
4068 Jump addresses of the routines are generated by adding 0x0F0000 (in
4069 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4070 2-byte addresses set in the vector table. Therefore you need to ensure
4071 that all the special page vector routines should get mapped within the
4072 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4075 In the following example 2 bytes are saved for each call to
4076 function @code{foo}.
4079 void foo (void) __attribute__((function_vector(0x18)));
4090 If functions are defined in one file and are called in another file,
4091 then be sure to write this declaration in both files.
4093 This attribute is ignored for R8C target.
4096 @cindex @code{interrupt} function attribute, M32C
4097 Use this attribute to indicate
4098 that the specified function is an interrupt handler. The compiler generates
4099 function entry and exit sequences suitable for use in an interrupt handler
4100 when this attribute is present.
4103 @node M32R/D Function Attributes
4104 @subsection M32R/D Function Attributes
4106 These function attributes are supported by the M32R/D back end:
4110 @cindex @code{interrupt} function attribute, M32R/D
4111 Use this attribute to indicate
4112 that the specified function is an interrupt handler. The compiler generates
4113 function entry and exit sequences suitable for use in an interrupt handler
4114 when this attribute is present.
4116 @item model (@var{model-name})
4117 @cindex @code{model} function attribute, M32R/D
4118 @cindex function addressability on the M32R/D
4120 On the M32R/D, use this attribute to set the addressability of an
4121 object, and of the code generated for a function. The identifier
4122 @var{model-name} is one of @code{small}, @code{medium}, or
4123 @code{large}, representing each of the code models.
4125 Small model objects live in the lower 16MB of memory (so that their
4126 addresses can be loaded with the @code{ld24} instruction), and are
4127 callable with the @code{bl} instruction.
4129 Medium model objects may live anywhere in the 32-bit address space (the
4130 compiler generates @code{seth/add3} instructions to load their addresses),
4131 and are callable with the @code{bl} instruction.
4133 Large model objects may live anywhere in the 32-bit address space (the
4134 compiler generates @code{seth/add3} instructions to load their addresses),
4135 and may not be reachable with the @code{bl} instruction (the compiler
4136 generates the much slower @code{seth/add3/jl} instruction sequence).
4139 @node m68k Function Attributes
4140 @subsection m68k Function Attributes
4142 These function attributes are supported by the m68k back end:
4146 @itemx interrupt_handler
4147 @cindex @code{interrupt} function attribute, m68k
4148 @cindex @code{interrupt_handler} function attribute, m68k
4149 Use this attribute to
4150 indicate that the specified function is an interrupt handler. The compiler
4151 generates function entry and exit sequences suitable for use in an
4152 interrupt handler when this attribute is present. Either name may be used.
4154 @item interrupt_thread
4155 @cindex @code{interrupt_thread} function attribute, fido
4156 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4157 that the specified function is an interrupt handler that is designed
4158 to run as a thread. The compiler omits generate prologue/epilogue
4159 sequences and replaces the return instruction with a @code{sleep}
4160 instruction. This attribute is available only on fido.
4163 @node MCORE Function Attributes
4164 @subsection MCORE Function Attributes
4166 These function attributes are supported by the MCORE back end:
4170 @cindex @code{naked} function attribute, MCORE
4171 This attribute allows the compiler to construct the
4172 requisite function declaration, while allowing the body of the
4173 function to be assembly code. The specified function will not have
4174 prologue/epilogue sequences generated by the compiler. Only basic
4175 @code{asm} statements can safely be included in naked functions
4176 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4177 basic @code{asm} and C code may appear to work, they cannot be
4178 depended upon to work reliably and are not supported.
4181 @node MeP Function Attributes
4182 @subsection MeP Function Attributes
4184 These function attributes are supported by the MeP back end:
4188 @cindex @code{disinterrupt} function attribute, MeP
4189 On MeP targets, this attribute causes the compiler to emit
4190 instructions to disable interrupts for the duration of the given
4194 @cindex @code{interrupt} function attribute, MeP
4195 Use this attribute to indicate
4196 that the specified function is an interrupt handler. The compiler generates
4197 function entry and exit sequences suitable for use in an interrupt handler
4198 when this attribute is present.
4201 @cindex @code{near} function attribute, MeP
4202 This attribute causes the compiler to assume the called
4203 function is close enough to use the normal calling convention,
4204 overriding the @option{-mtf} command-line option.
4207 @cindex @code{far} function attribute, MeP
4208 On MeP targets this causes the compiler to use a calling convention
4209 that assumes the called function is too far away for the built-in
4213 @cindex @code{vliw} function attribute, MeP
4214 The @code{vliw} attribute tells the compiler to emit
4215 instructions in VLIW mode instead of core mode. Note that this
4216 attribute is not allowed unless a VLIW coprocessor has been configured
4217 and enabled through command-line options.
4220 @node MicroBlaze Function Attributes
4221 @subsection MicroBlaze Function Attributes
4223 These function attributes are supported on MicroBlaze targets:
4226 @item save_volatiles
4227 @cindex @code{save_volatiles} function attribute, MicroBlaze
4228 Use this attribute to indicate that the function is
4229 an interrupt handler. All volatile registers (in addition to non-volatile
4230 registers) are saved in the function prologue. If the function is a leaf
4231 function, only volatiles used by the function are saved. A normal function
4232 return is generated instead of a return from interrupt.
4235 @cindex @code{break_handler} function attribute, MicroBlaze
4236 @cindex break handler functions
4237 Use this attribute to indicate that
4238 the specified function is a break handler. The compiler generates function
4239 entry and exit sequences suitable for use in an break handler when this
4240 attribute is present. The return from @code{break_handler} is done through
4241 the @code{rtbd} instead of @code{rtsd}.
4244 void f () __attribute__ ((break_handler));
4248 @node Microsoft Windows Function Attributes
4249 @subsection Microsoft Windows Function Attributes
4251 The following attributes are available on Microsoft Windows and Symbian OS
4256 @cindex @code{dllexport} function attribute
4257 @cindex @code{__declspec(dllexport)}
4258 On Microsoft Windows targets and Symbian OS targets the
4259 @code{dllexport} attribute causes the compiler to provide a global
4260 pointer to a pointer in a DLL, so that it can be referenced with the
4261 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4262 name is formed by combining @code{_imp__} and the function or variable
4265 You can use @code{__declspec(dllexport)} as a synonym for
4266 @code{__attribute__ ((dllexport))} for compatibility with other
4269 On systems that support the @code{visibility} attribute, this
4270 attribute also implies ``default'' visibility. It is an error to
4271 explicitly specify any other visibility.
4273 GCC's default behavior is to emit all inline functions with the
4274 @code{dllexport} attribute. Since this can cause object file-size bloat,
4275 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4276 ignore the attribute for inlined functions unless the
4277 @option{-fkeep-inline-functions} flag is used instead.
4279 The attribute is ignored for undefined symbols.
4281 When applied to C++ classes, the attribute marks defined non-inlined
4282 member functions and static data members as exports. Static consts
4283 initialized in-class are not marked unless they are also defined
4286 For Microsoft Windows targets there are alternative methods for
4287 including the symbol in the DLL's export table such as using a
4288 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4289 the @option{--export-all} linker flag.
4292 @cindex @code{dllimport} function attribute
4293 @cindex @code{__declspec(dllimport)}
4294 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4295 attribute causes the compiler to reference a function or variable via
4296 a global pointer to a pointer that is set up by the DLL exporting the
4297 symbol. The attribute implies @code{extern}. On Microsoft Windows
4298 targets, the pointer name is formed by combining @code{_imp__} and the
4299 function or variable name.
4301 You can use @code{__declspec(dllimport)} as a synonym for
4302 @code{__attribute__ ((dllimport))} for compatibility with other
4305 On systems that support the @code{visibility} attribute, this
4306 attribute also implies ``default'' visibility. It is an error to
4307 explicitly specify any other visibility.
4309 Currently, the attribute is ignored for inlined functions. If the
4310 attribute is applied to a symbol @emph{definition}, an error is reported.
4311 If a symbol previously declared @code{dllimport} is later defined, the
4312 attribute is ignored in subsequent references, and a warning is emitted.
4313 The attribute is also overridden by a subsequent declaration as
4316 When applied to C++ classes, the attribute marks non-inlined
4317 member functions and static data members as imports. However, the
4318 attribute is ignored for virtual methods to allow creation of vtables
4321 On the SH Symbian OS target the @code{dllimport} attribute also has
4322 another affect---it can cause the vtable and run-time type information
4323 for a class to be exported. This happens when the class has a
4324 dllimported constructor or a non-inline, non-pure virtual function
4325 and, for either of those two conditions, the class also has an inline
4326 constructor or destructor and has a key function that is defined in
4327 the current translation unit.
4329 For Microsoft Windows targets the use of the @code{dllimport}
4330 attribute on functions is not necessary, but provides a small
4331 performance benefit by eliminating a thunk in the DLL@. The use of the
4332 @code{dllimport} attribute on imported variables can be avoided by passing the
4333 @option{--enable-auto-import} switch to the GNU linker. As with
4334 functions, using the attribute for a variable eliminates a thunk in
4337 One drawback to using this attribute is that a pointer to a
4338 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4339 address. However, a pointer to a @emph{function} with the
4340 @code{dllimport} attribute can be used as a constant initializer; in
4341 this case, the address of a stub function in the import lib is
4342 referenced. On Microsoft Windows targets, the attribute can be disabled
4343 for functions by setting the @option{-mnop-fun-dllimport} flag.
4346 @node MIPS Function Attributes
4347 @subsection MIPS Function Attributes
4349 These function attributes are supported by the MIPS back end:
4353 @cindex @code{interrupt} function attribute, MIPS
4354 Use this attribute to indicate that the specified function is an interrupt
4355 handler. The compiler generates function entry and exit sequences suitable
4356 for use in an interrupt handler when this attribute is present.
4357 An optional argument is supported for the interrupt attribute which allows
4358 the interrupt mode to be described. By default GCC assumes the external
4359 interrupt controller (EIC) mode is in use, this can be explicitly set using
4360 @code{eic}. When interrupts are non-masked then the requested Interrupt
4361 Priority Level (IPL) is copied to the current IPL which has the effect of only
4362 enabling higher priority interrupts. To use vectored interrupt mode use
4363 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4364 the behaviour of the non-masked interrupt support and GCC will arrange to mask
4365 all interrupts from sw0 up to and including the specified interrupt vector.
4367 You can use the following attributes to modify the behavior
4368 of an interrupt handler:
4370 @item use_shadow_register_set
4371 @cindex @code{use_shadow_register_set} function attribute, MIPS
4372 Assume that the handler uses a shadow register set, instead of
4373 the main general-purpose registers. An optional argument @code{intstack} is
4374 supported to indicate that the shadow register set contains a valid stack
4377 @item keep_interrupts_masked
4378 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4379 Keep interrupts masked for the whole function. Without this attribute,
4380 GCC tries to reenable interrupts for as much of the function as it can.
4382 @item use_debug_exception_return
4383 @cindex @code{use_debug_exception_return} function attribute, MIPS
4384 Return using the @code{deret} instruction. Interrupt handlers that don't
4385 have this attribute return using @code{eret} instead.
4388 You can use any combination of these attributes, as shown below:
4390 void __attribute__ ((interrupt)) v0 ();
4391 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4392 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4393 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4394 void __attribute__ ((interrupt, use_shadow_register_set,
4395 keep_interrupts_masked)) v4 ();
4396 void __attribute__ ((interrupt, use_shadow_register_set,
4397 use_debug_exception_return)) v5 ();
4398 void __attribute__ ((interrupt, keep_interrupts_masked,
4399 use_debug_exception_return)) v6 ();
4400 void __attribute__ ((interrupt, use_shadow_register_set,
4401 keep_interrupts_masked,
4402 use_debug_exception_return)) v7 ();
4403 void __attribute__ ((interrupt("eic"))) v8 ();
4404 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4410 @cindex indirect calls, MIPS
4411 @cindex @code{long_call} function attribute, MIPS
4412 @cindex @code{near} function attribute, MIPS
4413 @cindex @code{far} function attribute, MIPS
4414 These attributes specify how a particular function is called on MIPS@.
4415 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4416 command-line switch. The @code{long_call} and @code{far} attributes are
4417 synonyms, and cause the compiler to always call
4418 the function by first loading its address into a register, and then using
4419 the contents of that register. The @code{near} attribute has the opposite
4420 effect; it specifies that non-PIC calls should be made using the more
4421 efficient @code{jal} instruction.
4425 @cindex @code{mips16} function attribute, MIPS
4426 @cindex @code{nomips16} function attribute, MIPS
4428 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4429 function attributes to locally select or turn off MIPS16 code generation.
4430 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4431 while MIPS16 code generation is disabled for functions with the
4432 @code{nomips16} attribute. These attributes override the
4433 @option{-mips16} and @option{-mno-mips16} options on the command line
4434 (@pxref{MIPS Options}).
4436 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4437 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4438 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4439 may interact badly with some GCC extensions such as @code{__builtin_apply}
4440 (@pxref{Constructing Calls}).
4442 @item micromips, MIPS
4443 @itemx nomicromips, MIPS
4444 @cindex @code{micromips} function attribute
4445 @cindex @code{nomicromips} function attribute
4447 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4448 function attributes to locally select or turn off microMIPS code generation.
4449 A function with the @code{micromips} attribute is emitted as microMIPS code,
4450 while microMIPS code generation is disabled for functions with the
4451 @code{nomicromips} attribute. These attributes override the
4452 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4453 (@pxref{MIPS Options}).
4455 When compiling files containing mixed microMIPS and non-microMIPS code, the
4456 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4458 not that within individual functions. Mixed microMIPS and non-microMIPS code
4459 may interact badly with some GCC extensions such as @code{__builtin_apply}
4460 (@pxref{Constructing Calls}).
4463 @cindex @code{nocompression} function attribute, MIPS
4464 On MIPS targets, you can use the @code{nocompression} function attribute
4465 to locally turn off MIPS16 and microMIPS code generation. This attribute
4466 overrides the @option{-mips16} and @option{-mmicromips} options on the
4467 command line (@pxref{MIPS Options}).
4470 @node MSP430 Function Attributes
4471 @subsection MSP430 Function Attributes
4473 These function attributes are supported by the MSP430 back end:
4477 @cindex @code{critical} function attribute, MSP430
4478 Critical functions disable interrupts upon entry and restore the
4479 previous interrupt state upon exit. Critical functions cannot also
4480 have the @code{naked} or @code{reentrant} attributes. They can have
4481 the @code{interrupt} attribute.
4484 @cindex @code{interrupt} function attribute, MSP430
4485 Use this attribute to indicate
4486 that the specified function is an interrupt handler. The compiler generates
4487 function entry and exit sequences suitable for use in an interrupt handler
4488 when this attribute is present.
4490 You can provide an argument to the interrupt
4491 attribute which specifies a name or number. If the argument is a
4492 number it indicates the slot in the interrupt vector table (0 - 31) to
4493 which this handler should be assigned. If the argument is a name it
4494 is treated as a symbolic name for the vector slot. These names should
4495 match up with appropriate entries in the linker script. By default
4496 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4497 @code{reset} for vector 31 are recognized.
4500 @cindex @code{naked} function attribute, MSP430
4501 This attribute allows the compiler to construct the
4502 requisite function declaration, while allowing the body of the
4503 function to be assembly code. The specified function will not have
4504 prologue/epilogue sequences generated by the compiler. Only basic
4505 @code{asm} statements can safely be included in naked functions
4506 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4507 basic @code{asm} and C code may appear to work, they cannot be
4508 depended upon to work reliably and are not supported.
4511 @cindex @code{reentrant} function attribute, MSP430
4512 Reentrant functions disable interrupts upon entry and enable them
4513 upon exit. Reentrant functions cannot also have the @code{naked}
4514 or @code{critical} attributes. They can have the @code{interrupt}
4518 @cindex @code{wakeup} function attribute, MSP430
4519 This attribute only applies to interrupt functions. It is silently
4520 ignored if applied to a non-interrupt function. A wakeup interrupt
4521 function will rouse the processor from any low-power state that it
4522 might be in when the function exits.
4525 @node NDS32 Function Attributes
4526 @subsection NDS32 Function Attributes
4528 These function attributes are supported by the NDS32 back end:
4532 @cindex @code{exception} function attribute
4533 @cindex exception handler functions, NDS32
4534 Use this attribute on the NDS32 target to indicate that the specified function
4535 is an exception handler. The compiler will generate corresponding sections
4536 for use in an exception handler.
4539 @cindex @code{interrupt} function attribute, NDS32
4540 On NDS32 target, this attribute indicates that the specified function
4541 is an interrupt handler. The compiler generates corresponding sections
4542 for use in an interrupt handler. You can use the following attributes
4543 to modify the behavior:
4546 @cindex @code{nested} function attribute, NDS32
4547 This interrupt service routine is interruptible.
4549 @cindex @code{not_nested} function attribute, NDS32
4550 This interrupt service routine is not interruptible.
4552 @cindex @code{nested_ready} function attribute, NDS32
4553 This interrupt service routine is interruptible after @code{PSW.GIE}
4554 (global interrupt enable) is set. This allows interrupt service routine to
4555 finish some short critical code before enabling interrupts.
4557 @cindex @code{save_all} function attribute, NDS32
4558 The system will help save all registers into stack before entering
4561 @cindex @code{partial_save} function attribute, NDS32
4562 The system will help save caller registers into stack before entering
4567 @cindex @code{naked} function attribute, NDS32
4568 This attribute allows the compiler to construct the
4569 requisite function declaration, while allowing the body of the
4570 function to be assembly code. The specified function will not have
4571 prologue/epilogue sequences generated by the compiler. Only basic
4572 @code{asm} statements can safely be included in naked functions
4573 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4574 basic @code{asm} and C code may appear to work, they cannot be
4575 depended upon to work reliably and are not supported.
4578 @cindex @code{reset} function attribute, NDS32
4579 @cindex reset handler functions
4580 Use this attribute on the NDS32 target to indicate that the specified function
4581 is a reset handler. The compiler will generate corresponding sections
4582 for use in a reset handler. You can use the following attributes
4583 to provide extra exception handling:
4586 @cindex @code{nmi} function attribute, NDS32
4587 Provide a user-defined function to handle NMI exception.
4589 @cindex @code{warm} function attribute, NDS32
4590 Provide a user-defined function to handle warm reset exception.
4594 @node Nios II Function Attributes
4595 @subsection Nios II Function Attributes
4597 These function attributes are supported by the Nios II back end:
4600 @item target (@var{options})
4601 @cindex @code{target} function attribute
4602 As discussed in @ref{Common Function Attributes}, this attribute
4603 allows specification of target-specific compilation options.
4605 When compiling for Nios II, the following options are allowed:
4608 @item custom-@var{insn}=@var{N}
4609 @itemx no-custom-@var{insn}
4610 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4611 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4612 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4613 custom instruction with encoding @var{N} when generating code that uses
4614 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4615 the custom instruction @var{insn}.
4616 These target attributes correspond to the
4617 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4618 command-line options, and support the same set of @var{insn} keywords.
4619 @xref{Nios II Options}, for more information.
4621 @item custom-fpu-cfg=@var{name}
4622 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4623 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4624 command-line option, to select a predefined set of custom instructions
4626 @xref{Nios II Options}, for more information.
4630 @node PowerPC Function Attributes
4631 @subsection PowerPC Function Attributes
4633 These function attributes are supported by the PowerPC back end:
4638 @cindex indirect calls, PowerPC
4639 @cindex @code{longcall} function attribute, PowerPC
4640 @cindex @code{shortcall} function attribute, PowerPC
4641 The @code{longcall} attribute
4642 indicates that the function might be far away from the call site and
4643 require a different (more expensive) calling sequence. The
4644 @code{shortcall} attribute indicates that the function is always close
4645 enough for the shorter calling sequence to be used. These attributes
4646 override both the @option{-mlongcall} switch and
4647 the @code{#pragma longcall} setting.
4649 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4650 calls are necessary.
4652 @item target (@var{options})
4653 @cindex @code{target} function attribute
4654 As discussed in @ref{Common Function Attributes}, this attribute
4655 allows specification of target-specific compilation options.
4657 On the PowerPC, the following options are allowed:
4662 @cindex @code{target("altivec")} function attribute, PowerPC
4663 Generate code that uses (does not use) AltiVec instructions. In
4664 32-bit code, you cannot enable AltiVec instructions unless
4665 @option{-mabi=altivec} is used on the command line.
4669 @cindex @code{target("cmpb")} function attribute, PowerPC
4670 Generate code that uses (does not use) the compare bytes instruction
4671 implemented on the POWER6 processor and other processors that support
4672 the PowerPC V2.05 architecture.
4676 @cindex @code{target("dlmzb")} function attribute, PowerPC
4677 Generate code that uses (does not use) the string-search @samp{dlmzb}
4678 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4679 generated by default when targeting those processors.
4683 @cindex @code{target("fprnd")} function attribute, PowerPC
4684 Generate code that uses (does not use) the FP round to integer
4685 instructions implemented on the POWER5+ processor and other processors
4686 that support the PowerPC V2.03 architecture.
4690 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4691 Generate code that uses (does not use) the decimal floating-point
4692 instructions implemented on some POWER processors.
4696 @cindex @code{target("isel")} function attribute, PowerPC
4697 Generate code that uses (does not use) ISEL instruction.
4701 @cindex @code{target("mfcrf")} function attribute, PowerPC
4702 Generate code that uses (does not use) the move from condition
4703 register field instruction implemented on the POWER4 processor and
4704 other processors that support the PowerPC V2.01 architecture.
4708 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4709 Generate code that uses (does not use) the FP move to/from general
4710 purpose register instructions implemented on the POWER6X processor and
4711 other processors that support the extended PowerPC V2.05 architecture.
4715 @cindex @code{target("mulhw")} function attribute, PowerPC
4716 Generate code that uses (does not use) the half-word multiply and
4717 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4718 These instructions are generated by default when targeting those
4723 @cindex @code{target("multiple")} function attribute, PowerPC
4724 Generate code that uses (does not use) the load multiple word
4725 instructions and the store multiple word instructions.
4729 @cindex @code{target("update")} function attribute, PowerPC
4730 Generate code that uses (does not use) the load or store instructions
4731 that update the base register to the address of the calculated memory
4736 @cindex @code{target("popcntb")} function attribute, PowerPC
4737 Generate code that uses (does not use) the popcount and double-precision
4738 FP reciprocal estimate instruction implemented on the POWER5
4739 processor and other processors that support the PowerPC V2.02
4744 @cindex @code{target("popcntd")} function attribute, PowerPC
4745 Generate code that uses (does not use) the popcount instruction
4746 implemented on the POWER7 processor and other processors that support
4747 the PowerPC V2.06 architecture.
4749 @item powerpc-gfxopt
4750 @itemx no-powerpc-gfxopt
4751 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4752 Generate code that uses (does not use) the optional PowerPC
4753 architecture instructions in the Graphics group, including
4754 floating-point select.
4757 @itemx no-powerpc-gpopt
4758 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4759 Generate code that uses (does not use) the optional PowerPC
4760 architecture instructions in the General Purpose group, including
4761 floating-point square root.
4763 @item recip-precision
4764 @itemx no-recip-precision
4765 @cindex @code{target("recip-precision")} function attribute, PowerPC
4766 Assume (do not assume) that the reciprocal estimate instructions
4767 provide higher-precision estimates than is mandated by the PowerPC
4772 @cindex @code{target("string")} function attribute, PowerPC
4773 Generate code that uses (does not use) the load string instructions
4774 and the store string word instructions to save multiple registers and
4775 do small block moves.
4779 @cindex @code{target("vsx")} function attribute, PowerPC
4780 Generate code that uses (does not use) vector/scalar (VSX)
4781 instructions, and also enable the use of built-in functions that allow
4782 more direct access to the VSX instruction set. In 32-bit code, you
4783 cannot enable VSX or AltiVec instructions unless
4784 @option{-mabi=altivec} is used on the command line.
4788 @cindex @code{target("friz")} function attribute, PowerPC
4789 Generate (do not generate) the @code{friz} instruction when the
4790 @option{-funsafe-math-optimizations} option is used to optimize
4791 rounding a floating-point value to 64-bit integer and back to floating
4792 point. The @code{friz} instruction does not return the same value if
4793 the floating-point number is too large to fit in an integer.
4795 @item avoid-indexed-addresses
4796 @itemx no-avoid-indexed-addresses
4797 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4798 Generate code that tries to avoid (not avoid) the use of indexed load
4799 or store instructions.
4803 @cindex @code{target("paired")} function attribute, PowerPC
4804 Generate code that uses (does not use) the generation of PAIRED simd
4809 @cindex @code{target("longcall")} function attribute, PowerPC
4810 Generate code that assumes (does not assume) that all calls are far
4811 away so that a longer more expensive calling sequence is required.
4814 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4815 Specify the architecture to generate code for when compiling the
4816 function. If you select the @code{target("cpu=power7")} attribute when
4817 generating 32-bit code, VSX and AltiVec instructions are not generated
4818 unless you use the @option{-mabi=altivec} option on the command line.
4820 @item tune=@var{TUNE}
4821 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4822 Specify the architecture to tune for when compiling the function. If
4823 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4824 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4825 compilation tunes for the @var{CPU} architecture, and not the
4826 default tuning specified on the command line.
4829 On the PowerPC, the inliner does not inline a
4830 function that has different target options than the caller, unless the
4831 callee has a subset of the target options of the caller.
4834 @node RL78 Function Attributes
4835 @subsection RL78 Function Attributes
4837 These function attributes are supported by the RL78 back end:
4841 @itemx brk_interrupt
4842 @cindex @code{interrupt} function attribute, RL78
4843 @cindex @code{brk_interrupt} function attribute, RL78
4844 These attributes indicate
4845 that the specified function is an interrupt handler. The compiler generates
4846 function entry and exit sequences suitable for use in an interrupt handler
4847 when this attribute is present.
4849 Use @code{brk_interrupt} instead of @code{interrupt} for
4850 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4851 that must end with @code{RETB} instead of @code{RETI}).
4854 @cindex @code{naked} function attribute, RL78
4855 This attribute allows the compiler to construct the
4856 requisite function declaration, while allowing the body of the
4857 function to be assembly code. The specified function will not have
4858 prologue/epilogue sequences generated by the compiler. Only basic
4859 @code{asm} statements can safely be included in naked functions
4860 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4861 basic @code{asm} and C code may appear to work, they cannot be
4862 depended upon to work reliably and are not supported.
4865 @node RX Function Attributes
4866 @subsection RX Function Attributes
4868 These function attributes are supported by the RX back end:
4871 @item fast_interrupt
4872 @cindex @code{fast_interrupt} function attribute, RX
4873 Use this attribute on the RX port to indicate that the specified
4874 function is a fast interrupt handler. This is just like the
4875 @code{interrupt} attribute, except that @code{freit} is used to return
4876 instead of @code{reit}.
4879 @cindex @code{interrupt} function attribute, RX
4880 Use this attribute to indicate
4881 that the specified function is an interrupt handler. The compiler generates
4882 function entry and exit sequences suitable for use in an interrupt handler
4883 when this attribute is present.
4885 On RX targets, you may specify one or more vector numbers as arguments
4886 to the attribute, as well as naming an alternate table name.
4887 Parameters are handled sequentially, so one handler can be assigned to
4888 multiple entries in multiple tables. One may also pass the magic
4889 string @code{"$default"} which causes the function to be used for any
4890 unfilled slots in the current table.
4892 This example shows a simple assignment of a function to one vector in
4893 the default table (note that preprocessor macros may be used for
4894 chip-specific symbolic vector names):
4896 void __attribute__ ((interrupt (5))) txd1_handler ();
4899 This example assigns a function to two slots in the default table
4900 (using preprocessor macros defined elsewhere) and makes it the default
4901 for the @code{dct} table:
4903 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4908 @cindex @code{naked} function attribute, RX
4909 This attribute allows the compiler to construct the
4910 requisite function declaration, while allowing the body of the
4911 function to be assembly code. The specified function will not have
4912 prologue/epilogue sequences generated by the compiler. Only basic
4913 @code{asm} statements can safely be included in naked functions
4914 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4915 basic @code{asm} and C code may appear to work, they cannot be
4916 depended upon to work reliably and are not supported.
4919 @cindex @code{vector} function attribute, RX
4920 This RX attribute is similar to the @code{interrupt} attribute, including its
4921 parameters, but does not make the function an interrupt-handler type
4922 function (i.e. it retains the normal C function calling ABI). See the
4923 @code{interrupt} attribute for a description of its arguments.
4926 @node S/390 Function Attributes
4927 @subsection S/390 Function Attributes
4929 These function attributes are supported on the S/390:
4932 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4933 @cindex @code{hotpatch} function attribute, S/390
4935 On S/390 System z targets, you can use this function attribute to
4936 make GCC generate a ``hot-patching'' function prologue. If the
4937 @option{-mhotpatch=} command-line option is used at the same time,
4938 the @code{hotpatch} attribute takes precedence. The first of the
4939 two arguments specifies the number of halfwords to be added before
4940 the function label. A second argument can be used to specify the
4941 number of halfwords to be added after the function label. For
4942 both arguments the maximum allowed value is 1000000.
4944 If both arguments are zero, hotpatching is disabled.
4946 @item target (@var{options})
4947 @cindex @code{target} function attribute
4948 As discussed in @ref{Common Function Attributes}, this attribute
4949 allows specification of target-specific compilation options.
4951 On S/390, the following options are supported:
4959 @item warn-framesize=
4971 @itemx no-packed-stack
4973 @itemx no-small-exec
4976 @item warn-dynamicstack
4977 @itemx no-warn-dynamicstack
4980 The options work exactly like the S/390 specific command line
4981 options (without the prefix @option{-m}) except that they do not
4982 change any feature macros. For example,
4985 @code{target("no-vx")}
4988 does not undefine the @code{__VEC__} macro.
4991 @node SH Function Attributes
4992 @subsection SH Function Attributes
4994 These function attributes are supported on the SH family of processors:
4997 @item function_vector
4998 @cindex @code{function_vector} function attribute, SH
4999 @cindex calling functions through the function vector on SH2A
5000 On SH2A targets, this attribute declares a function to be called using the
5001 TBR relative addressing mode. The argument to this attribute is the entry
5002 number of the same function in a vector table containing all the TBR
5003 relative addressable functions. For correct operation the TBR must be setup
5004 accordingly to point to the start of the vector table before any functions with
5005 this attribute are invoked. Usually a good place to do the initialization is
5006 the startup routine. The TBR relative vector table can have at max 256 function
5007 entries. The jumps to these functions are generated using a SH2A specific,
5008 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5009 from GNU binutils version 2.7 or later for this attribute to work correctly.
5011 In an application, for a function being called once, this attribute
5012 saves at least 8 bytes of code; and if other successive calls are being
5013 made to the same function, it saves 2 bytes of code per each of these
5016 @item interrupt_handler
5017 @cindex @code{interrupt_handler} function attribute, SH
5018 Use this attribute to
5019 indicate that the specified function is an interrupt handler. The compiler
5020 generates function entry and exit sequences suitable for use in an
5021 interrupt handler when this attribute is present.
5023 @item nosave_low_regs
5024 @cindex @code{nosave_low_regs} function attribute, SH
5025 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5026 function should not save and restore registers R0..R7. This can be used on SH3*
5027 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5031 @cindex @code{renesas} function attribute, SH
5032 On SH targets this attribute specifies that the function or struct follows the
5036 @cindex @code{resbank} function attribute, SH
5037 On the SH2A target, this attribute enables the high-speed register
5038 saving and restoration using a register bank for @code{interrupt_handler}
5039 routines. Saving to the bank is performed automatically after the CPU
5040 accepts an interrupt that uses a register bank.
5042 The nineteen 32-bit registers comprising general register R0 to R14,
5043 control register GBR, and system registers MACH, MACL, and PR and the
5044 vector table address offset are saved into a register bank. Register
5045 banks are stacked in first-in last-out (FILO) sequence. Restoration
5046 from the bank is executed by issuing a RESBANK instruction.
5049 @cindex @code{sp_switch} function attribute, SH
5050 Use this attribute on the SH to indicate an @code{interrupt_handler}
5051 function should switch to an alternate stack. It expects a string
5052 argument that names a global variable holding the address of the
5057 void f () __attribute__ ((interrupt_handler,
5058 sp_switch ("alt_stack")));
5062 @cindex @code{trap_exit} function attribute, SH
5063 Use this attribute on the SH for an @code{interrupt_handler} to return using
5064 @code{trapa} instead of @code{rte}. This attribute expects an integer
5065 argument specifying the trap number to be used.
5068 @cindex @code{trapa_handler} function attribute, SH
5069 On SH targets this function attribute is similar to @code{interrupt_handler}
5070 but it does not save and restore all registers.
5073 @node SPU Function Attributes
5074 @subsection SPU Function Attributes
5076 These function attributes are supported by the SPU back end:
5080 @cindex @code{naked} function attribute, SPU
5081 This attribute allows the compiler to construct the
5082 requisite function declaration, while allowing the body of the
5083 function to be assembly code. The specified function will not have
5084 prologue/epilogue sequences generated by the compiler. Only basic
5085 @code{asm} statements can safely be included in naked functions
5086 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5087 basic @code{asm} and C code may appear to work, they cannot be
5088 depended upon to work reliably and are not supported.
5091 @node Symbian OS Function Attributes
5092 @subsection Symbian OS Function Attributes
5094 @xref{Microsoft Windows Function Attributes}, for discussion of the
5095 @code{dllexport} and @code{dllimport} attributes.
5097 @node Visium Function Attributes
5098 @subsection Visium Function Attributes
5100 These function attributes are supported by the Visium back end:
5104 @cindex @code{interrupt} function attribute, Visium
5105 Use this attribute to indicate
5106 that the specified function is an interrupt handler. The compiler generates
5107 function entry and exit sequences suitable for use in an interrupt handler
5108 when this attribute is present.
5111 @node x86 Function Attributes
5112 @subsection x86 Function Attributes
5114 These function attributes are supported by the x86 back end:
5118 @cindex @code{cdecl} function attribute, x86-32
5119 @cindex functions that pop the argument stack on x86-32
5121 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5122 assume that the calling function pops off the stack space used to
5123 pass arguments. This is
5124 useful to override the effects of the @option{-mrtd} switch.
5127 @cindex @code{fastcall} function attribute, x86-32
5128 @cindex functions that pop the argument stack on x86-32
5129 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5130 pass the first argument (if of integral type) in the register ECX and
5131 the second argument (if of integral type) in the register EDX@. Subsequent
5132 and other typed arguments are passed on the stack. The called function
5133 pops the arguments off the stack. If the number of arguments is variable all
5134 arguments are pushed on the stack.
5137 @cindex @code{thiscall} function attribute, x86-32
5138 @cindex functions that pop the argument stack on x86-32
5139 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5140 pass the first argument (if of integral type) in the register ECX.
5141 Subsequent and other typed arguments are passed on the stack. The called
5142 function pops the arguments off the stack.
5143 If the number of arguments is variable all arguments are pushed on the
5145 The @code{thiscall} attribute is intended for C++ non-static member functions.
5146 As a GCC extension, this calling convention can be used for C functions
5147 and for static member methods.
5151 @cindex @code{ms_abi} function attribute, x86
5152 @cindex @code{sysv_abi} function attribute, x86
5154 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5155 to indicate which calling convention should be used for a function. The
5156 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5157 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5158 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5159 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5161 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5162 requires the @option{-maccumulate-outgoing-args} option.
5164 @item callee_pop_aggregate_return (@var{number})
5165 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5167 On x86-32 targets, you can use this attribute to control how
5168 aggregates are returned in memory. If the caller is responsible for
5169 popping the hidden pointer together with the rest of the arguments, specify
5170 @var{number} equal to zero. If callee is responsible for popping the
5171 hidden pointer, specify @var{number} equal to one.
5173 The default x86-32 ABI assumes that the callee pops the
5174 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5175 the compiler assumes that the
5176 caller pops the stack for hidden pointer.
5178 @item ms_hook_prologue
5179 @cindex @code{ms_hook_prologue} function attribute, x86
5181 On 32-bit and 64-bit x86 targets, you can use
5182 this function attribute to make GCC generate the ``hot-patching'' function
5183 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5186 @item regparm (@var{number})
5187 @cindex @code{regparm} function attribute, x86
5188 @cindex functions that are passed arguments in registers on x86-32
5189 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5190 pass arguments number one to @var{number} if they are of integral type
5191 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5192 take a variable number of arguments continue to be passed all of their
5193 arguments on the stack.
5195 Beware that on some ELF systems this attribute is unsuitable for
5196 global functions in shared libraries with lazy binding (which is the
5197 default). Lazy binding sends the first call via resolving code in
5198 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5199 per the standard calling conventions. Solaris 8 is affected by this.
5200 Systems with the GNU C Library version 2.1 or higher
5201 and FreeBSD are believed to be
5202 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5203 disabled with the linker or the loader if desired, to avoid the
5207 @cindex @code{sseregparm} function attribute, x86
5208 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5209 causes the compiler to pass up to 3 floating-point arguments in
5210 SSE registers instead of on the stack. Functions that take a
5211 variable number of arguments continue to pass all of their
5212 floating-point arguments on the stack.
5214 @item force_align_arg_pointer
5215 @cindex @code{force_align_arg_pointer} function attribute, x86
5216 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5217 applied to individual function definitions, generating an alternate
5218 prologue and epilogue that realigns the run-time stack if necessary.
5219 This supports mixing legacy codes that run with a 4-byte aligned stack
5220 with modern codes that keep a 16-byte stack for SSE compatibility.
5223 @cindex @code{stdcall} function attribute, x86-32
5224 @cindex functions that pop the argument stack on x86-32
5225 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5226 assume that the called function pops off the stack space used to
5227 pass arguments, unless it takes a variable number of arguments.
5229 @item target (@var{options})
5230 @cindex @code{target} function attribute
5231 As discussed in @ref{Common Function Attributes}, this attribute
5232 allows specification of target-specific compilation options.
5234 On the x86, the following options are allowed:
5238 @cindex @code{target("abm")} function attribute, x86
5239 Enable/disable the generation of the advanced bit instructions.
5243 @cindex @code{target("aes")} function attribute, x86
5244 Enable/disable the generation of the AES instructions.
5247 @cindex @code{target("default")} function attribute, x86
5248 @xref{Function Multiversioning}, where it is used to specify the
5249 default function version.
5253 @cindex @code{target("mmx")} function attribute, x86
5254 Enable/disable the generation of the MMX instructions.
5258 @cindex @code{target("pclmul")} function attribute, x86
5259 Enable/disable the generation of the PCLMUL instructions.
5263 @cindex @code{target("popcnt")} function attribute, x86
5264 Enable/disable the generation of the POPCNT instruction.
5268 @cindex @code{target("sse")} function attribute, x86
5269 Enable/disable the generation of the SSE instructions.
5273 @cindex @code{target("sse2")} function attribute, x86
5274 Enable/disable the generation of the SSE2 instructions.
5278 @cindex @code{target("sse3")} function attribute, x86
5279 Enable/disable the generation of the SSE3 instructions.
5283 @cindex @code{target("sse4")} function attribute, x86
5284 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5289 @cindex @code{target("sse4.1")} function attribute, x86
5290 Enable/disable the generation of the sse4.1 instructions.
5294 @cindex @code{target("sse4.2")} function attribute, x86
5295 Enable/disable the generation of the sse4.2 instructions.
5299 @cindex @code{target("sse4a")} function attribute, x86
5300 Enable/disable the generation of the SSE4A instructions.
5304 @cindex @code{target("fma4")} function attribute, x86
5305 Enable/disable the generation of the FMA4 instructions.
5309 @cindex @code{target("xop")} function attribute, x86
5310 Enable/disable the generation of the XOP instructions.
5314 @cindex @code{target("lwp")} function attribute, x86
5315 Enable/disable the generation of the LWP instructions.
5319 @cindex @code{target("ssse3")} function attribute, x86
5320 Enable/disable the generation of the SSSE3 instructions.
5324 @cindex @code{target("cld")} function attribute, x86
5325 Enable/disable the generation of the CLD before string moves.
5327 @item fancy-math-387
5328 @itemx no-fancy-math-387
5329 @cindex @code{target("fancy-math-387")} function attribute, x86
5330 Enable/disable the generation of the @code{sin}, @code{cos}, and
5331 @code{sqrt} instructions on the 387 floating-point unit.
5334 @itemx no-fused-madd
5335 @cindex @code{target("fused-madd")} function attribute, x86
5336 Enable/disable the generation of the fused multiply/add instructions.
5340 @cindex @code{target("ieee-fp")} function attribute, x86
5341 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5343 @item inline-all-stringops
5344 @itemx no-inline-all-stringops
5345 @cindex @code{target("inline-all-stringops")} function attribute, x86
5346 Enable/disable inlining of string operations.
5348 @item inline-stringops-dynamically
5349 @itemx no-inline-stringops-dynamically
5350 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5351 Enable/disable the generation of the inline code to do small string
5352 operations and calling the library routines for large operations.
5354 @item align-stringops
5355 @itemx no-align-stringops
5356 @cindex @code{target("align-stringops")} function attribute, x86
5357 Do/do not align destination of inlined string operations.
5361 @cindex @code{target("recip")} function attribute, x86
5362 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5363 instructions followed an additional Newton-Raphson step instead of
5364 doing a floating-point division.
5366 @item arch=@var{ARCH}
5367 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5368 Specify the architecture to generate code for in compiling the function.
5370 @item tune=@var{TUNE}
5371 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5372 Specify the architecture to tune for in compiling the function.
5374 @item fpmath=@var{FPMATH}
5375 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5376 Specify which floating-point unit to use. You must specify the
5377 @code{target("fpmath=sse,387")} option as
5378 @code{target("fpmath=sse+387")} because the comma would separate
5382 On the x86, the inliner does not inline a
5383 function that has different target options than the caller, unless the
5384 callee has a subset of the target options of the caller. For example
5385 a function declared with @code{target("sse3")} can inline a function
5386 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5389 @node Xstormy16 Function Attributes
5390 @subsection Xstormy16 Function Attributes
5392 These function attributes are supported by the Xstormy16 back end:
5396 @cindex @code{interrupt} function attribute, Xstormy16
5397 Use this attribute to indicate
5398 that the specified function is an interrupt handler. The compiler generates
5399 function entry and exit sequences suitable for use in an interrupt handler
5400 when this attribute is present.
5403 @node Variable Attributes
5404 @section Specifying Attributes of Variables
5405 @cindex attribute of variables
5406 @cindex variable attributes
5408 The keyword @code{__attribute__} allows you to specify special
5409 attributes of variables or structure fields. This keyword is followed
5410 by an attribute specification inside double parentheses. Some
5411 attributes are currently defined generically for variables.
5412 Other attributes are defined for variables on particular target
5413 systems. Other attributes are available for functions
5414 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5415 enumerators (@pxref{Enumerator Attributes}), and for types
5416 (@pxref{Type Attributes}).
5417 Other front ends might define more attributes
5418 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5420 @xref{Attribute Syntax}, for details of the exact syntax for using
5424 * Common Variable Attributes::
5425 * AVR Variable Attributes::
5426 * Blackfin Variable Attributes::
5427 * H8/300 Variable Attributes::
5428 * IA-64 Variable Attributes::
5429 * M32R/D Variable Attributes::
5430 * MeP Variable Attributes::
5431 * Microsoft Windows Variable Attributes::
5432 * MSP430 Variable Attributes::
5433 * PowerPC Variable Attributes::
5434 * SPU Variable Attributes::
5435 * x86 Variable Attributes::
5436 * Xstormy16 Variable Attributes::
5439 @node Common Variable Attributes
5440 @subsection Common Variable Attributes
5442 The following attributes are supported on most targets.
5445 @cindex @code{aligned} variable attribute
5446 @item aligned (@var{alignment})
5447 This attribute specifies a minimum alignment for the variable or
5448 structure field, measured in bytes. For example, the declaration:
5451 int x __attribute__ ((aligned (16))) = 0;
5455 causes the compiler to allocate the global variable @code{x} on a
5456 16-byte boundary. On a 68040, this could be used in conjunction with
5457 an @code{asm} expression to access the @code{move16} instruction which
5458 requires 16-byte aligned operands.
5460 You can also specify the alignment of structure fields. For example, to
5461 create a double-word aligned @code{int} pair, you could write:
5464 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5468 This is an alternative to creating a union with a @code{double} member,
5469 which forces the union to be double-word aligned.
5471 As in the preceding examples, you can explicitly specify the alignment
5472 (in bytes) that you wish the compiler to use for a given variable or
5473 structure field. Alternatively, you can leave out the alignment factor
5474 and just ask the compiler to align a variable or field to the
5475 default alignment for the target architecture you are compiling for.
5476 The default alignment is sufficient for all scalar types, but may not be
5477 enough for all vector types on a target that supports vector operations.
5478 The default alignment is fixed for a particular target ABI.
5480 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5481 which is the largest alignment ever used for any data type on the
5482 target machine you are compiling for. For example, you could write:
5485 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5488 The compiler automatically sets the alignment for the declared
5489 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5490 often make copy operations more efficient, because the compiler can
5491 use whatever instructions copy the biggest chunks of memory when
5492 performing copies to or from the variables or fields that you have
5493 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5494 may change depending on command-line options.
5496 When used on a struct, or struct member, the @code{aligned} attribute can
5497 only increase the alignment; in order to decrease it, the @code{packed}
5498 attribute must be specified as well. When used as part of a typedef, the
5499 @code{aligned} attribute can both increase and decrease alignment, and
5500 specifying the @code{packed} attribute generates a warning.
5502 Note that the effectiveness of @code{aligned} attributes may be limited
5503 by inherent limitations in your linker. On many systems, the linker is
5504 only able to arrange for variables to be aligned up to a certain maximum
5505 alignment. (For some linkers, the maximum supported alignment may
5506 be very very small.) If your linker is only able to align variables
5507 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5508 in an @code{__attribute__} still only provides you with 8-byte
5509 alignment. See your linker documentation for further information.
5511 The @code{aligned} attribute can also be used for functions
5512 (@pxref{Common Function Attributes}.)
5514 @item cleanup (@var{cleanup_function})
5515 @cindex @code{cleanup} variable attribute
5516 The @code{cleanup} attribute runs a function when the variable goes
5517 out of scope. This attribute can only be applied to auto function
5518 scope variables; it may not be applied to parameters or variables
5519 with static storage duration. The function must take one parameter,
5520 a pointer to a type compatible with the variable. The return value
5521 of the function (if any) is ignored.
5523 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5524 is run during the stack unwinding that happens during the
5525 processing of the exception. Note that the @code{cleanup} attribute
5526 does not allow the exception to be caught, only to perform an action.
5527 It is undefined what happens if @var{cleanup_function} does not
5532 @cindex @code{common} variable attribute
5533 @cindex @code{nocommon} variable attribute
5536 The @code{common} attribute requests GCC to place a variable in
5537 ``common'' storage. The @code{nocommon} attribute requests the
5538 opposite---to allocate space for it directly.
5540 These attributes override the default chosen by the
5541 @option{-fno-common} and @option{-fcommon} flags respectively.
5544 @itemx deprecated (@var{msg})
5545 @cindex @code{deprecated} variable attribute
5546 The @code{deprecated} attribute results in a warning if the variable
5547 is used anywhere in the source file. This is useful when identifying
5548 variables that are expected to be removed in a future version of a
5549 program. The warning also includes the location of the declaration
5550 of the deprecated variable, to enable users to easily find further
5551 information about why the variable is deprecated, or what they should
5552 do instead. Note that the warning only occurs for uses:
5555 extern int old_var __attribute__ ((deprecated));
5557 int new_fn () @{ return old_var; @}
5561 results in a warning on line 3 but not line 2. The optional @var{msg}
5562 argument, which must be a string, is printed in the warning if
5565 The @code{deprecated} attribute can also be used for functions and
5566 types (@pxref{Common Function Attributes},
5567 @pxref{Common Type Attributes}).
5569 @item mode (@var{mode})
5570 @cindex @code{mode} variable attribute
5571 This attribute specifies the data type for the declaration---whichever
5572 type corresponds to the mode @var{mode}. This in effect lets you
5573 request an integer or floating-point type according to its width.
5575 You may also specify a mode of @code{byte} or @code{__byte__} to
5576 indicate the mode corresponding to a one-byte integer, @code{word} or
5577 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5578 or @code{__pointer__} for the mode used to represent pointers.
5581 @cindex @code{packed} variable attribute
5582 The @code{packed} attribute specifies that a variable or structure field
5583 should have the smallest possible alignment---one byte for a variable,
5584 and one bit for a field, unless you specify a larger value with the
5585 @code{aligned} attribute.
5587 Here is a structure in which the field @code{x} is packed, so that it
5588 immediately follows @code{a}:
5594 int x[2] __attribute__ ((packed));
5598 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5599 @code{packed} attribute on bit-fields of type @code{char}. This has
5600 been fixed in GCC 4.4 but the change can lead to differences in the
5601 structure layout. See the documentation of
5602 @option{-Wpacked-bitfield-compat} for more information.
5604 @item section ("@var{section-name}")
5605 @cindex @code{section} variable attribute
5606 Normally, the compiler places the objects it generates in sections like
5607 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5608 or you need certain particular variables to appear in special sections,
5609 for example to map to special hardware. The @code{section}
5610 attribute specifies that a variable (or function) lives in a particular
5611 section. For example, this small program uses several specific section names:
5614 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5615 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5616 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5617 int init_data __attribute__ ((section ("INITDATA")));
5621 /* @r{Initialize stack pointer} */
5622 init_sp (stack + sizeof (stack));
5624 /* @r{Initialize initialized data} */
5625 memcpy (&init_data, &data, &edata - &data);
5627 /* @r{Turn on the serial ports} */
5634 Use the @code{section} attribute with
5635 @emph{global} variables and not @emph{local} variables,
5636 as shown in the example.
5638 You may use the @code{section} attribute with initialized or
5639 uninitialized global variables but the linker requires
5640 each object be defined once, with the exception that uninitialized
5641 variables tentatively go in the @code{common} (or @code{bss}) section
5642 and can be multiply ``defined''. Using the @code{section} attribute
5643 changes what section the variable goes into and may cause the
5644 linker to issue an error if an uninitialized variable has multiple
5645 definitions. You can force a variable to be initialized with the
5646 @option{-fno-common} flag or the @code{nocommon} attribute.
5648 Some file formats do not support arbitrary sections so the @code{section}
5649 attribute is not available on all platforms.
5650 If you need to map the entire contents of a module to a particular
5651 section, consider using the facilities of the linker instead.
5653 @item tls_model ("@var{tls_model}")
5654 @cindex @code{tls_model} variable attribute
5655 The @code{tls_model} attribute sets thread-local storage model
5656 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5657 overriding @option{-ftls-model=} command-line switch on a per-variable
5659 The @var{tls_model} argument should be one of @code{global-dynamic},
5660 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5662 Not all targets support this attribute.
5665 @cindex @code{unused} variable attribute
5666 This attribute, attached to a variable, means that the variable is meant
5667 to be possibly unused. GCC does not produce a warning for this
5671 @cindex @code{used} variable attribute
5672 This attribute, attached to a variable with static storage, means that
5673 the variable must be emitted even if it appears that the variable is not
5676 When applied to a static data member of a C++ class template, the
5677 attribute also means that the member is instantiated if the
5678 class itself is instantiated.
5680 @item vector_size (@var{bytes})
5681 @cindex @code{vector_size} variable attribute
5682 This attribute specifies the vector size for the variable, measured in
5683 bytes. For example, the declaration:
5686 int foo __attribute__ ((vector_size (16)));
5690 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5691 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5692 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5694 This attribute is only applicable to integral and float scalars,
5695 although arrays, pointers, and function return values are allowed in
5696 conjunction with this construct.
5698 Aggregates with this attribute are invalid, even if they are of the same
5699 size as a corresponding scalar. For example, the declaration:
5702 struct S @{ int a; @};
5703 struct S __attribute__ ((vector_size (16))) foo;
5707 is invalid even if the size of the structure is the same as the size of
5710 @item visibility ("@var{visibility_type}")
5711 @cindex @code{visibility} variable attribute
5712 This attribute affects the linkage of the declaration to which it is attached.
5713 The @code{visibility} attribute is described in
5714 @ref{Common Function Attributes}.
5717 @cindex @code{weak} variable attribute
5718 The @code{weak} attribute is described in
5719 @ref{Common Function Attributes}.
5723 @node AVR Variable Attributes
5724 @subsection AVR Variable Attributes
5728 @cindex @code{progmem} variable attribute, AVR
5729 The @code{progmem} attribute is used on the AVR to place read-only
5730 data in the non-volatile program memory (flash). The @code{progmem}
5731 attribute accomplishes this by putting respective variables into a
5732 section whose name starts with @code{.progmem}.
5734 This attribute works similar to the @code{section} attribute
5735 but adds additional checking. Notice that just like the
5736 @code{section} attribute, @code{progmem} affects the location
5737 of the data but not how this data is accessed.
5739 In order to read data located with the @code{progmem} attribute
5740 (inline) assembler must be used.
5742 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5743 #include <avr/pgmspace.h>
5745 /* Locate var in flash memory */
5746 const int var[2] PROGMEM = @{ 1, 2 @};
5748 int read_var (int i)
5750 /* Access var[] by accessor macro from avr/pgmspace.h */
5751 return (int) pgm_read_word (& var[i]);
5755 AVR is a Harvard architecture processor and data and read-only data
5756 normally resides in the data memory (RAM).
5758 See also the @ref{AVR Named Address Spaces} section for
5759 an alternate way to locate and access data in flash memory.
5762 @itemx io (@var{addr})
5763 @cindex @code{io} variable attribute, AVR
5764 Variables with the @code{io} attribute are used to address
5765 memory-mapped peripherals in the io address range.
5766 If an address is specified, the variable
5767 is assigned that address, and the value is interpreted as an
5768 address in the data address space.
5772 volatile int porta __attribute__((io (0x22)));
5775 The address specified in the address in the data address range.
5777 Otherwise, the variable it is not assigned an address, but the
5778 compiler will still use in/out instructions where applicable,
5779 assuming some other module assigns an address in the io address range.
5783 extern volatile int porta __attribute__((io));
5787 @itemx io_low (@var{addr})
5788 @cindex @code{io_low} variable attribute, AVR
5789 This is like the @code{io} attribute, but additionally it informs the
5790 compiler that the object lies in the lower half of the I/O area,
5791 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5795 @itemx address (@var{addr})
5796 @cindex @code{address} variable attribute, AVR
5797 Variables with the @code{address} attribute are used to address
5798 memory-mapped peripherals that may lie outside the io address range.
5801 volatile int porta __attribute__((address (0x600)));
5806 @node Blackfin Variable Attributes
5807 @subsection Blackfin Variable Attributes
5809 Three attributes are currently defined for the Blackfin.
5815 @cindex @code{l1_data} variable attribute, Blackfin
5816 @cindex @code{l1_data_A} variable attribute, Blackfin
5817 @cindex @code{l1_data_B} variable attribute, Blackfin
5818 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5819 Variables with @code{l1_data} attribute are put into the specific section
5820 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5821 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5822 attribute are put into the specific section named @code{.l1.data.B}.
5825 @cindex @code{l2} variable attribute, Blackfin
5826 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5827 Variables with @code{l2} attribute are put into the specific section
5828 named @code{.l2.data}.
5831 @node H8/300 Variable Attributes
5832 @subsection H8/300 Variable Attributes
5834 These variable attributes are available for H8/300 targets:
5838 @cindex @code{eightbit_data} variable attribute, H8/300
5839 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5840 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5841 variable should be placed into the eight-bit data section.
5842 The compiler generates more efficient code for certain operations
5843 on data in the eight-bit data area. Note the eight-bit data area is limited to
5846 You must use GAS and GLD from GNU binutils version 2.7 or later for
5847 this attribute to work correctly.
5850 @cindex @code{tiny_data} variable attribute, H8/300
5851 @cindex tiny data section on the H8/300H and H8S
5852 Use this attribute on the H8/300H and H8S to indicate that the specified
5853 variable should be placed into the tiny data section.
5854 The compiler generates more efficient code for loads and stores
5855 on data in the tiny data section. Note the tiny data area is limited to
5856 slightly under 32KB of data.
5860 @node IA-64 Variable Attributes
5861 @subsection IA-64 Variable Attributes
5863 The IA-64 back end supports the following variable attribute:
5866 @item model (@var{model-name})
5867 @cindex @code{model} variable attribute, IA-64
5869 On IA-64, use this attribute to set the addressability of an object.
5870 At present, the only supported identifier for @var{model-name} is
5871 @code{small}, indicating addressability via ``small'' (22-bit)
5872 addresses (so that their addresses can be loaded with the @code{addl}
5873 instruction). Caveat: such addressing is by definition not position
5874 independent and hence this attribute must not be used for objects
5875 defined by shared libraries.
5879 @node M32R/D Variable Attributes
5880 @subsection M32R/D Variable Attributes
5882 One attribute is currently defined for the M32R/D@.
5885 @item model (@var{model-name})
5886 @cindex @code{model-name} variable attribute, M32R/D
5887 @cindex variable addressability on the M32R/D
5888 Use this attribute on the M32R/D to set the addressability of an object.
5889 The identifier @var{model-name} is one of @code{small}, @code{medium},
5890 or @code{large}, representing each of the code models.
5892 Small model objects live in the lower 16MB of memory (so that their
5893 addresses can be loaded with the @code{ld24} instruction).
5895 Medium and large model objects may live anywhere in the 32-bit address space
5896 (the compiler generates @code{seth/add3} instructions to load their
5900 @node MeP Variable Attributes
5901 @subsection MeP Variable Attributes
5903 The MeP target has a number of addressing modes and busses. The
5904 @code{near} space spans the standard memory space's first 16 megabytes
5905 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5906 The @code{based} space is a 128-byte region in the memory space that
5907 is addressed relative to the @code{$tp} register. The @code{tiny}
5908 space is a 65536-byte region relative to the @code{$gp} register. In
5909 addition to these memory regions, the MeP target has a separate 16-bit
5910 control bus which is specified with @code{cb} attributes.
5915 @cindex @code{based} variable attribute, MeP
5916 Any variable with the @code{based} attribute is assigned to the
5917 @code{.based} section, and is accessed with relative to the
5918 @code{$tp} register.
5921 @cindex @code{tiny} variable attribute, MeP
5922 Likewise, the @code{tiny} attribute assigned variables to the
5923 @code{.tiny} section, relative to the @code{$gp} register.
5926 @cindex @code{near} variable attribute, MeP
5927 Variables with the @code{near} attribute are assumed to have addresses
5928 that fit in a 24-bit addressing mode. This is the default for large
5929 variables (@code{-mtiny=4} is the default) but this attribute can
5930 override @code{-mtiny=} for small variables, or override @code{-ml}.
5933 @cindex @code{far} variable attribute, MeP
5934 Variables with the @code{far} attribute are addressed using a full
5935 32-bit address. Since this covers the entire memory space, this
5936 allows modules to make no assumptions about where variables might be
5940 @cindex @code{io} variable attribute, MeP
5941 @itemx io (@var{addr})
5942 Variables with the @code{io} attribute are used to address
5943 memory-mapped peripherals. If an address is specified, the variable
5944 is assigned that address, else it is not assigned an address (it is
5945 assumed some other module assigns an address). Example:
5948 int timer_count __attribute__((io(0x123)));
5952 @itemx cb (@var{addr})
5953 @cindex @code{cb} variable attribute, MeP
5954 Variables with the @code{cb} attribute are used to access the control
5955 bus, using special instructions. @code{addr} indicates the control bus
5959 int cpu_clock __attribute__((cb(0x123)));
5964 @node Microsoft Windows Variable Attributes
5965 @subsection Microsoft Windows Variable Attributes
5967 You can use these attributes on Microsoft Windows targets.
5968 @ref{x86 Variable Attributes} for additional Windows compatibility
5969 attributes available on all x86 targets.
5974 @cindex @code{dllimport} variable attribute
5975 @cindex @code{dllexport} variable attribute
5976 The @code{dllimport} and @code{dllexport} attributes are described in
5977 @ref{Microsoft Windows Function Attributes}.
5980 @cindex @code{selectany} variable attribute
5981 The @code{selectany} attribute causes an initialized global variable to
5982 have link-once semantics. When multiple definitions of the variable are
5983 encountered by the linker, the first is selected and the remainder are
5984 discarded. Following usage by the Microsoft compiler, the linker is told
5985 @emph{not} to warn about size or content differences of the multiple
5988 Although the primary usage of this attribute is for POD types, the
5989 attribute can also be applied to global C++ objects that are initialized
5990 by a constructor. In this case, the static initialization and destruction
5991 code for the object is emitted in each translation defining the object,
5992 but the calls to the constructor and destructor are protected by a
5993 link-once guard variable.
5995 The @code{selectany} attribute is only available on Microsoft Windows
5996 targets. You can use @code{__declspec (selectany)} as a synonym for
5997 @code{__attribute__ ((selectany))} for compatibility with other
6001 @cindex @code{shared} variable attribute
6002 On Microsoft Windows, in addition to putting variable definitions in a named
6003 section, the section can also be shared among all running copies of an
6004 executable or DLL@. For example, this small program defines shared data
6005 by putting it in a named section @code{shared} and marking the section
6009 int foo __attribute__((section ("shared"), shared)) = 0;
6014 /* @r{Read and write foo. All running
6015 copies see the same value.} */
6021 You may only use the @code{shared} attribute along with @code{section}
6022 attribute with a fully-initialized global definition because of the way
6023 linkers work. See @code{section} attribute for more information.
6025 The @code{shared} attribute is only available on Microsoft Windows@.
6029 @node MSP430 Variable Attributes
6030 @subsection MSP430 Variable Attributes
6034 @cindex @code{noinit} MSP430 variable attribute
6035 Any data with the @code{noinit} attribute will not be initialised by
6036 the C runtime startup code, or the program loader. Not initialising
6037 data in this way can reduce program startup times.
6040 @cindex @code{persistent} MSP430 variable attribute
6041 Any variable with the @code{persistent} attribute will not be
6042 initialised by the C runtime startup code. Instead its value will be
6043 set once, when the application is loaded, and then never initialised
6044 again, even if the processor is reset or the program restarts.
6045 Persistent data is intended to be placed into FLASH RAM, where its
6046 value will be retained across resets. The linker script being used to
6047 create the application should ensure that persistent data is correctly
6053 @cindex @code{lower} memory region on the MSP430
6054 @cindex @code{upper} memory region on the MSP430
6055 @cindex @code{either} memory region on the MSP430
6056 These attributes are the same as the MSP430 function attributes of the
6057 same name. These attributes can be applied to both functions and
6061 @node PowerPC Variable Attributes
6062 @subsection PowerPC Variable Attributes
6064 Three attributes currently are defined for PowerPC configurations:
6065 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6067 @cindex @code{ms_struct} variable attribute, PowerPC
6068 @cindex @code{gcc_struct} variable attribute, PowerPC
6069 For full documentation of the struct attributes please see the
6070 documentation in @ref{x86 Variable Attributes}.
6072 @cindex @code{altivec} variable attribute, PowerPC
6073 For documentation of @code{altivec} attribute please see the
6074 documentation in @ref{PowerPC Type Attributes}.
6076 @node SPU Variable Attributes
6077 @subsection SPU Variable Attributes
6079 @cindex @code{spu_vector} variable attribute, SPU
6080 The SPU supports the @code{spu_vector} attribute for variables. For
6081 documentation of this attribute please see the documentation in
6082 @ref{SPU Type Attributes}.
6084 @node x86 Variable Attributes
6085 @subsection x86 Variable Attributes
6087 Two attributes are currently defined for x86 configurations:
6088 @code{ms_struct} and @code{gcc_struct}.
6093 @cindex @code{ms_struct} variable attribute, x86
6094 @cindex @code{gcc_struct} variable attribute, x86
6096 If @code{packed} is used on a structure, or if bit-fields are used,
6097 it may be that the Microsoft ABI lays out the structure differently
6098 than the way GCC normally does. Particularly when moving packed
6099 data between functions compiled with GCC and the native Microsoft compiler
6100 (either via function call or as data in a file), it may be necessary to access
6103 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6104 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6105 command-line options, respectively;
6106 see @ref{x86 Options}, for details of how structure layout is affected.
6107 @xref{x86 Type Attributes}, for information about the corresponding
6108 attributes on types.
6112 @node Xstormy16 Variable Attributes
6113 @subsection Xstormy16 Variable Attributes
6115 One attribute is currently defined for xstormy16 configurations:
6120 @cindex @code{below100} variable attribute, Xstormy16
6122 If a variable has the @code{below100} attribute (@code{BELOW100} is
6123 allowed also), GCC places the variable in the first 0x100 bytes of
6124 memory and use special opcodes to access it. Such variables are
6125 placed in either the @code{.bss_below100} section or the
6126 @code{.data_below100} section.
6130 @node Type Attributes
6131 @section Specifying Attributes of Types
6132 @cindex attribute of types
6133 @cindex type attributes
6135 The keyword @code{__attribute__} allows you to specify special
6136 attributes of types. Some type attributes apply only to @code{struct}
6137 and @code{union} types, while others can apply to any type defined
6138 via a @code{typedef} declaration. Other attributes are defined for
6139 functions (@pxref{Function Attributes}), labels (@pxref{Label
6140 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6141 variables (@pxref{Variable Attributes}).
6143 The @code{__attribute__} keyword is followed by an attribute specification
6144 inside double parentheses.
6146 You may specify type attributes in an enum, struct or union type
6147 declaration or definition by placing them immediately after the
6148 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6149 syntax is to place them just past the closing curly brace of the
6152 You can also include type attributes in a @code{typedef} declaration.
6153 @xref{Attribute Syntax}, for details of the exact syntax for using
6157 * Common Type Attributes::
6158 * ARM Type Attributes::
6159 * MeP Type Attributes::
6160 * PowerPC Type Attributes::
6161 * SPU Type Attributes::
6162 * x86 Type Attributes::
6165 @node Common Type Attributes
6166 @subsection Common Type Attributes
6168 The following type attributes are supported on most targets.
6171 @cindex @code{aligned} type attribute
6172 @item aligned (@var{alignment})
6173 This attribute specifies a minimum alignment (in bytes) for variables
6174 of the specified type. For example, the declarations:
6177 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6178 typedef int more_aligned_int __attribute__ ((aligned (8)));
6182 force the compiler to ensure (as far as it can) that each variable whose
6183 type is @code{struct S} or @code{more_aligned_int} is allocated and
6184 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6185 variables of type @code{struct S} aligned to 8-byte boundaries allows
6186 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6187 store) instructions when copying one variable of type @code{struct S} to
6188 another, thus improving run-time efficiency.
6190 Note that the alignment of any given @code{struct} or @code{union} type
6191 is required by the ISO C standard to be at least a perfect multiple of
6192 the lowest common multiple of the alignments of all of the members of
6193 the @code{struct} or @code{union} in question. This means that you @emph{can}
6194 effectively adjust the alignment of a @code{struct} or @code{union}
6195 type by attaching an @code{aligned} attribute to any one of the members
6196 of such a type, but the notation illustrated in the example above is a
6197 more obvious, intuitive, and readable way to request the compiler to
6198 adjust the alignment of an entire @code{struct} or @code{union} type.
6200 As in the preceding example, you can explicitly specify the alignment
6201 (in bytes) that you wish the compiler to use for a given @code{struct}
6202 or @code{union} type. Alternatively, you can leave out the alignment factor
6203 and just ask the compiler to align a type to the maximum
6204 useful alignment for the target machine you are compiling for. For
6205 example, you could write:
6208 struct S @{ short f[3]; @} __attribute__ ((aligned));
6211 Whenever you leave out the alignment factor in an @code{aligned}
6212 attribute specification, the compiler automatically sets the alignment
6213 for the type to the largest alignment that is ever used for any data
6214 type on the target machine you are compiling for. Doing this can often
6215 make copy operations more efficient, because the compiler can use
6216 whatever instructions copy the biggest chunks of memory when performing
6217 copies to or from the variables that have types that you have aligned
6220 In the example above, if the size of each @code{short} is 2 bytes, then
6221 the size of the entire @code{struct S} type is 6 bytes. The smallest
6222 power of two that is greater than or equal to that is 8, so the
6223 compiler sets the alignment for the entire @code{struct S} type to 8
6226 Note that although you can ask the compiler to select a time-efficient
6227 alignment for a given type and then declare only individual stand-alone
6228 objects of that type, the compiler's ability to select a time-efficient
6229 alignment is primarily useful only when you plan to create arrays of
6230 variables having the relevant (efficiently aligned) type. If you
6231 declare or use arrays of variables of an efficiently-aligned type, then
6232 it is likely that your program also does pointer arithmetic (or
6233 subscripting, which amounts to the same thing) on pointers to the
6234 relevant type, and the code that the compiler generates for these
6235 pointer arithmetic operations is often more efficient for
6236 efficiently-aligned types than for other types.
6238 The @code{aligned} attribute can only increase the alignment; but you
6239 can decrease it by specifying @code{packed} as well. See below.
6241 Note that the effectiveness of @code{aligned} attributes may be limited
6242 by inherent limitations in your linker. On many systems, the linker is
6243 only able to arrange for variables to be aligned up to a certain maximum
6244 alignment. (For some linkers, the maximum supported alignment may
6245 be very very small.) If your linker is only able to align variables
6246 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6247 in an @code{__attribute__} still only provides you with 8-byte
6248 alignment. See your linker documentation for further information.
6250 @opindex fshort-enums
6251 Specifying this attribute for @code{struct} and @code{union} types is
6252 equivalent to specifying the @code{packed} attribute on each of the
6253 structure or union members. Specifying the @option{-fshort-enums}
6254 flag on the line is equivalent to specifying the @code{packed}
6255 attribute on all @code{enum} definitions.
6257 In the following example @code{struct my_packed_struct}'s members are
6258 packed closely together, but the internal layout of its @code{s} member
6259 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6263 struct my_unpacked_struct
6269 struct __attribute__ ((__packed__)) my_packed_struct
6273 struct my_unpacked_struct s;
6277 You may only specify this attribute on the definition of an @code{enum},
6278 @code{struct} or @code{union}, not on a @code{typedef} that does not
6279 also define the enumerated type, structure or union.
6281 @item bnd_variable_size
6282 @cindex @code{bnd_variable_size} type attribute
6283 @cindex Pointer Bounds Checker attributes
6284 When applied to a structure field, this attribute tells Pointer
6285 Bounds Checker that the size of this field should not be computed
6286 using static type information. It may be used to mark variably-sized
6287 static array fields placed at the end of a structure.
6295 S *p = (S *)malloc (sizeof(S) + 100);
6296 p->data[10] = 0; //Bounds violation
6300 By using an attribute for the field we may avoid unwanted bound
6307 char data[1] __attribute__((bnd_variable_size));
6309 S *p = (S *)malloc (sizeof(S) + 100);
6310 p->data[10] = 0; //OK
6314 @itemx deprecated (@var{msg})
6315 @cindex @code{deprecated} type attribute
6316 The @code{deprecated} attribute results in a warning if the type
6317 is used anywhere in the source file. This is useful when identifying
6318 types that are expected to be removed in a future version of a program.
6319 If possible, the warning also includes the location of the declaration
6320 of the deprecated type, to enable users to easily find further
6321 information about why the type is deprecated, or what they should do
6322 instead. Note that the warnings only occur for uses and then only
6323 if the type is being applied to an identifier that itself is not being
6324 declared as deprecated.
6327 typedef int T1 __attribute__ ((deprecated));
6331 typedef T1 T3 __attribute__ ((deprecated));
6332 T3 z __attribute__ ((deprecated));
6336 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6337 warning is issued for line 4 because T2 is not explicitly
6338 deprecated. Line 5 has no warning because T3 is explicitly
6339 deprecated. Similarly for line 6. The optional @var{msg}
6340 argument, which must be a string, is printed in the warning if
6343 The @code{deprecated} attribute can also be used for functions and
6344 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6346 @item designated_init
6347 @cindex @code{designated_init} type attribute
6348 This attribute may only be applied to structure types. It indicates
6349 that any initialization of an object of this type must use designated
6350 initializers rather than positional initializers. The intent of this
6351 attribute is to allow the programmer to indicate that a structure's
6352 layout may change, and that therefore relying on positional
6353 initialization will result in future breakage.
6355 GCC emits warnings based on this attribute by default; use
6356 @option{-Wno-designated-init} to suppress them.
6359 @cindex @code{may_alias} type attribute
6360 Accesses through pointers to types with this attribute are not subject
6361 to type-based alias analysis, but are instead assumed to be able to alias
6362 any other type of objects.
6363 In the context of section 6.5 paragraph 7 of the C99 standard,
6364 an lvalue expression
6365 dereferencing such a pointer is treated like having a character type.
6366 See @option{-fstrict-aliasing} for more information on aliasing issues.
6367 This extension exists to support some vector APIs, in which pointers to
6368 one vector type are permitted to alias pointers to a different vector type.
6370 Note that an object of a type with this attribute does not have any
6376 typedef short __attribute__((__may_alias__)) short_a;
6382 short_a *b = (short_a *) &a;
6386 if (a == 0x12345678)
6394 If you replaced @code{short_a} with @code{short} in the variable
6395 declaration, the above program would abort when compiled with
6396 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6400 @cindex @code{packed} type attribute
6401 This attribute, attached to @code{struct} or @code{union} type
6402 definition, specifies that each member (other than zero-width bit-fields)
6403 of the structure or union is placed to minimize the memory required. When
6404 attached to an @code{enum} definition, it indicates that the smallest
6405 integral type should be used.
6407 @item scalar_storage_order ("@var{endianness}")
6408 @cindex @code{scalar_storage_order} type attribute
6409 When attached to a @code{union} or a @code{struct}, this attribute sets
6410 the storage order, aka endianness, of the scalar fields of the type, as
6411 well as the array fields whose component is scalar. The supported
6412 endianness are @code{big-endian} and @code{little-endian}. The attribute
6413 has no effects on fields which are themselves a @code{union}, a @code{struct}
6414 or an array whose component is a @code{union} or a @code{struct}, and it is
6415 possible to have fields with a different scalar storage order than the
6418 This attribute is supported only for targets that use a uniform default
6419 scalar storage order (fortunately, most of them), i.e. targets that store
6420 the scalars either all in big-endian or all in little-endian.
6422 Additional restrictions are enforced for types with the reverse scalar
6423 storage order with regard to the scalar storage order of the target:
6426 @item Taking the address of a scalar field of a @code{union} or a
6427 @code{struct} with reverse scalar storage order is not permitted and will
6429 @item Taking the address of an array field, whose component is scalar, of
6430 a @code{union} or a @code{struct} with reverse scalar storage order is
6431 permitted but will yield a warning, unless @option{-Wno-scalar-storage-order}
6433 @item Taking the address of a @code{union} or a @code{struct} with reverse
6434 scalar storage order is permitted.
6437 These restrictions exist because the storage order attribute is lost when
6438 the address of a scalar or the address of an array with scalar component
6439 is taken, so storing indirectly through this address will generally not work.
6440 The second case is nevertheless allowed to be able to perform a block copy
6441 from or to the array.
6443 @item transparent_union
6444 @cindex @code{transparent_union} type attribute
6446 This attribute, attached to a @code{union} type definition, indicates
6447 that any function parameter having that union type causes calls to that
6448 function to be treated in a special way.
6450 First, the argument corresponding to a transparent union type can be of
6451 any type in the union; no cast is required. Also, if the union contains
6452 a pointer type, the corresponding argument can be a null pointer
6453 constant or a void pointer expression; and if the union contains a void
6454 pointer type, the corresponding argument can be any pointer expression.
6455 If the union member type is a pointer, qualifiers like @code{const} on
6456 the referenced type must be respected, just as with normal pointer
6459 Second, the argument is passed to the function using the calling
6460 conventions of the first member of the transparent union, not the calling
6461 conventions of the union itself. All members of the union must have the
6462 same machine representation; this is necessary for this argument passing
6465 Transparent unions are designed for library functions that have multiple
6466 interfaces for compatibility reasons. For example, suppose the
6467 @code{wait} function must accept either a value of type @code{int *} to
6468 comply with POSIX, or a value of type @code{union wait *} to comply with
6469 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6470 @code{wait} would accept both kinds of arguments, but it would also
6471 accept any other pointer type and this would make argument type checking
6472 less useful. Instead, @code{<sys/wait.h>} might define the interface
6476 typedef union __attribute__ ((__transparent_union__))
6480 @} wait_status_ptr_t;
6482 pid_t wait (wait_status_ptr_t);
6486 This interface allows either @code{int *} or @code{union wait *}
6487 arguments to be passed, using the @code{int *} calling convention.
6488 The program can call @code{wait} with arguments of either type:
6491 int w1 () @{ int w; return wait (&w); @}
6492 int w2 () @{ union wait w; return wait (&w); @}
6496 With this interface, @code{wait}'s implementation might look like this:
6499 pid_t wait (wait_status_ptr_t p)
6501 return waitpid (-1, p.__ip, 0);
6506 @cindex @code{unused} type attribute
6507 When attached to a type (including a @code{union} or a @code{struct}),
6508 this attribute means that variables of that type are meant to appear
6509 possibly unused. GCC does not produce a warning for any variables of
6510 that type, even if the variable appears to do nothing. This is often
6511 the case with lock or thread classes, which are usually defined and then
6512 not referenced, but contain constructors and destructors that have
6513 nontrivial bookkeeping functions.
6516 @cindex @code{visibility} type attribute
6517 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6518 applied to class, struct, union and enum types. Unlike other type
6519 attributes, the attribute must appear between the initial keyword and
6520 the name of the type; it cannot appear after the body of the type.
6522 Note that the type visibility is applied to vague linkage entities
6523 associated with the class (vtable, typeinfo node, etc.). In
6524 particular, if a class is thrown as an exception in one shared object
6525 and caught in another, the class must have default visibility.
6526 Otherwise the two shared objects are unable to use the same
6527 typeinfo node and exception handling will break.
6531 To specify multiple attributes, separate them by commas within the
6532 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6535 @node ARM Type Attributes
6536 @subsection ARM Type Attributes
6538 @cindex @code{notshared} type attribute, ARM
6539 On those ARM targets that support @code{dllimport} (such as Symbian
6540 OS), you can use the @code{notshared} attribute to indicate that the
6541 virtual table and other similar data for a class should not be
6542 exported from a DLL@. For example:
6545 class __declspec(notshared) C @{
6547 __declspec(dllimport) C();
6551 __declspec(dllexport)
6556 In this code, @code{C::C} is exported from the current DLL, but the
6557 virtual table for @code{C} is not exported. (You can use
6558 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6559 most Symbian OS code uses @code{__declspec}.)
6561 @node MeP Type Attributes
6562 @subsection MeP Type Attributes
6564 @cindex @code{based} type attribute, MeP
6565 @cindex @code{tiny} type attribute, MeP
6566 @cindex @code{near} type attribute, MeP
6567 @cindex @code{far} type attribute, MeP
6568 Many of the MeP variable attributes may be applied to types as well.
6569 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6570 @code{far} attributes may be applied to either. The @code{io} and
6571 @code{cb} attributes may not be applied to types.
6573 @node PowerPC Type Attributes
6574 @subsection PowerPC Type Attributes
6576 Three attributes currently are defined for PowerPC configurations:
6577 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6579 @cindex @code{ms_struct} type attribute, PowerPC
6580 @cindex @code{gcc_struct} type attribute, PowerPC
6581 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6582 attributes please see the documentation in @ref{x86 Type Attributes}.
6584 @cindex @code{altivec} type attribute, PowerPC
6585 The @code{altivec} attribute allows one to declare AltiVec vector data
6586 types supported by the AltiVec Programming Interface Manual. The
6587 attribute requires an argument to specify one of three vector types:
6588 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6589 and @code{bool__} (always followed by unsigned).
6592 __attribute__((altivec(vector__)))
6593 __attribute__((altivec(pixel__))) unsigned short
6594 __attribute__((altivec(bool__))) unsigned
6597 These attributes mainly are intended to support the @code{__vector},
6598 @code{__pixel}, and @code{__bool} AltiVec keywords.
6600 @node SPU Type Attributes
6601 @subsection SPU Type Attributes
6603 @cindex @code{spu_vector} type attribute, SPU
6604 The SPU supports the @code{spu_vector} attribute for types. This attribute
6605 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6606 Language Extensions Specification. It is intended to support the
6607 @code{__vector} keyword.
6609 @node x86 Type Attributes
6610 @subsection x86 Type Attributes
6612 Two attributes are currently defined for x86 configurations:
6613 @code{ms_struct} and @code{gcc_struct}.
6619 @cindex @code{ms_struct} type attribute, x86
6620 @cindex @code{gcc_struct} type attribute, x86
6622 If @code{packed} is used on a structure, or if bit-fields are used
6623 it may be that the Microsoft ABI packs them differently
6624 than GCC normally packs them. Particularly when moving packed
6625 data between functions compiled with GCC and the native Microsoft compiler
6626 (either via function call or as data in a file), it may be necessary to access
6629 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6630 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6631 command-line options, respectively;
6632 see @ref{x86 Options}, for details of how structure layout is affected.
6633 @xref{x86 Variable Attributes}, for information about the corresponding
6634 attributes on variables.
6638 @node Label Attributes
6639 @section Label Attributes
6640 @cindex Label Attributes
6642 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6643 details of the exact syntax for using attributes. Other attributes are
6644 available for functions (@pxref{Function Attributes}), variables
6645 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6646 and for types (@pxref{Type Attributes}).
6648 This example uses the @code{cold} label attribute to indicate the
6649 @code{ErrorHandling} branch is unlikely to be taken and that the
6650 @code{ErrorHandling} label is unused:
6654 asm goto ("some asm" : : : : NoError);
6656 /* This branch (the fall-through from the asm) is less commonly used */
6658 __attribute__((cold, unused)); /* Semi-colon is required here */
6663 printf("no error\n");
6669 @cindex @code{unused} label attribute
6670 This feature is intended for program-generated code that may contain
6671 unused labels, but which is compiled with @option{-Wall}. It is
6672 not normally appropriate to use in it human-written code, though it
6673 could be useful in cases where the code that jumps to the label is
6674 contained within an @code{#ifdef} conditional.
6677 @cindex @code{hot} label attribute
6678 The @code{hot} attribute on a label is used to inform the compiler that
6679 the path following the label is more likely than paths that are not so
6680 annotated. This attribute is used in cases where @code{__builtin_expect}
6681 cannot be used, for instance with computed goto or @code{asm goto}.
6684 @cindex @code{cold} label attribute
6685 The @code{cold} attribute on labels is used to inform the compiler that
6686 the path following the label is unlikely to be executed. This attribute
6687 is used in cases where @code{__builtin_expect} cannot be used, for instance
6688 with computed goto or @code{asm goto}.
6692 @node Enumerator Attributes
6693 @section Enumerator Attributes
6694 @cindex Enumerator Attributes
6696 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6697 details of the exact syntax for using attributes. Other attributes are
6698 available for functions (@pxref{Function Attributes}), variables
6699 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6700 and for types (@pxref{Type Attributes}).
6702 This example uses the @code{deprecated} enumerator attribute to indicate the
6703 @code{oldval} enumerator is deprecated:
6707 oldval __attribute__((deprecated)),
6720 @cindex @code{deprecated} enumerator attribute
6721 The @code{deprecated} attribute results in a warning if the enumerator
6722 is used anywhere in the source file. This is useful when identifying
6723 enumerators that are expected to be removed in a future version of a
6724 program. The warning also includes the location of the declaration
6725 of the deprecated enumerator, to enable users to easily find further
6726 information about why the enumerator is deprecated, or what they should
6727 do instead. Note that the warnings only occurs for uses.
6731 @node Attribute Syntax
6732 @section Attribute Syntax
6733 @cindex attribute syntax
6735 This section describes the syntax with which @code{__attribute__} may be
6736 used, and the constructs to which attribute specifiers bind, for the C
6737 language. Some details may vary for C++ and Objective-C@. Because of
6738 infelicities in the grammar for attributes, some forms described here
6739 may not be successfully parsed in all cases.
6741 There are some problems with the semantics of attributes in C++. For
6742 example, there are no manglings for attributes, although they may affect
6743 code generation, so problems may arise when attributed types are used in
6744 conjunction with templates or overloading. Similarly, @code{typeid}
6745 does not distinguish between types with different attributes. Support
6746 for attributes in C++ may be restricted in future to attributes on
6747 declarations only, but not on nested declarators.
6749 @xref{Function Attributes}, for details of the semantics of attributes
6750 applying to functions. @xref{Variable Attributes}, for details of the
6751 semantics of attributes applying to variables. @xref{Type Attributes},
6752 for details of the semantics of attributes applying to structure, union
6753 and enumerated types.
6754 @xref{Label Attributes}, for details of the semantics of attributes
6756 @xref{Enumerator Attributes}, for details of the semantics of attributes
6757 applying to enumerators.
6759 An @dfn{attribute specifier} is of the form
6760 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6761 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6762 each attribute is one of the following:
6766 Empty. Empty attributes are ignored.
6770 (which may be an identifier such as @code{unused}, or a reserved
6771 word such as @code{const}).
6774 An attribute name followed by a parenthesized list of
6775 parameters for the attribute.
6776 These parameters take one of the following forms:
6780 An identifier. For example, @code{mode} attributes use this form.
6783 An identifier followed by a comma and a non-empty comma-separated list
6784 of expressions. For example, @code{format} attributes use this form.
6787 A possibly empty comma-separated list of expressions. For example,
6788 @code{format_arg} attributes use this form with the list being a single
6789 integer constant expression, and @code{alias} attributes use this form
6790 with the list being a single string constant.
6794 An @dfn{attribute specifier list} is a sequence of one or more attribute
6795 specifiers, not separated by any other tokens.
6797 You may optionally specify attribute names with @samp{__}
6798 preceding and following the name.
6799 This allows you to use them in header files without
6800 being concerned about a possible macro of the same name. For example,
6801 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6804 @subsubheading Label Attributes
6806 In GNU C, an attribute specifier list may appear after the colon following a
6807 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6808 attributes on labels if the attribute specifier is immediately
6809 followed by a semicolon (i.e., the label applies to an empty
6810 statement). If the semicolon is missing, C++ label attributes are
6811 ambiguous, as it is permissible for a declaration, which could begin
6812 with an attribute list, to be labelled in C++. Declarations cannot be
6813 labelled in C90 or C99, so the ambiguity does not arise there.
6815 @subsubheading Enumerator Attributes
6817 In GNU C, an attribute specifier list may appear as part of an enumerator.
6818 The attribute goes after the enumeration constant, before @code{=}, if
6819 present. The optional attribute in the enumerator appertains to the
6820 enumeration constant. It is not possible to place the attribute after
6821 the constant expression, if present.
6823 @subsubheading Type Attributes
6825 An attribute specifier list may appear as part of a @code{struct},
6826 @code{union} or @code{enum} specifier. It may go either immediately
6827 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6828 the closing brace. The former syntax is preferred.
6829 Where attribute specifiers follow the closing brace, they are considered
6830 to relate to the structure, union or enumerated type defined, not to any
6831 enclosing declaration the type specifier appears in, and the type
6832 defined is not complete until after the attribute specifiers.
6833 @c Otherwise, there would be the following problems: a shift/reduce
6834 @c conflict between attributes binding the struct/union/enum and
6835 @c binding to the list of specifiers/qualifiers; and "aligned"
6836 @c attributes could use sizeof for the structure, but the size could be
6837 @c changed later by "packed" attributes.
6840 @subsubheading All other attributes
6842 Otherwise, an attribute specifier appears as part of a declaration,
6843 counting declarations of unnamed parameters and type names, and relates
6844 to that declaration (which may be nested in another declaration, for
6845 example in the case of a parameter declaration), or to a particular declarator
6846 within a declaration. Where an
6847 attribute specifier is applied to a parameter declared as a function or
6848 an array, it should apply to the function or array rather than the
6849 pointer to which the parameter is implicitly converted, but this is not
6850 yet correctly implemented.
6852 Any list of specifiers and qualifiers at the start of a declaration may
6853 contain attribute specifiers, whether or not such a list may in that
6854 context contain storage class specifiers. (Some attributes, however,
6855 are essentially in the nature of storage class specifiers, and only make
6856 sense where storage class specifiers may be used; for example,
6857 @code{section}.) There is one necessary limitation to this syntax: the
6858 first old-style parameter declaration in a function definition cannot
6859 begin with an attribute specifier, because such an attribute applies to
6860 the function instead by syntax described below (which, however, is not
6861 yet implemented in this case). In some other cases, attribute
6862 specifiers are permitted by this grammar but not yet supported by the
6863 compiler. All attribute specifiers in this place relate to the
6864 declaration as a whole. In the obsolescent usage where a type of
6865 @code{int} is implied by the absence of type specifiers, such a list of
6866 specifiers and qualifiers may be an attribute specifier list with no
6867 other specifiers or qualifiers.
6869 At present, the first parameter in a function prototype must have some
6870 type specifier that is not an attribute specifier; this resolves an
6871 ambiguity in the interpretation of @code{void f(int
6872 (__attribute__((foo)) x))}, but is subject to change. At present, if
6873 the parentheses of a function declarator contain only attributes then
6874 those attributes are ignored, rather than yielding an error or warning
6875 or implying a single parameter of type int, but this is subject to
6878 An attribute specifier list may appear immediately before a declarator
6879 (other than the first) in a comma-separated list of declarators in a
6880 declaration of more than one identifier using a single list of
6881 specifiers and qualifiers. Such attribute specifiers apply
6882 only to the identifier before whose declarator they appear. For
6886 __attribute__((noreturn)) void d0 (void),
6887 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6892 the @code{noreturn} attribute applies to all the functions
6893 declared; the @code{format} attribute only applies to @code{d1}.
6895 An attribute specifier list may appear immediately before the comma,
6896 @code{=} or semicolon terminating the declaration of an identifier other
6897 than a function definition. Such attribute specifiers apply
6898 to the declared object or function. Where an
6899 assembler name for an object or function is specified (@pxref{Asm
6900 Labels}), the attribute must follow the @code{asm}
6903 An attribute specifier list may, in future, be permitted to appear after
6904 the declarator in a function definition (before any old-style parameter
6905 declarations or the function body).
6907 Attribute specifiers may be mixed with type qualifiers appearing inside
6908 the @code{[]} of a parameter array declarator, in the C99 construct by
6909 which such qualifiers are applied to the pointer to which the array is
6910 implicitly converted. Such attribute specifiers apply to the pointer,
6911 not to the array, but at present this is not implemented and they are
6914 An attribute specifier list may appear at the start of a nested
6915 declarator. At present, there are some limitations in this usage: the
6916 attributes correctly apply to the declarator, but for most individual
6917 attributes the semantics this implies are not implemented.
6918 When attribute specifiers follow the @code{*} of a pointer
6919 declarator, they may be mixed with any type qualifiers present.
6920 The following describes the formal semantics of this syntax. It makes the
6921 most sense if you are familiar with the formal specification of
6922 declarators in the ISO C standard.
6924 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
6925 D1}, where @code{T} contains declaration specifiers that specify a type
6926 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
6927 contains an identifier @var{ident}. The type specified for @var{ident}
6928 for derived declarators whose type does not include an attribute
6929 specifier is as in the ISO C standard.
6931 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
6932 and the declaration @code{T D} specifies the type
6933 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6934 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6935 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
6937 If @code{D1} has the form @code{*
6938 @var{type-qualifier-and-attribute-specifier-list} D}, and the
6939 declaration @code{T D} specifies the type
6940 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6941 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6942 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
6948 void (__attribute__((noreturn)) ****f) (void);
6952 specifies the type ``pointer to pointer to pointer to pointer to
6953 non-returning function returning @code{void}''. As another example,
6956 char *__attribute__((aligned(8))) *f;
6960 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
6961 Note again that this does not work with most attributes; for example,
6962 the usage of @samp{aligned} and @samp{noreturn} attributes given above
6963 is not yet supported.
6965 For compatibility with existing code written for compiler versions that
6966 did not implement attributes on nested declarators, some laxity is
6967 allowed in the placing of attributes. If an attribute that only applies
6968 to types is applied to a declaration, it is treated as applying to
6969 the type of that declaration. If an attribute that only applies to
6970 declarations is applied to the type of a declaration, it is treated
6971 as applying to that declaration; and, for compatibility with code
6972 placing the attributes immediately before the identifier declared, such
6973 an attribute applied to a function return type is treated as
6974 applying to the function type, and such an attribute applied to an array
6975 element type is treated as applying to the array type. If an
6976 attribute that only applies to function types is applied to a
6977 pointer-to-function type, it is treated as applying to the pointer
6978 target type; if such an attribute is applied to a function return type
6979 that is not a pointer-to-function type, it is treated as applying
6980 to the function type.
6982 @node Function Prototypes
6983 @section Prototypes and Old-Style Function Definitions
6984 @cindex function prototype declarations
6985 @cindex old-style function definitions
6986 @cindex promotion of formal parameters
6988 GNU C extends ISO C to allow a function prototype to override a later
6989 old-style non-prototype definition. Consider the following example:
6992 /* @r{Use prototypes unless the compiler is old-fashioned.} */
6999 /* @r{Prototype function declaration.} */
7000 int isroot P((uid_t));
7002 /* @r{Old-style function definition.} */
7004 isroot (x) /* @r{??? lossage here ???} */
7011 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7012 not allow this example, because subword arguments in old-style
7013 non-prototype definitions are promoted. Therefore in this example the
7014 function definition's argument is really an @code{int}, which does not
7015 match the prototype argument type of @code{short}.
7017 This restriction of ISO C makes it hard to write code that is portable
7018 to traditional C compilers, because the programmer does not know
7019 whether the @code{uid_t} type is @code{short}, @code{int}, or
7020 @code{long}. Therefore, in cases like these GNU C allows a prototype
7021 to override a later old-style definition. More precisely, in GNU C, a
7022 function prototype argument type overrides the argument type specified
7023 by a later old-style definition if the former type is the same as the
7024 latter type before promotion. Thus in GNU C the above example is
7025 equivalent to the following:
7038 GNU C++ does not support old-style function definitions, so this
7039 extension is irrelevant.
7042 @section C++ Style Comments
7044 @cindex C++ comments
7045 @cindex comments, C++ style
7047 In GNU C, you may use C++ style comments, which start with @samp{//} and
7048 continue until the end of the line. Many other C implementations allow
7049 such comments, and they are included in the 1999 C standard. However,
7050 C++ style comments are not recognized if you specify an @option{-std}
7051 option specifying a version of ISO C before C99, or @option{-ansi}
7052 (equivalent to @option{-std=c90}).
7055 @section Dollar Signs in Identifier Names
7057 @cindex dollar signs in identifier names
7058 @cindex identifier names, dollar signs in
7060 In GNU C, you may normally use dollar signs in identifier names.
7061 This is because many traditional C implementations allow such identifiers.
7062 However, dollar signs in identifiers are not supported on a few target
7063 machines, typically because the target assembler does not allow them.
7065 @node Character Escapes
7066 @section The Character @key{ESC} in Constants
7068 You can use the sequence @samp{\e} in a string or character constant to
7069 stand for the ASCII character @key{ESC}.
7072 @section Inquiring on Alignment of Types or Variables
7074 @cindex type alignment
7075 @cindex variable alignment
7077 The keyword @code{__alignof__} allows you to inquire about how an object
7078 is aligned, or the minimum alignment usually required by a type. Its
7079 syntax is just like @code{sizeof}.
7081 For example, if the target machine requires a @code{double} value to be
7082 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7083 This is true on many RISC machines. On more traditional machine
7084 designs, @code{__alignof__ (double)} is 4 or even 2.
7086 Some machines never actually require alignment; they allow reference to any
7087 data type even at an odd address. For these machines, @code{__alignof__}
7088 reports the smallest alignment that GCC gives the data type, usually as
7089 mandated by the target ABI.
7091 If the operand of @code{__alignof__} is an lvalue rather than a type,
7092 its value is the required alignment for its type, taking into account
7093 any minimum alignment specified with GCC's @code{__attribute__}
7094 extension (@pxref{Variable Attributes}). For example, after this
7098 struct foo @{ int x; char y; @} foo1;
7102 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7103 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7105 It is an error to ask for the alignment of an incomplete type.
7109 @section An Inline Function is As Fast As a Macro
7110 @cindex inline functions
7111 @cindex integrating function code
7113 @cindex macros, inline alternative
7115 By declaring a function inline, you can direct GCC to make
7116 calls to that function faster. One way GCC can achieve this is to
7117 integrate that function's code into the code for its callers. This
7118 makes execution faster by eliminating the function-call overhead; in
7119 addition, if any of the actual argument values are constant, their
7120 known values may permit simplifications at compile time so that not
7121 all of the inline function's code needs to be included. The effect on
7122 code size is less predictable; object code may be larger or smaller
7123 with function inlining, depending on the particular case. You can
7124 also direct GCC to try to integrate all ``simple enough'' functions
7125 into their callers with the option @option{-finline-functions}.
7127 GCC implements three different semantics of declaring a function
7128 inline. One is available with @option{-std=gnu89} or
7129 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7130 on all inline declarations, another when
7131 @option{-std=c99}, @option{-std=c11},
7132 @option{-std=gnu99} or @option{-std=gnu11}
7133 (without @option{-fgnu89-inline}), and the third
7134 is used when compiling C++.
7136 To declare a function inline, use the @code{inline} keyword in its
7137 declaration, like this:
7147 If you are writing a header file to be included in ISO C90 programs, write
7148 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7150 The three types of inlining behave similarly in two important cases:
7151 when the @code{inline} keyword is used on a @code{static} function,
7152 like the example above, and when a function is first declared without
7153 using the @code{inline} keyword and then is defined with
7154 @code{inline}, like this:
7157 extern int inc (int *a);
7165 In both of these common cases, the program behaves the same as if you
7166 had not used the @code{inline} keyword, except for its speed.
7168 @cindex inline functions, omission of
7169 @opindex fkeep-inline-functions
7170 When a function is both inline and @code{static}, if all calls to the
7171 function are integrated into the caller, and the function's address is
7172 never used, then the function's own assembler code is never referenced.
7173 In this case, GCC does not actually output assembler code for the
7174 function, unless you specify the option @option{-fkeep-inline-functions}.
7175 If there is a nonintegrated call, then the function is compiled to
7176 assembler code as usual. The function must also be compiled as usual if
7177 the program refers to its address, because that can't be inlined.
7180 Note that certain usages in a function definition can make it unsuitable
7181 for inline substitution. Among these usages are: variadic functions,
7182 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7183 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7184 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7185 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7186 function marked @code{inline} could not be substituted, and gives the
7187 reason for the failure.
7189 @cindex automatic @code{inline} for C++ member fns
7190 @cindex @code{inline} automatic for C++ member fns
7191 @cindex member fns, automatically @code{inline}
7192 @cindex C++ member fns, automatically @code{inline}
7193 @opindex fno-default-inline
7194 As required by ISO C++, GCC considers member functions defined within
7195 the body of a class to be marked inline even if they are
7196 not explicitly declared with the @code{inline} keyword. You can
7197 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7198 Options,,Options Controlling C++ Dialect}.
7200 GCC does not inline any functions when not optimizing unless you specify
7201 the @samp{always_inline} attribute for the function, like this:
7204 /* @r{Prototype.} */
7205 inline void foo (const char) __attribute__((always_inline));
7208 The remainder of this section is specific to GNU C90 inlining.
7210 @cindex non-static inline function
7211 When an inline function is not @code{static}, then the compiler must assume
7212 that there may be calls from other source files; since a global symbol can
7213 be defined only once in any program, the function must not be defined in
7214 the other source files, so the calls therein cannot be integrated.
7215 Therefore, a non-@code{static} inline function is always compiled on its
7216 own in the usual fashion.
7218 If you specify both @code{inline} and @code{extern} in the function
7219 definition, then the definition is used only for inlining. In no case
7220 is the function compiled on its own, not even if you refer to its
7221 address explicitly. Such an address becomes an external reference, as
7222 if you had only declared the function, and had not defined it.
7224 This combination of @code{inline} and @code{extern} has almost the
7225 effect of a macro. The way to use it is to put a function definition in
7226 a header file with these keywords, and put another copy of the
7227 definition (lacking @code{inline} and @code{extern}) in a library file.
7228 The definition in the header file causes most calls to the function
7229 to be inlined. If any uses of the function remain, they refer to
7230 the single copy in the library.
7233 @section When is a Volatile Object Accessed?
7234 @cindex accessing volatiles
7235 @cindex volatile read
7236 @cindex volatile write
7237 @cindex volatile access
7239 C has the concept of volatile objects. These are normally accessed by
7240 pointers and used for accessing hardware or inter-thread
7241 communication. The standard encourages compilers to refrain from
7242 optimizations concerning accesses to volatile objects, but leaves it
7243 implementation defined as to what constitutes a volatile access. The
7244 minimum requirement is that at a sequence point all previous accesses
7245 to volatile objects have stabilized and no subsequent accesses have
7246 occurred. Thus an implementation is free to reorder and combine
7247 volatile accesses that occur between sequence points, but cannot do
7248 so for accesses across a sequence point. The use of volatile does
7249 not allow you to violate the restriction on updating objects multiple
7250 times between two sequence points.
7252 Accesses to non-volatile objects are not ordered with respect to
7253 volatile accesses. You cannot use a volatile object as a memory
7254 barrier to order a sequence of writes to non-volatile memory. For
7258 int *ptr = @var{something};
7260 *ptr = @var{something};
7265 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7266 that the write to @var{*ptr} occurs by the time the update
7267 of @var{vobj} happens. If you need this guarantee, you must use
7268 a stronger memory barrier such as:
7271 int *ptr = @var{something};
7273 *ptr = @var{something};
7274 asm volatile ("" : : : "memory");
7278 A scalar volatile object is read when it is accessed in a void context:
7281 volatile int *src = @var{somevalue};
7285 Such expressions are rvalues, and GCC implements this as a
7286 read of the volatile object being pointed to.
7288 Assignments are also expressions and have an rvalue. However when
7289 assigning to a scalar volatile, the volatile object is not reread,
7290 regardless of whether the assignment expression's rvalue is used or
7291 not. If the assignment's rvalue is used, the value is that assigned
7292 to the volatile object. For instance, there is no read of @var{vobj}
7293 in all the following cases:
7298 vobj = @var{something};
7299 obj = vobj = @var{something};
7300 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7301 obj = (@var{something}, vobj = @var{anotherthing});
7304 If you need to read the volatile object after an assignment has
7305 occurred, you must use a separate expression with an intervening
7308 As bit-fields are not individually addressable, volatile bit-fields may
7309 be implicitly read when written to, or when adjacent bit-fields are
7310 accessed. Bit-field operations may be optimized such that adjacent
7311 bit-fields are only partially accessed, if they straddle a storage unit
7312 boundary. For these reasons it is unwise to use volatile bit-fields to
7315 @node Using Assembly Language with C
7316 @section How to Use Inline Assembly Language in C Code
7317 @cindex @code{asm} keyword
7318 @cindex assembly language in C
7319 @cindex inline assembly language
7320 @cindex mixing assembly language and C
7322 The @code{asm} keyword allows you to embed assembler instructions
7323 within C code. GCC provides two forms of inline @code{asm}
7324 statements. A @dfn{basic @code{asm}} statement is one with no
7325 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7326 statement (@pxref{Extended Asm}) includes one or more operands.
7327 The extended form is preferred for mixing C and assembly language
7328 within a function, but to include assembly language at
7329 top level you must use basic @code{asm}.
7331 You can also use the @code{asm} keyword to override the assembler name
7332 for a C symbol, or to place a C variable in a specific register.
7335 * Basic Asm:: Inline assembler without operands.
7336 * Extended Asm:: Inline assembler with operands.
7337 * Constraints:: Constraints for @code{asm} operands
7338 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7339 * Explicit Register Variables:: Defining variables residing in specified
7341 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7345 @subsection Basic Asm --- Assembler Instructions Without Operands
7346 @cindex basic @code{asm}
7347 @cindex assembly language in C, basic
7349 A basic @code{asm} statement has the following syntax:
7352 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7355 The @code{asm} keyword is a GNU extension.
7356 When writing code that can be compiled with @option{-ansi} and the
7357 various @option{-std} options, use @code{__asm__} instead of
7358 @code{asm} (@pxref{Alternate Keywords}).
7360 @subsubheading Qualifiers
7363 The optional @code{volatile} qualifier has no effect.
7364 All basic @code{asm} blocks are implicitly volatile.
7367 @subsubheading Parameters
7370 @item AssemblerInstructions
7371 This is a literal string that specifies the assembler code. The string can
7372 contain any instructions recognized by the assembler, including directives.
7373 GCC does not parse the assembler instructions themselves and
7374 does not know what they mean or even whether they are valid assembler input.
7376 You may place multiple assembler instructions together in a single @code{asm}
7377 string, separated by the characters normally used in assembly code for the
7378 system. A combination that works in most places is a newline to break the
7379 line, plus a tab character (written as @samp{\n\t}).
7380 Some assemblers allow semicolons as a line separator. However,
7381 note that some assembler dialects use semicolons to start a comment.
7384 @subsubheading Remarks
7385 Using extended @code{asm} typically produces smaller, safer, and more
7386 efficient code, and in most cases it is a better solution than basic
7387 @code{asm}. However, there are two situations where only basic @code{asm}
7392 Extended @code{asm} statements have to be inside a C
7393 function, so to write inline assembly language at file scope (``top-level''),
7394 outside of C functions, you must use basic @code{asm}.
7395 You can use this technique to emit assembler directives,
7396 define assembly language macros that can be invoked elsewhere in the file,
7397 or write entire functions in assembly language.
7401 with the @code{naked} attribute also require basic @code{asm}
7402 (@pxref{Function Attributes}).
7405 Safely accessing C data and calling functions from basic @code{asm} is more
7406 complex than it may appear. To access C data, it is better to use extended
7409 Do not expect a sequence of @code{asm} statements to remain perfectly
7410 consecutive after compilation. If certain instructions need to remain
7411 consecutive in the output, put them in a single multi-instruction @code{asm}
7412 statement. Note that GCC's optimizers can move @code{asm} statements
7413 relative to other code, including across jumps.
7415 @code{asm} statements may not perform jumps into other @code{asm} statements.
7416 GCC does not know about these jumps, and therefore cannot take
7417 account of them when deciding how to optimize. Jumps from @code{asm} to C
7418 labels are only supported in extended @code{asm}.
7420 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7421 assembly code when optimizing. This can lead to unexpected duplicate
7422 symbol errors during compilation if your assembly code defines symbols or
7425 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7426 visibility of any symbols it references. This may result in GCC discarding
7427 those symbols as unreferenced.
7429 The compiler copies the assembler instructions in a basic @code{asm}
7430 verbatim to the assembly language output file, without
7431 processing dialects or any of the @samp{%} operators that are available with
7432 extended @code{asm}. This results in minor differences between basic
7433 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7434 registers you might use @samp{%eax} in basic @code{asm} and
7435 @samp{%%eax} in extended @code{asm}.
7437 On targets such as x86 that support multiple assembler dialects,
7438 all basic @code{asm} blocks use the assembler dialect specified by the
7439 @option{-masm} command-line option (@pxref{x86 Options}).
7440 Basic @code{asm} provides no
7441 mechanism to provide different assembler strings for different dialects.
7443 Here is an example of basic @code{asm} for i386:
7446 /* Note that this code will not compile with -masm=intel */
7447 #define DebugBreak() asm("int $3")
7451 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7452 @cindex extended @code{asm}
7453 @cindex assembly language in C, extended
7455 With extended @code{asm} you can read and write C variables from
7456 assembler and perform jumps from assembler code to C labels.
7457 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7458 the operand parameters after the assembler template:
7461 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7462 : @var{OutputOperands}
7463 @r{[} : @var{InputOperands}
7464 @r{[} : @var{Clobbers} @r{]} @r{]})
7466 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7468 : @var{InputOperands}
7473 The @code{asm} keyword is a GNU extension.
7474 When writing code that can be compiled with @option{-ansi} and the
7475 various @option{-std} options, use @code{__asm__} instead of
7476 @code{asm} (@pxref{Alternate Keywords}).
7478 @subsubheading Qualifiers
7482 The typical use of extended @code{asm} statements is to manipulate input
7483 values to produce output values. However, your @code{asm} statements may
7484 also produce side effects. If so, you may need to use the @code{volatile}
7485 qualifier to disable certain optimizations. @xref{Volatile}.
7488 This qualifier informs the compiler that the @code{asm} statement may
7489 perform a jump to one of the labels listed in the @var{GotoLabels}.
7493 @subsubheading Parameters
7495 @item AssemblerTemplate
7496 This is a literal string that is the template for the assembler code. It is a
7497 combination of fixed text and tokens that refer to the input, output,
7498 and goto parameters. @xref{AssemblerTemplate}.
7500 @item OutputOperands
7501 A comma-separated list of the C variables modified by the instructions in the
7502 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7505 A comma-separated list of C expressions read by the instructions in the
7506 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7509 A comma-separated list of registers or other values changed by the
7510 @var{AssemblerTemplate}, beyond those listed as outputs.
7511 An empty list is permitted. @xref{Clobbers}.
7514 When you are using the @code{goto} form of @code{asm}, this section contains
7515 the list of all C labels to which the code in the
7516 @var{AssemblerTemplate} may jump.
7519 @code{asm} statements may not perform jumps into other @code{asm} statements,
7520 only to the listed @var{GotoLabels}.
7521 GCC's optimizers do not know about other jumps; therefore they cannot take
7522 account of them when deciding how to optimize.
7525 The total number of input + output + goto operands is limited to 30.
7527 @subsubheading Remarks
7528 The @code{asm} statement allows you to include assembly instructions directly
7529 within C code. This may help you to maximize performance in time-sensitive
7530 code or to access assembly instructions that are not readily available to C
7533 Note that extended @code{asm} statements must be inside a function. Only
7534 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7535 Functions declared with the @code{naked} attribute also require basic
7536 @code{asm} (@pxref{Function Attributes}).
7538 While the uses of @code{asm} are many and varied, it may help to think of an
7539 @code{asm} statement as a series of low-level instructions that convert input
7540 parameters to output parameters. So a simple (if not particularly useful)
7541 example for i386 using @code{asm} might look like this:
7547 asm ("mov %1, %0\n\t"
7552 printf("%d\n", dst);
7555 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7558 @subsubsection Volatile
7559 @cindex volatile @code{asm}
7560 @cindex @code{asm} volatile
7562 GCC's optimizers sometimes discard @code{asm} statements if they determine
7563 there is no need for the output variables. Also, the optimizers may move
7564 code out of loops if they believe that the code will always return the same
7565 result (i.e. none of its input values change between calls). Using the
7566 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7567 that have no output operands, including @code{asm goto} statements,
7568 are implicitly volatile.
7570 This i386 code demonstrates a case that does not use (or require) the
7571 @code{volatile} qualifier. If it is performing assertion checking, this code
7572 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7573 unreferenced by any code. As a result, the optimizers can discard the
7574 @code{asm} statement, which in turn removes the need for the entire
7575 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7576 isn't needed you allow the optimizers to produce the most efficient code
7580 void DoCheck(uint32_t dwSomeValue)
7584 // Assumes dwSomeValue is not zero.
7594 The next example shows a case where the optimizers can recognize that the input
7595 (@code{dwSomeValue}) never changes during the execution of the function and can
7596 therefore move the @code{asm} outside the loop to produce more efficient code.
7597 Again, using @code{volatile} disables this type of optimization.
7600 void do_print(uint32_t dwSomeValue)
7604 for (uint32_t x=0; x < 5; x++)
7606 // Assumes dwSomeValue is not zero.
7612 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7617 The following example demonstrates a case where you need to use the
7618 @code{volatile} qualifier.
7619 It uses the x86 @code{rdtsc} instruction, which reads
7620 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7621 the optimizers might assume that the @code{asm} block will always return the
7622 same value and therefore optimize away the second call.
7627 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7628 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7629 "or %%rdx, %0" // 'Or' in the lower bits.
7634 printf("msr: %llx\n", msr);
7638 // Reprint the timestamp
7639 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7640 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7641 "or %%rdx, %0" // 'Or' in the lower bits.
7646 printf("msr: %llx\n", msr);
7649 GCC's optimizers do not treat this code like the non-volatile code in the
7650 earlier examples. They do not move it out of loops or omit it on the
7651 assumption that the result from a previous call is still valid.
7653 Note that the compiler can move even volatile @code{asm} instructions relative
7654 to other code, including across jump instructions. For example, on many
7655 targets there is a system register that controls the rounding mode of
7656 floating-point operations. Setting it with a volatile @code{asm}, as in the
7657 following PowerPC example, does not work reliably.
7660 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7664 The compiler may move the addition back before the volatile @code{asm}. To
7665 make it work as expected, add an artificial dependency to the @code{asm} by
7666 referencing a variable in the subsequent code, for example:
7669 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7673 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7674 assembly code when optimizing. This can lead to unexpected duplicate symbol
7675 errors during compilation if your asm code defines symbols or labels.
7677 (@pxref{AssemblerTemplate}) may help resolve this problem.
7679 @anchor{AssemblerTemplate}
7680 @subsubsection Assembler Template
7681 @cindex @code{asm} assembler template
7683 An assembler template is a literal string containing assembler instructions.
7684 The compiler replaces tokens in the template that refer
7685 to inputs, outputs, and goto labels,
7686 and then outputs the resulting string to the assembler. The
7687 string can contain any instructions recognized by the assembler, including
7688 directives. GCC does not parse the assembler instructions
7689 themselves and does not know what they mean or even whether they are valid
7690 assembler input. However, it does count the statements
7691 (@pxref{Size of an asm}).
7693 You may place multiple assembler instructions together in a single @code{asm}
7694 string, separated by the characters normally used in assembly code for the
7695 system. A combination that works in most places is a newline to break the
7696 line, plus a tab character to move to the instruction field (written as
7698 Some assemblers allow semicolons as a line separator. However, note
7699 that some assembler dialects use semicolons to start a comment.
7701 Do not expect a sequence of @code{asm} statements to remain perfectly
7702 consecutive after compilation, even when you are using the @code{volatile}
7703 qualifier. If certain instructions need to remain consecutive in the output,
7704 put them in a single multi-instruction asm statement.
7706 Accessing data from C programs without using input/output operands (such as
7707 by using global symbols directly from the assembler template) may not work as
7708 expected. Similarly, calling functions directly from an assembler template
7709 requires a detailed understanding of the target assembler and ABI.
7711 Since GCC does not parse the assembler template,
7712 it has no visibility of any
7713 symbols it references. This may result in GCC discarding those symbols as
7714 unreferenced unless they are also listed as input, output, or goto operands.
7716 @subsubheading Special format strings
7718 In addition to the tokens described by the input, output, and goto operands,
7719 these tokens have special meanings in the assembler template:
7723 Outputs a single @samp{%} into the assembler code.
7726 Outputs a number that is unique to each instance of the @code{asm}
7727 statement in the entire compilation. This option is useful when creating local
7728 labels and referring to them multiple times in a single template that
7729 generates multiple assembler instructions.
7734 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7735 into the assembler code. When unescaped, these characters have special
7736 meaning to indicate multiple assembler dialects, as described below.
7739 @subsubheading Multiple assembler dialects in @code{asm} templates
7741 On targets such as x86, GCC supports multiple assembler dialects.
7742 The @option{-masm} option controls which dialect GCC uses as its
7743 default for inline assembler. The target-specific documentation for the
7744 @option{-masm} option contains the list of supported dialects, as well as the
7745 default dialect if the option is not specified. This information may be
7746 important to understand, since assembler code that works correctly when
7747 compiled using one dialect will likely fail if compiled using another.
7750 If your code needs to support multiple assembler dialects (for example, if
7751 you are writing public headers that need to support a variety of compilation
7752 options), use constructs of this form:
7755 @{ dialect0 | dialect1 | dialect2... @}
7758 This construct outputs @code{dialect0}
7759 when using dialect #0 to compile the code,
7760 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7761 braces than the number of dialects the compiler supports, the construct
7764 For example, if an x86 compiler supports two dialects
7765 (@samp{att}, @samp{intel}), an
7766 assembler template such as this:
7769 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7773 is equivalent to one of
7776 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7777 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7780 Using that same compiler, this code:
7783 "xchg@{l@}\t@{%%@}ebx, %1"
7787 corresponds to either
7790 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7791 "xchg\tebx, %1" @r{/* intel dialect */}
7794 There is no support for nesting dialect alternatives.
7796 @anchor{OutputOperands}
7797 @subsubsection Output Operands
7798 @cindex @code{asm} output operands
7800 An @code{asm} statement has zero or more output operands indicating the names
7801 of C variables modified by the assembler code.
7803 In this i386 example, @code{old} (referred to in the template string as
7804 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7805 (@code{%2}) is an input:
7810 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7811 "sbb %0,%0" // Use the CF to calculate old.
7812 : "=r" (old), "+rm" (*Base)
7819 Operands are separated by commas. Each operand has this format:
7822 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7826 @item asmSymbolicName
7827 Specifies a symbolic name for the operand.
7828 Reference the name in the assembler template
7829 by enclosing it in square brackets
7830 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7831 that contains the definition. Any valid C variable name is acceptable,
7832 including names already defined in the surrounding code. No two operands
7833 within the same @code{asm} statement can use the same symbolic name.
7835 When not using an @var{asmSymbolicName}, use the (zero-based) position
7837 in the list of operands in the assembler template. For example if there are
7838 three output operands, use @samp{%0} in the template to refer to the first,
7839 @samp{%1} for the second, and @samp{%2} for the third.
7842 A string constant specifying constraints on the placement of the operand;
7843 @xref{Constraints}, for details.
7845 Output constraints must begin with either @samp{=} (a variable overwriting an
7846 existing value) or @samp{+} (when reading and writing). When using
7847 @samp{=}, do not assume the location contains the existing value
7848 on entry to the @code{asm}, except
7849 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7851 After the prefix, there must be one or more additional constraints
7852 (@pxref{Constraints}) that describe where the value resides. Common
7853 constraints include @samp{r} for register and @samp{m} for memory.
7854 When you list more than one possible location (for example, @code{"=rm"}),
7855 the compiler chooses the most efficient one based on the current context.
7856 If you list as many alternates as the @code{asm} statement allows, you permit
7857 the optimizers to produce the best possible code.
7858 If you must use a specific register, but your Machine Constraints do not
7859 provide sufficient control to select the specific register you want,
7860 local register variables may provide a solution (@pxref{Local Register
7864 Specifies a C lvalue expression to hold the output, typically a variable name.
7865 The enclosing parentheses are a required part of the syntax.
7869 When the compiler selects the registers to use to
7870 represent the output operands, it does not use any of the clobbered registers
7873 Output operand expressions must be lvalues. The compiler cannot check whether
7874 the operands have data types that are reasonable for the instruction being
7875 executed. For output expressions that are not directly addressable (for
7876 example a bit-field), the constraint must allow a register. In that case, GCC
7877 uses the register as the output of the @code{asm}, and then stores that
7878 register into the output.
7880 Operands using the @samp{+} constraint modifier count as two operands
7881 (that is, both as input and output) towards the total maximum of 30 operands
7882 per @code{asm} statement.
7884 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7885 operands that must not overlap an input. Otherwise,
7886 GCC may allocate the output operand in the same register as an unrelated
7887 input operand, on the assumption that the assembler code consumes its
7888 inputs before producing outputs. This assumption may be false if the assembler
7889 code actually consists of more than one instruction.
7891 The same problem can occur if one output parameter (@var{a}) allows a register
7892 constraint and another output parameter (@var{b}) allows a memory constraint.
7893 The code generated by GCC to access the memory address in @var{b} can contain
7894 registers which @emph{might} be shared by @var{a}, and GCC considers those
7895 registers to be inputs to the asm. As above, GCC assumes that such input
7896 registers are consumed before any outputs are written. This assumption may
7897 result in incorrect behavior if the asm writes to @var{a} before using
7898 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7899 ensures that modifying @var{a} does not affect the address referenced by
7900 @var{b}. Otherwise, the location of @var{b}
7901 is undefined if @var{a} is modified before using @var{b}.
7903 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7904 instead of simply @samp{%2}). Typically these qualifiers are hardware
7905 dependent. The list of supported modifiers for x86 is found at
7906 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7908 If the C code that follows the @code{asm} makes no use of any of the output
7909 operands, use @code{volatile} for the @code{asm} statement to prevent the
7910 optimizers from discarding the @code{asm} statement as unneeded
7911 (see @ref{Volatile}).
7913 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7914 references the first output operand as @code{%0} (were there a second, it
7915 would be @code{%1}, etc). The number of the first input operand is one greater
7916 than that of the last output operand. In this i386 example, that makes
7917 @code{Mask} referenced as @code{%1}:
7920 uint32_t Mask = 1234;
7929 That code overwrites the variable @code{Index} (@samp{=}),
7930 placing the value in a register (@samp{r}).
7931 Using the generic @samp{r} constraint instead of a constraint for a specific
7932 register allows the compiler to pick the register to use, which can result
7933 in more efficient code. This may not be possible if an assembler instruction
7934 requires a specific register.
7936 The following i386 example uses the @var{asmSymbolicName} syntax.
7938 same result as the code above, but some may consider it more readable or more
7939 maintainable since reordering index numbers is not necessary when adding or
7940 removing operands. The names @code{aIndex} and @code{aMask}
7941 are only used in this example to emphasize which
7942 names get used where.
7943 It is acceptable to reuse the names @code{Index} and @code{Mask}.
7946 uint32_t Mask = 1234;
7949 asm ("bsfl %[aMask], %[aIndex]"
7950 : [aIndex] "=r" (Index)
7951 : [aMask] "r" (Mask)
7955 Here are some more examples of output operands.
7962 asm ("mov %[e], %[d]"
7967 Here, @code{d} may either be in a register or in memory. Since the compiler
7968 might already have the current value of the @code{uint32_t} location
7969 pointed to by @code{e}
7970 in a register, you can enable it to choose the best location
7971 for @code{d} by specifying both constraints.
7973 @anchor{FlagOutputOperands}
7974 @subsection Flag Output Operands
7975 @cindex @code{asm} flag output operands
7977 Some targets have a special register that holds the ``flags'' for the
7978 result of an operation or comparison. Normally, the contents of that
7979 register are either unmodifed by the asm, or the asm is considered to
7980 clobber the contents.
7982 On some targets, a special form of output operand exists by which
7983 conditions in the flags register may be outputs of the asm. The set of
7984 conditions supported are target specific, but the general rule is that
7985 the output variable must be a scalar integer, and the value will be boolean.
7986 When supported, the target will define the preprocessor symbol
7987 @code{__GCC_ASM_FLAG_OUTPUTS__}.
7989 Because of the special nature of the flag output operands, the constraint
7990 may not include alternatives.
7992 Most often, the target has only one flags register, and thus is an implied
7993 operand of many instructions. In this case, the operand should not be
7994 referenced within the assembler template via @code{%0} etc, as there's
7995 no corresponding text in the assembly language.
7999 The flag output constraints for the x86 family are of the form
8000 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8001 conditions defined in the ISA manual for @code{j@var{cc}} or
8006 ``above'' or unsigned greater than
8008 ``above or equal'' or unsigned greater than or equal
8010 ``below'' or unsigned less than
8012 ``below or equal'' or unsigned less than or equal
8017 ``equal'' or zero flag set
8021 signed greater than or equal
8025 signed less than or equal
8046 ``not'' @var{flag}, or inverted versions of those above
8051 @anchor{InputOperands}
8052 @subsubsection Input Operands
8053 @cindex @code{asm} input operands
8054 @cindex @code{asm} expressions
8056 Input operands make values from C variables and expressions available to the
8059 Operands are separated by commas. Each operand has this format:
8062 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8066 @item asmSymbolicName
8067 Specifies a symbolic name for the operand.
8068 Reference the name in the assembler template
8069 by enclosing it in square brackets
8070 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8071 that contains the definition. Any valid C variable name is acceptable,
8072 including names already defined in the surrounding code. No two operands
8073 within the same @code{asm} statement can use the same symbolic name.
8075 When not using an @var{asmSymbolicName}, use the (zero-based) position
8077 in the list of operands in the assembler template. For example if there are
8078 two output operands and three inputs,
8079 use @samp{%2} in the template to refer to the first input operand,
8080 @samp{%3} for the second, and @samp{%4} for the third.
8083 A string constant specifying constraints on the placement of the operand;
8084 @xref{Constraints}, for details.
8086 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8087 When you list more than one possible location (for example, @samp{"irm"}),
8088 the compiler chooses the most efficient one based on the current context.
8089 If you must use a specific register, but your Machine Constraints do not
8090 provide sufficient control to select the specific register you want,
8091 local register variables may provide a solution (@pxref{Local Register
8094 Input constraints can also be digits (for example, @code{"0"}). This indicates
8095 that the specified input must be in the same place as the output constraint
8096 at the (zero-based) index in the output constraint list.
8097 When using @var{asmSymbolicName} syntax for the output operands,
8098 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8101 This is the C variable or expression being passed to the @code{asm} statement
8102 as input. The enclosing parentheses are a required part of the syntax.
8106 When the compiler selects the registers to use to represent the input
8107 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8109 If there are no output operands but there are input operands, place two
8110 consecutive colons where the output operands would go:
8113 __asm__ ("some instructions"
8115 : "r" (Offset / 8));
8118 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8119 (except for inputs tied to outputs). The compiler assumes that on exit from
8120 the @code{asm} statement these operands contain the same values as they
8121 had before executing the statement.
8122 It is @emph{not} possible to use clobbers
8123 to inform the compiler that the values in these inputs are changing. One
8124 common work-around is to tie the changing input variable to an output variable
8125 that never gets used. Note, however, that if the code that follows the
8126 @code{asm} statement makes no use of any of the output operands, the GCC
8127 optimizers may discard the @code{asm} statement as unneeded
8128 (see @ref{Volatile}).
8130 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8131 instead of simply @samp{%2}). Typically these qualifiers are hardware
8132 dependent. The list of supported modifiers for x86 is found at
8133 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8135 In this example using the fictitious @code{combine} instruction, the
8136 constraint @code{"0"} for input operand 1 says that it must occupy the same
8137 location as output operand 0. Only input operands may use numbers in
8138 constraints, and they must each refer to an output operand. Only a number (or
8139 the symbolic assembler name) in the constraint can guarantee that one operand
8140 is in the same place as another. The mere fact that @code{foo} is the value of
8141 both operands is not enough to guarantee that they are in the same place in
8142 the generated assembler code.
8145 asm ("combine %2, %0"
8147 : "0" (foo), "g" (bar));
8150 Here is an example using symbolic names.
8153 asm ("cmoveq %1, %2, %[result]"
8154 : [result] "=r"(result)
8155 : "r" (test), "r" (new), "[result]" (old));
8159 @subsubsection Clobbers
8160 @cindex @code{asm} clobbers
8162 While the compiler is aware of changes to entries listed in the output
8163 operands, the inline @code{asm} code may modify more than just the outputs. For
8164 example, calculations may require additional registers, or the processor may
8165 overwrite a register as a side effect of a particular assembler instruction.
8166 In order to inform the compiler of these changes, list them in the clobber
8167 list. Clobber list items are either register names or the special clobbers
8168 (listed below). Each clobber list item is a string constant
8169 enclosed in double quotes and separated by commas.
8171 Clobber descriptions may not in any way overlap with an input or output
8172 operand. For example, you may not have an operand describing a register class
8173 with one member when listing that register in the clobber list. Variables
8174 declared to live in specific registers (@pxref{Explicit Register
8175 Variables}) and used
8176 as @code{asm} input or output operands must have no part mentioned in the
8177 clobber description. In particular, there is no way to specify that input
8178 operands get modified without also specifying them as output operands.
8180 When the compiler selects which registers to use to represent input and output
8181 operands, it does not use any of the clobbered registers. As a result,
8182 clobbered registers are available for any use in the assembler code.
8184 Here is a realistic example for the VAX showing the use of clobbered
8188 asm volatile ("movc3 %0, %1, %2"
8190 : "g" (from), "g" (to), "g" (count)
8191 : "r0", "r1", "r2", "r3", "r4", "r5");
8194 Also, there are two special clobber arguments:
8198 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8199 register. On some machines, GCC represents the condition codes as a specific
8200 hardware register; @code{"cc"} serves to name this register.
8201 On other machines, condition code handling is different,
8202 and specifying @code{"cc"} has no effect. But
8203 it is valid no matter what the target.
8206 The @code{"memory"} clobber tells the compiler that the assembly code
8208 reads or writes to items other than those listed in the input and output
8209 operands (for example, accessing the memory pointed to by one of the input
8210 parameters). To ensure memory contains correct values, GCC may need to flush
8211 specific register values to memory before executing the @code{asm}. Further,
8212 the compiler does not assume that any values read from memory before an
8213 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8215 Using the @code{"memory"} clobber effectively forms a read/write
8216 memory barrier for the compiler.
8218 Note that this clobber does not prevent the @emph{processor} from doing
8219 speculative reads past the @code{asm} statement. To prevent that, you need
8220 processor-specific fence instructions.
8222 Flushing registers to memory has performance implications and may be an issue
8223 for time-sensitive code. You can use a trick to avoid this if the size of
8224 the memory being accessed is known at compile time. For example, if accessing
8225 ten bytes of a string, use a memory input like:
8227 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8232 @subsubsection Goto Labels
8233 @cindex @code{asm} goto labels
8235 @code{asm goto} allows assembly code to jump to one or more C labels. The
8236 @var{GotoLabels} section in an @code{asm goto} statement contains
8238 list of all C labels to which the assembler code may jump. GCC assumes that
8239 @code{asm} execution falls through to the next statement (if this is not the
8240 case, consider using the @code{__builtin_unreachable} intrinsic after the
8241 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8242 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8245 An @code{asm goto} statement cannot have outputs.
8246 This is due to an internal restriction of
8247 the compiler: control transfer instructions cannot have outputs.
8248 If the assembler code does modify anything, use the @code{"memory"} clobber
8250 optimizers to flush all register values to memory and reload them if
8251 necessary after the @code{asm} statement.
8253 Also note that an @code{asm goto} statement is always implicitly
8254 considered volatile.
8256 To reference a label in the assembler template,
8257 prefix it with @samp{%l} (lowercase @samp{L}) followed
8258 by its (zero-based) position in @var{GotoLabels} plus the number of input
8259 operands. For example, if the @code{asm} has three inputs and references two
8260 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8262 Alternately, you can reference labels using the actual C label name enclosed
8263 in brackets. For example, to reference a label named @code{carry}, you can
8264 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8265 section when using this approach.
8267 Here is an example of @code{asm goto} for i386:
8274 : "r" (p1), "r" (p2)
8284 The following example shows an @code{asm goto} that uses a memory clobber.
8290 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8301 @anchor{x86Operandmodifiers}
8302 @subsubsection x86 Operand Modifiers
8304 References to input, output, and goto operands in the assembler template
8305 of extended @code{asm} statements can use
8306 modifiers to affect the way the operands are formatted in
8307 the code output to the assembler. For example, the
8308 following code uses the @samp{h} and @samp{b} modifiers for x86:
8312 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8316 These modifiers generate this assembler code:
8322 The rest of this discussion uses the following code for illustrative purposes.
8331 asm volatile goto ("some assembler instructions here"
8333 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8334 : /* No clobbers. */
8339 With no modifiers, this is what the output from the operands would be for the
8340 @samp{att} and @samp{intel} dialects of assembler:
8342 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8343 @headitem Operand @tab masm=att @tab masm=intel
8352 @tab @code{OFFSET FLAT:.L2}
8355 The table below shows the list of supported modifiers and their effects.
8357 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8358 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8360 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8365 @tab Print the QImode name of the register.
8370 @tab Print the QImode name for a ``high'' register.
8375 @tab Print the HImode name of the register.
8380 @tab Print the SImode name of the register.
8385 @tab Print the DImode name of the register.
8390 @tab Print the label name with no punctuation.
8395 @tab Require a constant operand and print the constant expression with no punctuation.
8401 @anchor{x86floatingpointasmoperands}
8402 @subsubsection x86 Floating-Point @code{asm} Operands
8404 On x86 targets, there are several rules on the usage of stack-like registers
8405 in the operands of an @code{asm}. These rules apply only to the operands
8406 that are stack-like registers:
8410 Given a set of input registers that die in an @code{asm}, it is
8411 necessary to know which are implicitly popped by the @code{asm}, and
8412 which must be explicitly popped by GCC@.
8414 An input register that is implicitly popped by the @code{asm} must be
8415 explicitly clobbered, unless it is constrained to match an
8419 For any input register that is implicitly popped by an @code{asm}, it is
8420 necessary to know how to adjust the stack to compensate for the pop.
8421 If any non-popped input is closer to the top of the reg-stack than
8422 the implicitly popped register, it would not be possible to know what the
8423 stack looked like---it's not clear how the rest of the stack ``slides
8426 All implicitly popped input registers must be closer to the top of
8427 the reg-stack than any input that is not implicitly popped.
8429 It is possible that if an input dies in an @code{asm}, the compiler might
8430 use the input register for an output reload. Consider this example:
8433 asm ("foo" : "=t" (a) : "f" (b));
8437 This code says that input @code{b} is not popped by the @code{asm}, and that
8438 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8439 deeper after the @code{asm} than it was before. But, it is possible that
8440 reload may think that it can use the same register for both the input and
8443 To prevent this from happening,
8444 if any input operand uses the @samp{f} constraint, all output register
8445 constraints must use the @samp{&} early-clobber modifier.
8447 The example above is correctly written as:
8450 asm ("foo" : "=&t" (a) : "f" (b));
8454 Some operands need to be in particular places on the stack. All
8455 output operands fall in this category---GCC has no other way to
8456 know which registers the outputs appear in unless you indicate
8457 this in the constraints.
8459 Output operands must specifically indicate which register an output
8460 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8461 constraints must select a class with a single register.
8464 Output operands may not be ``inserted'' between existing stack registers.
8465 Since no 387 opcode uses a read/write operand, all output operands
8466 are dead before the @code{asm}, and are pushed by the @code{asm}.
8467 It makes no sense to push anywhere but the top of the reg-stack.
8469 Output operands must start at the top of the reg-stack: output
8470 operands may not ``skip'' a register.
8473 Some @code{asm} statements may need extra stack space for internal
8474 calculations. This can be guaranteed by clobbering stack registers
8475 unrelated to the inputs and outputs.
8480 takes one input, which is internally popped, and produces two outputs.
8483 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8487 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8488 and replaces them with one output. The @code{st(1)} clobber is necessary
8489 for the compiler to know that @code{fyl2xp1} pops both inputs.
8492 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8500 @subsection Controlling Names Used in Assembler Code
8501 @cindex assembler names for identifiers
8502 @cindex names used in assembler code
8503 @cindex identifiers, names in assembler code
8505 You can specify the name to be used in the assembler code for a C
8506 function or variable by writing the @code{asm} (or @code{__asm__})
8507 keyword after the declarator.
8508 It is up to you to make sure that the assembler names you choose do not
8509 conflict with any other assembler symbols, or reference registers.
8511 @subsubheading Assembler names for data:
8513 This sample shows how to specify the assembler name for data:
8516 int foo asm ("myfoo") = 2;
8520 This specifies that the name to be used for the variable @code{foo} in
8521 the assembler code should be @samp{myfoo} rather than the usual
8524 On systems where an underscore is normally prepended to the name of a C
8525 variable, this feature allows you to define names for the
8526 linker that do not start with an underscore.
8528 GCC does not support using this feature with a non-static local variable
8529 since such variables do not have assembler names. If you are
8530 trying to put the variable in a particular register, see
8531 @ref{Explicit Register Variables}.
8533 @subsubheading Assembler names for functions:
8535 To specify the assembler name for functions, write a declaration for the
8536 function before its definition and put @code{asm} there, like this:
8539 int func (int x, int y) asm ("MYFUNC");
8541 int func (int x, int y)
8547 This specifies that the name to be used for the function @code{func} in
8548 the assembler code should be @code{MYFUNC}.
8550 @node Explicit Register Variables
8551 @subsection Variables in Specified Registers
8552 @anchor{Explicit Reg Vars}
8553 @cindex explicit register variables
8554 @cindex variables in specified registers
8555 @cindex specified registers
8557 GNU C allows you to associate specific hardware registers with C
8558 variables. In almost all cases, allowing the compiler to assign
8559 registers produces the best code. However under certain unusual
8560 circumstances, more precise control over the variable storage is
8563 Both global and local variables can be associated with a register. The
8564 consequences of performing this association are very different between
8565 the two, as explained in the sections below.
8568 * Global Register Variables:: Variables declared at global scope.
8569 * Local Register Variables:: Variables declared within a function.
8572 @node Global Register Variables
8573 @subsubsection Defining Global Register Variables
8574 @anchor{Global Reg Vars}
8575 @cindex global register variables
8576 @cindex registers, global variables in
8577 @cindex registers, global allocation
8579 You can define a global register variable and associate it with a specified
8583 register int *foo asm ("r12");
8587 Here @code{r12} is the name of the register that should be used. Note that
8588 this is the same syntax used for defining local register variables, but for
8589 a global variable the declaration appears outside a function. The
8590 @code{register} keyword is required, and cannot be combined with
8591 @code{static}. The register name must be a valid register name for the
8594 Registers are a scarce resource on most systems and allowing the
8595 compiler to manage their usage usually results in the best code. However,
8596 under special circumstances it can make sense to reserve some globally.
8597 For example this may be useful in programs such as programming language
8598 interpreters that have a couple of global variables that are accessed
8601 After defining a global register variable, for the current compilation
8605 @item The register is reserved entirely for this use, and will not be
8606 allocated for any other purpose.
8607 @item The register is not saved and restored by any functions.
8608 @item Stores into this register are never deleted even if they appear to be
8609 dead, but references may be deleted, moved or simplified.
8612 Note that these points @emph{only} apply to code that is compiled with the
8613 definition. The behavior of code that is merely linked in (for example
8614 code from libraries) is not affected.
8616 If you want to recompile source files that do not actually use your global
8617 register variable so they do not use the specified register for any other
8618 purpose, you need not actually add the global register declaration to
8619 their source code. It suffices to specify the compiler option
8620 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8623 @subsubheading Declaring the variable
8625 Global register variables can not have initial values, because an
8626 executable file has no means to supply initial contents for a register.
8628 When selecting a register, choose one that is normally saved and
8629 restored by function calls on your machine. This ensures that code
8630 which is unaware of this reservation (such as library routines) will
8631 restore it before returning.
8633 On machines with register windows, be sure to choose a global
8634 register that is not affected magically by the function call mechanism.
8636 @subsubheading Using the variable
8638 @cindex @code{qsort}, and global register variables
8639 When calling routines that are not aware of the reservation, be
8640 cautious if those routines call back into code which uses them. As an
8641 example, if you call the system library version of @code{qsort}, it may
8642 clobber your registers during execution, but (if you have selected
8643 appropriate registers) it will restore them before returning. However
8644 it will @emph{not} restore them before calling @code{qsort}'s comparison
8645 function. As a result, global values will not reliably be available to
8646 the comparison function unless the @code{qsort} function itself is rebuilt.
8648 Similarly, it is not safe to access the global register variables from signal
8649 handlers or from more than one thread of control. Unless you recompile
8650 them specially for the task at hand, the system library routines may
8651 temporarily use the register for other things.
8653 @cindex register variable after @code{longjmp}
8654 @cindex global register after @code{longjmp}
8655 @cindex value after @code{longjmp}
8658 On most machines, @code{longjmp} restores to each global register
8659 variable the value it had at the time of the @code{setjmp}. On some
8660 machines, however, @code{longjmp} does not change the value of global
8661 register variables. To be portable, the function that called @code{setjmp}
8662 should make other arrangements to save the values of the global register
8663 variables, and to restore them in a @code{longjmp}. This way, the same
8664 thing happens regardless of what @code{longjmp} does.
8666 Eventually there may be a way of asking the compiler to choose a register
8667 automatically, but first we need to figure out how it should choose and
8668 how to enable you to guide the choice. No solution is evident.
8670 @node Local Register Variables
8671 @subsubsection Specifying Registers for Local Variables
8672 @anchor{Local Reg Vars}
8673 @cindex local variables, specifying registers
8674 @cindex specifying registers for local variables
8675 @cindex registers for local variables
8677 You can define a local register variable and associate it with a specified
8681 register int *foo asm ("r12");
8685 Here @code{r12} is the name of the register that should be used. Note
8686 that this is the same syntax used for defining global register variables,
8687 but for a local variable the declaration appears within a function. The
8688 @code{register} keyword is required, and cannot be combined with
8689 @code{static}. The register name must be a valid register name for the
8692 As with global register variables, it is recommended that you choose
8693 a register that is normally saved and restored by function calls on your
8694 machine, so that calls to library routines will not clobber it.
8696 The only supported use for this feature is to specify registers
8697 for input and output operands when calling Extended @code{asm}
8698 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8699 particular machine don't provide sufficient control to select the desired
8700 register. To force an operand into a register, create a local variable
8701 and specify the register name after the variable's declaration. Then use
8702 the local variable for the @code{asm} operand and specify any constraint
8703 letter that matches the register:
8706 register int *p1 asm ("r0") = @dots{};
8707 register int *p2 asm ("r1") = @dots{};
8708 register int *result asm ("r0");
8709 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8712 @emph{Warning:} In the above example, be aware that a register (for example
8713 @code{r0}) can be call-clobbered by subsequent code, including function
8714 calls and library calls for arithmetic operators on other variables (for
8715 example the initialization of @code{p2}). In this case, use temporary
8716 variables for expressions between the register assignments:
8720 register int *p1 asm ("r0") = @dots{};
8721 register int *p2 asm ("r1") = t1;
8722 register int *result asm ("r0");
8723 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8726 Defining a register variable does not reserve the register. Other than
8727 when invoking the Extended @code{asm}, the contents of the specified
8728 register are not guaranteed. For this reason, the following uses
8729 are explicitly @emph{not} supported. If they appear to work, it is only
8730 happenstance, and may stop working as intended due to (seemingly)
8731 unrelated changes in surrounding code, or even minor changes in the
8732 optimization of a future version of gcc:
8735 @item Passing parameters to or from Basic @code{asm}
8736 @item Passing parameters to or from Extended @code{asm} without using input
8738 @item Passing parameters to or from routines written in assembler (or
8739 other languages) using non-standard calling conventions.
8742 Some developers use Local Register Variables in an attempt to improve
8743 gcc's allocation of registers, especially in large functions. In this
8744 case the register name is essentially a hint to the register allocator.
8745 While in some instances this can generate better code, improvements are
8746 subject to the whims of the allocator/optimizers. Since there are no
8747 guarantees that your improvements won't be lost, this usage of Local
8748 Register Variables is discouraged.
8750 On the MIPS platform, there is related use for local register variables
8751 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8752 Defining coprocessor specifics for MIPS targets, gccint,
8753 GNU Compiler Collection (GCC) Internals}).
8755 @node Size of an asm
8756 @subsection Size of an @code{asm}
8758 Some targets require that GCC track the size of each instruction used
8759 in order to generate correct code. Because the final length of the
8760 code produced by an @code{asm} statement is only known by the
8761 assembler, GCC must make an estimate as to how big it will be. It
8762 does this by counting the number of instructions in the pattern of the
8763 @code{asm} and multiplying that by the length of the longest
8764 instruction supported by that processor. (When working out the number
8765 of instructions, it assumes that any occurrence of a newline or of
8766 whatever statement separator character is supported by the assembler --
8767 typically @samp{;} --- indicates the end of an instruction.)
8769 Normally, GCC's estimate is adequate to ensure that correct
8770 code is generated, but it is possible to confuse the compiler if you use
8771 pseudo instructions or assembler macros that expand into multiple real
8772 instructions, or if you use assembler directives that expand to more
8773 space in the object file than is needed for a single instruction.
8774 If this happens then the assembler may produce a diagnostic saying that
8775 a label is unreachable.
8777 @node Alternate Keywords
8778 @section Alternate Keywords
8779 @cindex alternate keywords
8780 @cindex keywords, alternate
8782 @option{-ansi} and the various @option{-std} options disable certain
8783 keywords. This causes trouble when you want to use GNU C extensions, or
8784 a general-purpose header file that should be usable by all programs,
8785 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8786 @code{inline} are not available in programs compiled with
8787 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8788 program compiled with @option{-std=c99} or @option{-std=c11}). The
8790 @code{restrict} is only available when @option{-std=gnu99} (which will
8791 eventually be the default) or @option{-std=c99} (or the equivalent
8792 @option{-std=iso9899:1999}), or an option for a later standard
8795 The way to solve these problems is to put @samp{__} at the beginning and
8796 end of each problematical keyword. For example, use @code{__asm__}
8797 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8799 Other C compilers won't accept these alternative keywords; if you want to
8800 compile with another compiler, you can define the alternate keywords as
8801 macros to replace them with the customary keywords. It looks like this:
8809 @findex __extension__
8811 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8813 prevent such warnings within one expression by writing
8814 @code{__extension__} before the expression. @code{__extension__} has no
8815 effect aside from this.
8817 @node Incomplete Enums
8818 @section Incomplete @code{enum} Types
8820 You can define an @code{enum} tag without specifying its possible values.
8821 This results in an incomplete type, much like what you get if you write
8822 @code{struct foo} without describing the elements. A later declaration
8823 that does specify the possible values completes the type.
8825 You can't allocate variables or storage using the type while it is
8826 incomplete. However, you can work with pointers to that type.
8828 This extension may not be very useful, but it makes the handling of
8829 @code{enum} more consistent with the way @code{struct} and @code{union}
8832 This extension is not supported by GNU C++.
8834 @node Function Names
8835 @section Function Names as Strings
8836 @cindex @code{__func__} identifier
8837 @cindex @code{__FUNCTION__} identifier
8838 @cindex @code{__PRETTY_FUNCTION__} identifier
8840 GCC provides three magic variables that hold the name of the current
8841 function, as a string. The first of these is @code{__func__}, which
8842 is part of the C99 standard:
8844 The identifier @code{__func__} is implicitly declared by the translator
8845 as if, immediately following the opening brace of each function
8846 definition, the declaration
8849 static const char __func__[] = "function-name";
8853 appeared, where function-name is the name of the lexically-enclosing
8854 function. This name is the unadorned name of the function.
8856 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8857 backward compatibility with old versions of GCC.
8859 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8860 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8861 the type signature of the function as well as its bare name. For
8862 example, this program:
8866 extern int printf (char *, ...);
8873 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8874 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8892 __PRETTY_FUNCTION__ = void a::sub(int)
8895 These identifiers are variables, not preprocessor macros, and may not
8896 be used to initialize @code{char} arrays or be concatenated with other string
8899 @node Return Address
8900 @section Getting the Return or Frame Address of a Function
8902 These functions may be used to get information about the callers of a
8905 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8906 This function returns the return address of the current function, or of
8907 one of its callers. The @var{level} argument is number of frames to
8908 scan up the call stack. A value of @code{0} yields the return address
8909 of the current function, a value of @code{1} yields the return address
8910 of the caller of the current function, and so forth. When inlining
8911 the expected behavior is that the function returns the address of
8912 the function that is returned to. To work around this behavior use
8913 the @code{noinline} function attribute.
8915 The @var{level} argument must be a constant integer.
8917 On some machines it may be impossible to determine the return address of
8918 any function other than the current one; in such cases, or when the top
8919 of the stack has been reached, this function returns @code{0} or a
8920 random value. In addition, @code{__builtin_frame_address} may be used
8921 to determine if the top of the stack has been reached.
8923 Additional post-processing of the returned value may be needed, see
8924 @code{__builtin_extract_return_addr}.
8926 Calling this function with a nonzero argument can have unpredictable
8927 effects, including crashing the calling program. As a result, calls
8928 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8929 option is in effect. Such calls should only be made in debugging
8933 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
8934 The address as returned by @code{__builtin_return_address} may have to be fed
8935 through this function to get the actual encoded address. For example, on the
8936 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
8937 platforms an offset has to be added for the true next instruction to be
8940 If no fixup is needed, this function simply passes through @var{addr}.
8943 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
8944 This function does the reverse of @code{__builtin_extract_return_addr}.
8947 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
8948 This function is similar to @code{__builtin_return_address}, but it
8949 returns the address of the function frame rather than the return address
8950 of the function. Calling @code{__builtin_frame_address} with a value of
8951 @code{0} yields the frame address of the current function, a value of
8952 @code{1} yields the frame address of the caller of the current function,
8955 The frame is the area on the stack that holds local variables and saved
8956 registers. The frame address is normally the address of the first word
8957 pushed on to the stack by the function. However, the exact definition
8958 depends upon the processor and the calling convention. If the processor
8959 has a dedicated frame pointer register, and the function has a frame,
8960 then @code{__builtin_frame_address} returns the value of the frame
8963 On some machines it may be impossible to determine the frame address of
8964 any function other than the current one; in such cases, or when the top
8965 of the stack has been reached, this function returns @code{0} if
8966 the first frame pointer is properly initialized by the startup code.
8968 Calling this function with a nonzero argument can have unpredictable
8969 effects, including crashing the calling program. As a result, calls
8970 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8971 option is in effect. Such calls should only be made in debugging
8975 @node Vector Extensions
8976 @section Using Vector Instructions through Built-in Functions
8978 On some targets, the instruction set contains SIMD vector instructions which
8979 operate on multiple values contained in one large register at the same time.
8980 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
8983 The first step in using these extensions is to provide the necessary data
8984 types. This should be done using an appropriate @code{typedef}:
8987 typedef int v4si __attribute__ ((vector_size (16)));
8991 The @code{int} type specifies the base type, while the attribute specifies
8992 the vector size for the variable, measured in bytes. For example, the
8993 declaration above causes the compiler to set the mode for the @code{v4si}
8994 type to be 16 bytes wide and divided into @code{int} sized units. For
8995 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
8996 corresponding mode of @code{foo} is @acronym{V4SI}.
8998 The @code{vector_size} attribute is only applicable to integral and
8999 float scalars, although arrays, pointers, and function return values
9000 are allowed in conjunction with this construct. Only sizes that are
9001 a power of two are currently allowed.
9003 All the basic integer types can be used as base types, both as signed
9004 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9005 @code{long long}. In addition, @code{float} and @code{double} can be
9006 used to build floating-point vector types.
9008 Specifying a combination that is not valid for the current architecture
9009 causes GCC to synthesize the instructions using a narrower mode.
9010 For example, if you specify a variable of type @code{V4SI} and your
9011 architecture does not allow for this specific SIMD type, GCC
9012 produces code that uses 4 @code{SIs}.
9014 The types defined in this manner can be used with a subset of normal C
9015 operations. Currently, GCC allows using the following operators
9016 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9018 The operations behave like C++ @code{valarrays}. Addition is defined as
9019 the addition of the corresponding elements of the operands. For
9020 example, in the code below, each of the 4 elements in @var{a} is
9021 added to the corresponding 4 elements in @var{b} and the resulting
9022 vector is stored in @var{c}.
9025 typedef int v4si __attribute__ ((vector_size (16)));
9032 Subtraction, multiplication, division, and the logical operations
9033 operate in a similar manner. Likewise, the result of using the unary
9034 minus or complement operators on a vector type is a vector whose
9035 elements are the negative or complemented values of the corresponding
9036 elements in the operand.
9038 It is possible to use shifting operators @code{<<}, @code{>>} on
9039 integer-type vectors. The operation is defined as following: @code{@{a0,
9040 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9041 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9044 For convenience, it is allowed to use a binary vector operation
9045 where one operand is a scalar. In that case the compiler transforms
9046 the scalar operand into a vector where each element is the scalar from
9047 the operation. The transformation happens only if the scalar could be
9048 safely converted to the vector-element type.
9049 Consider the following code.
9052 typedef int v4si __attribute__ ((vector_size (16)));
9057 a = b + 1; /* a = b + @{1,1,1,1@}; */
9058 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9060 a = l + a; /* Error, cannot convert long to int. */
9063 Vectors can be subscripted as if the vector were an array with
9064 the same number of elements and base type. Out of bound accesses
9065 invoke undefined behavior at run time. Warnings for out of bound
9066 accesses for vector subscription can be enabled with
9067 @option{-Warray-bounds}.
9069 Vector comparison is supported with standard comparison
9070 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9071 vector expressions of integer-type or real-type. Comparison between
9072 integer-type vectors and real-type vectors are not supported. The
9073 result of the comparison is a vector of the same width and number of
9074 elements as the comparison operands with a signed integral element
9077 Vectors are compared element-wise producing 0 when comparison is false
9078 and -1 (constant of the appropriate type where all bits are set)
9079 otherwise. Consider the following example.
9082 typedef int v4si __attribute__ ((vector_size (16)));
9084 v4si a = @{1,2,3,4@};
9085 v4si b = @{3,2,1,4@};
9088 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9089 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9092 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9093 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9094 integer vector with the same number of elements of the same size as @code{b}
9095 and @code{c}, computes all three arguments and creates a vector
9096 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9097 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9098 As in the case of binary operations, this syntax is also accepted when
9099 one of @code{b} or @code{c} is a scalar that is then transformed into a
9100 vector. If both @code{b} and @code{c} are scalars and the type of
9101 @code{true?b:c} has the same size as the element type of @code{a}, then
9102 @code{b} and @code{c} are converted to a vector type whose elements have
9103 this type and with the same number of elements as @code{a}.
9105 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9106 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9107 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9108 For mixed operations between a scalar @code{s} and a vector @code{v},
9109 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9110 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9112 Vector shuffling is available using functions
9113 @code{__builtin_shuffle (vec, mask)} and
9114 @code{__builtin_shuffle (vec0, vec1, mask)}.
9115 Both functions construct a permutation of elements from one or two
9116 vectors and return a vector of the same type as the input vector(s).
9117 The @var{mask} is an integral vector with the same width (@var{W})
9118 and element count (@var{N}) as the output vector.
9120 The elements of the input vectors are numbered in memory ordering of
9121 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9122 elements of @var{mask} are considered modulo @var{N} in the single-operand
9123 case and modulo @math{2*@var{N}} in the two-operand case.
9125 Consider the following example,
9128 typedef int v4si __attribute__ ((vector_size (16)));
9130 v4si a = @{1,2,3,4@};
9131 v4si b = @{5,6,7,8@};
9132 v4si mask1 = @{0,1,1,3@};
9133 v4si mask2 = @{0,4,2,5@};
9136 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9137 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9140 Note that @code{__builtin_shuffle} is intentionally semantically
9141 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9143 You can declare variables and use them in function calls and returns, as
9144 well as in assignments and some casts. You can specify a vector type as
9145 a return type for a function. Vector types can also be used as function
9146 arguments. It is possible to cast from one vector type to another,
9147 provided they are of the same size (in fact, you can also cast vectors
9148 to and from other datatypes of the same size).
9150 You cannot operate between vectors of different lengths or different
9151 signedness without a cast.
9154 @section Support for @code{offsetof}
9155 @findex __builtin_offsetof
9157 GCC implements for both C and C++ a syntactic extension to implement
9158 the @code{offsetof} macro.
9162 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9164 offsetof_member_designator:
9166 | offsetof_member_designator "." @code{identifier}
9167 | offsetof_member_designator "[" @code{expr} "]"
9170 This extension is sufficient such that
9173 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9177 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9178 may be dependent. In either case, @var{member} may consist of a single
9179 identifier, or a sequence of member accesses and array references.
9181 @node __sync Builtins
9182 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9184 The following built-in functions
9185 are intended to be compatible with those described
9186 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9187 section 7.4. As such, they depart from normal GCC practice by not using
9188 the @samp{__builtin_} prefix and also by being overloaded so that they
9189 work on multiple types.
9191 The definition given in the Intel documentation allows only for the use of
9192 the types @code{int}, @code{long}, @code{long long} or their unsigned
9193 counterparts. GCC allows any integral scalar or pointer type that is
9194 1, 2, 4 or 8 bytes in length.
9196 These functions are implemented in terms of the @samp{__atomic}
9197 builtins (@pxref{__atomic Builtins}). They should not be used for new
9198 code which should use the @samp{__atomic} builtins instead.
9200 Not all operations are supported by all target processors. If a particular
9201 operation cannot be implemented on the target processor, a warning is
9202 generated and a call to an external function is generated. The external
9203 function carries the same name as the built-in version,
9204 with an additional suffix
9205 @samp{_@var{n}} where @var{n} is the size of the data type.
9207 @c ??? Should we have a mechanism to suppress this warning? This is almost
9208 @c useful for implementing the operation under the control of an external
9211 In most cases, these built-in functions are considered a @dfn{full barrier}.
9213 no memory operand is moved across the operation, either forward or
9214 backward. Further, instructions are issued as necessary to prevent the
9215 processor from speculating loads across the operation and from queuing stores
9216 after the operation.
9218 All of the routines are described in the Intel documentation to take
9219 ``an optional list of variables protected by the memory barrier''. It's
9220 not clear what is meant by that; it could mean that @emph{only} the
9221 listed variables are protected, or it could mean a list of additional
9222 variables to be protected. The list is ignored by GCC which treats it as
9223 empty. GCC interprets an empty list as meaning that all globally
9224 accessible variables should be protected.
9227 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9228 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9229 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9230 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9231 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9232 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9233 @findex __sync_fetch_and_add
9234 @findex __sync_fetch_and_sub
9235 @findex __sync_fetch_and_or
9236 @findex __sync_fetch_and_and
9237 @findex __sync_fetch_and_xor
9238 @findex __sync_fetch_and_nand
9239 These built-in functions perform the operation suggested by the name, and
9240 returns the value that had previously been in memory. That is,
9243 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9244 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9247 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9248 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9250 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9251 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9252 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9253 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9254 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9255 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9256 @findex __sync_add_and_fetch
9257 @findex __sync_sub_and_fetch
9258 @findex __sync_or_and_fetch
9259 @findex __sync_and_and_fetch
9260 @findex __sync_xor_and_fetch
9261 @findex __sync_nand_and_fetch
9262 These built-in functions perform the operation suggested by the name, and
9263 return the new value. That is,
9266 @{ *ptr @var{op}= value; return *ptr; @}
9267 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9270 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9271 as @code{*ptr = ~(*ptr & value)} instead of
9272 @code{*ptr = ~*ptr & value}.
9274 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9275 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9276 @findex __sync_bool_compare_and_swap
9277 @findex __sync_val_compare_and_swap
9278 These built-in functions perform an atomic compare and swap.
9279 That is, if the current
9280 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9283 The ``bool'' version returns true if the comparison is successful and
9284 @var{newval} is written. The ``val'' version returns the contents
9285 of @code{*@var{ptr}} before the operation.
9287 @item __sync_synchronize (...)
9288 @findex __sync_synchronize
9289 This built-in function issues a full memory barrier.
9291 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9292 @findex __sync_lock_test_and_set
9293 This built-in function, as described by Intel, is not a traditional test-and-set
9294 operation, but rather an atomic exchange operation. It writes @var{value}
9295 into @code{*@var{ptr}}, and returns the previous contents of
9298 Many targets have only minimal support for such locks, and do not support
9299 a full exchange operation. In this case, a target may support reduced
9300 functionality here by which the @emph{only} valid value to store is the
9301 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9302 is implementation defined.
9304 This built-in function is not a full barrier,
9305 but rather an @dfn{acquire barrier}.
9306 This means that references after the operation cannot move to (or be
9307 speculated to) before the operation, but previous memory stores may not
9308 be globally visible yet, and previous memory loads may not yet be
9311 @item void __sync_lock_release (@var{type} *ptr, ...)
9312 @findex __sync_lock_release
9313 This built-in function releases the lock acquired by
9314 @code{__sync_lock_test_and_set}.
9315 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9317 This built-in function is not a full barrier,
9318 but rather a @dfn{release barrier}.
9319 This means that all previous memory stores are globally visible, and all
9320 previous memory loads have been satisfied, but following memory reads
9321 are not prevented from being speculated to before the barrier.
9324 @node __atomic Builtins
9325 @section Built-in Functions for Memory Model Aware Atomic Operations
9327 The following built-in functions approximately match the requirements
9328 for the C++11 memory model. They are all
9329 identified by being prefixed with @samp{__atomic} and most are
9330 overloaded so that they work with multiple types.
9332 These functions are intended to replace the legacy @samp{__sync}
9333 builtins. The main difference is that the memory order that is requested
9334 is a parameter to the functions. New code should always use the
9335 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9337 Note that the @samp{__atomic} builtins assume that programs will
9338 conform to the C++11 memory model. In particular, they assume
9339 that programs are free of data races. See the C++11 standard for
9340 detailed requirements.
9342 The @samp{__atomic} builtins can be used with any integral scalar or
9343 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9344 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9345 supported by the architecture.
9347 The four non-arithmetic functions (load, store, exchange, and
9348 compare_exchange) all have a generic version as well. This generic
9349 version works on any data type. It uses the lock-free built-in function
9350 if the specific data type size makes that possible; otherwise, an
9351 external call is left to be resolved at run time. This external call is
9352 the same format with the addition of a @samp{size_t} parameter inserted
9353 as the first parameter indicating the size of the object being pointed to.
9354 All objects must be the same size.
9356 There are 6 different memory orders that can be specified. These map
9357 to the C++11 memory orders with the same names, see the C++11 standard
9358 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9359 on atomic synchronization} for detailed definitions. Individual
9360 targets may also support additional memory orders for use on specific
9361 architectures. Refer to the target documentation for details of
9364 An atomic operation can both constrain code motion and
9365 be mapped to hardware instructions for synchronization between threads
9366 (e.g., a fence). To which extent this happens is controlled by the
9367 memory orders, which are listed here in approximately ascending order of
9368 strength. The description of each memory order is only meant to roughly
9369 illustrate the effects and is not a specification; see the C++11
9370 memory model for precise semantics.
9373 @item __ATOMIC_RELAXED
9374 Implies no inter-thread ordering constraints.
9375 @item __ATOMIC_CONSUME
9376 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9377 memory order because of a deficiency in C++11's semantics for
9378 @code{memory_order_consume}.
9379 @item __ATOMIC_ACQUIRE
9380 Creates an inter-thread happens-before constraint from the release (or
9381 stronger) semantic store to this acquire load. Can prevent hoisting
9382 of code to before the operation.
9383 @item __ATOMIC_RELEASE
9384 Creates an inter-thread happens-before constraint to acquire (or stronger)
9385 semantic loads that read from this release store. Can prevent sinking
9386 of code to after the operation.
9387 @item __ATOMIC_ACQ_REL
9388 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9389 @code{__ATOMIC_RELEASE}.
9390 @item __ATOMIC_SEQ_CST
9391 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9394 Note that in the C++11 memory model, @emph{fences} (e.g.,
9395 @samp{__atomic_thread_fence}) take effect in combination with other
9396 atomic operations on specific memory locations (e.g., atomic loads);
9397 operations on specific memory locations do not necessarily affect other
9398 operations in the same way.
9400 Target architectures are encouraged to provide their own patterns for
9401 each of the atomic built-in functions. If no target is provided, the original
9402 non-memory model set of @samp{__sync} atomic built-in functions are
9403 used, along with any required synchronization fences surrounding it in
9404 order to achieve the proper behavior. Execution in this case is subject
9405 to the same restrictions as those built-in functions.
9407 If there is no pattern or mechanism to provide a lock-free instruction
9408 sequence, a call is made to an external routine with the same parameters
9409 to be resolved at run time.
9411 When implementing patterns for these built-in functions, the memory order
9412 parameter can be ignored as long as the pattern implements the most
9413 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9414 orders execute correctly with this memory order but they may not execute as
9415 efficiently as they could with a more appropriate implementation of the
9416 relaxed requirements.
9418 Note that the C++11 standard allows for the memory order parameter to be
9419 determined at run time rather than at compile time. These built-in
9420 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9421 than invoke a runtime library call or inline a switch statement. This is
9422 standard compliant, safe, and the simplest approach for now.
9424 The memory order parameter is a signed int, but only the lower 16 bits are
9425 reserved for the memory order. The remainder of the signed int is reserved
9426 for target use and should be 0. Use of the predefined atomic values
9427 ensures proper usage.
9429 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9430 This built-in function implements an atomic load operation. It returns the
9431 contents of @code{*@var{ptr}}.
9433 The valid memory order variants are
9434 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9435 and @code{__ATOMIC_CONSUME}.
9439 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9440 This is the generic version of an atomic load. It returns the
9441 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9445 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9446 This built-in function implements an atomic store operation. It writes
9447 @code{@var{val}} into @code{*@var{ptr}}.
9449 The valid memory order variants are
9450 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9454 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9455 This is the generic version of an atomic store. It stores the value
9456 of @code{*@var{val}} into @code{*@var{ptr}}.
9460 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9461 This built-in function implements an atomic exchange operation. It writes
9462 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9465 The valid memory order variants are
9466 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9467 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9471 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9472 This is the generic version of an atomic exchange. It stores the
9473 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9474 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9478 @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)
9479 This built-in function implements an atomic compare and exchange operation.
9480 This compares the contents of @code{*@var{ptr}} with the contents of
9481 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9482 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9483 equal, the operation is a @emph{read} and the current contents of
9484 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
9485 for weak compare_exchange, and false for the strong variation. Many targets
9486 only offer the strong variation and ignore the parameter. When in doubt, use
9487 the strong variation.
9489 True is returned if @var{desired} is written into
9490 @code{*@var{ptr}} and the operation is considered to conform to the
9491 memory order specified by @var{success_memorder}. There are no
9492 restrictions on what memory order can be used here.
9494 False is returned otherwise, and the operation is considered to conform
9495 to @var{failure_memorder}. This memory order cannot be
9496 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9497 stronger order than that specified by @var{success_memorder}.
9501 @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)
9502 This built-in function implements the generic version of
9503 @code{__atomic_compare_exchange}. The function is virtually identical to
9504 @code{__atomic_compare_exchange_n}, except the desired value is also a
9509 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9510 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9511 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9512 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9513 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9514 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9515 These built-in functions perform the operation suggested by the name, and
9516 return the result of the operation. That is,
9519 @{ *ptr @var{op}= val; return *ptr; @}
9522 All memory orders are valid.
9526 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9527 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9528 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9529 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9530 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9531 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9532 These built-in functions perform the operation suggested by the name, and
9533 return the value that had previously been in @code{*@var{ptr}}. That is,
9536 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9539 All memory orders are valid.
9543 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9545 This built-in function performs an atomic test-and-set operation on
9546 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9547 defined nonzero ``set'' value and the return value is @code{true} if and only
9548 if the previous contents were ``set''.
9549 It should be only used for operands of type @code{bool} or @code{char}. For
9550 other types only part of the value may be set.
9552 All memory orders are valid.
9556 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9558 This built-in function performs an atomic clear operation on
9559 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9560 It should be only used for operands of type @code{bool} or @code{char} and
9561 in conjunction with @code{__atomic_test_and_set}.
9562 For other types it may only clear partially. If the type is not @code{bool}
9563 prefer using @code{__atomic_store}.
9565 The valid memory order variants are
9566 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9567 @code{__ATOMIC_RELEASE}.
9571 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9573 This built-in function acts as a synchronization fence between threads
9574 based on the specified memory order.
9576 All memory orders are valid.
9580 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9582 This built-in function acts as a synchronization fence between a thread
9583 and signal handlers based in the same thread.
9585 All memory orders are valid.
9589 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9591 This built-in function returns true if objects of @var{size} bytes always
9592 generate lock-free atomic instructions for the target architecture.
9593 @var{size} must resolve to a compile-time constant and the result also
9594 resolves to a compile-time constant.
9596 @var{ptr} is an optional pointer to the object that may be used to determine
9597 alignment. A value of 0 indicates typical alignment should be used. The
9598 compiler may also ignore this parameter.
9601 if (_atomic_always_lock_free (sizeof (long long), 0))
9606 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9608 This built-in function returns true if objects of @var{size} bytes always
9609 generate lock-free atomic instructions for the target architecture. If
9610 the built-in function is not known to be lock-free, a call is made to a
9611 runtime routine named @code{__atomic_is_lock_free}.
9613 @var{ptr} is an optional pointer to the object that may be used to determine
9614 alignment. A value of 0 indicates typical alignment should be used. The
9615 compiler may also ignore this parameter.
9618 @node Integer Overflow Builtins
9619 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9621 The following built-in functions allow performing simple arithmetic operations
9622 together with checking whether the operations overflowed.
9624 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9625 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9626 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9627 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9628 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9629 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9630 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9632 These built-in functions promote the first two operands into infinite precision signed
9633 type and perform addition on those promoted operands. The result is then
9634 cast to the type the third pointer argument points to and stored there.
9635 If the stored result is equal to the infinite precision result, the built-in
9636 functions return false, otherwise they return true. As the addition is
9637 performed in infinite signed precision, these built-in functions have fully defined
9638 behavior for all argument values.
9640 The first built-in function allows arbitrary integral types for operands and
9641 the result type must be pointer to some integer type, the rest of the built-in
9642 functions have explicit integer types.
9644 The compiler will attempt to use hardware instructions to implement
9645 these built-in functions where possible, like conditional jump on overflow
9646 after addition, conditional jump on carry etc.
9650 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9651 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9652 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9653 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9654 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9655 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9656 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9658 These built-in functions are similar to the add overflow checking built-in
9659 functions above, except they perform subtraction, subtract the second argument
9660 from the first one, instead of addition.
9664 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9665 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9666 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9667 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9668 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9669 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9670 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9672 These built-in functions are similar to the add overflow checking built-in
9673 functions above, except they perform multiplication, instead of addition.
9677 @node x86 specific memory model extensions for transactional memory
9678 @section x86-Specific Memory Model Extensions for Transactional Memory
9680 The x86 architecture supports additional memory ordering flags
9681 to mark lock critical sections for hardware lock elision.
9682 These must be specified in addition to an existing memory order to
9686 @item __ATOMIC_HLE_ACQUIRE
9687 Start lock elision on a lock variable.
9688 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9689 @item __ATOMIC_HLE_RELEASE
9690 End lock elision on a lock variable.
9691 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9694 When a lock acquire fails, it is required for good performance to abort
9695 the transaction quickly. This can be done with a @code{_mm_pause}.
9698 #include <immintrin.h> // For _mm_pause
9702 /* Acquire lock with lock elision */
9703 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9704 _mm_pause(); /* Abort failed transaction */
9706 /* Free lock with lock elision */
9707 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9710 @node Object Size Checking
9711 @section Object Size Checking Built-in Functions
9712 @findex __builtin_object_size
9713 @findex __builtin___memcpy_chk
9714 @findex __builtin___mempcpy_chk
9715 @findex __builtin___memmove_chk
9716 @findex __builtin___memset_chk
9717 @findex __builtin___strcpy_chk
9718 @findex __builtin___stpcpy_chk
9719 @findex __builtin___strncpy_chk
9720 @findex __builtin___strcat_chk
9721 @findex __builtin___strncat_chk
9722 @findex __builtin___sprintf_chk
9723 @findex __builtin___snprintf_chk
9724 @findex __builtin___vsprintf_chk
9725 @findex __builtin___vsnprintf_chk
9726 @findex __builtin___printf_chk
9727 @findex __builtin___vprintf_chk
9728 @findex __builtin___fprintf_chk
9729 @findex __builtin___vfprintf_chk
9731 GCC implements a limited buffer overflow protection mechanism
9732 that can prevent some buffer overflow attacks.
9734 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9735 is a built-in construct that returns a constant number of bytes from
9736 @var{ptr} to the end of the object @var{ptr} pointer points to
9737 (if known at compile time). @code{__builtin_object_size} never evaluates
9738 its arguments for side-effects. If there are any side-effects in them, it
9739 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9740 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9741 point to and all of them are known at compile time, the returned number
9742 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9743 0 and minimum if nonzero. If it is not possible to determine which objects
9744 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9745 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9746 for @var{type} 2 or 3.
9748 @var{type} is an integer constant from 0 to 3. If the least significant
9749 bit is clear, objects are whole variables, if it is set, a closest
9750 surrounding subobject is considered the object a pointer points to.
9751 The second bit determines if maximum or minimum of remaining bytes
9755 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9756 char *p = &var.buf1[1], *q = &var.b;
9758 /* Here the object p points to is var. */
9759 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9760 /* The subobject p points to is var.buf1. */
9761 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9762 /* The object q points to is var. */
9763 assert (__builtin_object_size (q, 0)
9764 == (char *) (&var + 1) - (char *) &var.b);
9765 /* The subobject q points to is var.b. */
9766 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9770 There are built-in functions added for many common string operation
9771 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9772 built-in is provided. This built-in has an additional last argument,
9773 which is the number of bytes remaining in object the @var{dest}
9774 argument points to or @code{(size_t) -1} if the size is not known.
9776 The built-in functions are optimized into the normal string functions
9777 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9778 it is known at compile time that the destination object will not
9779 be overflown. If the compiler can determine at compile time the
9780 object will be always overflown, it issues a warning.
9782 The intended use can be e.g.@:
9786 #define bos0(dest) __builtin_object_size (dest, 0)
9787 #define memcpy(dest, src, n) \
9788 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9792 /* It is unknown what object p points to, so this is optimized
9793 into plain memcpy - no checking is possible. */
9794 memcpy (p, "abcde", n);
9795 /* Destination is known and length too. It is known at compile
9796 time there will be no overflow. */
9797 memcpy (&buf[5], "abcde", 5);
9798 /* Destination is known, but the length is not known at compile time.
9799 This will result in __memcpy_chk call that can check for overflow
9801 memcpy (&buf[5], "abcde", n);
9802 /* Destination is known and it is known at compile time there will
9803 be overflow. There will be a warning and __memcpy_chk call that
9804 will abort the program at run time. */
9805 memcpy (&buf[6], "abcde", 5);
9808 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9809 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9810 @code{strcat} and @code{strncat}.
9812 There are also checking built-in functions for formatted output functions.
9814 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9815 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9816 const char *fmt, ...);
9817 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9819 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9820 const char *fmt, va_list ap);
9823 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9824 etc.@: functions and can contain implementation specific flags on what
9825 additional security measures the checking function might take, such as
9826 handling @code{%n} differently.
9828 The @var{os} argument is the object size @var{s} points to, like in the
9829 other built-in functions. There is a small difference in the behavior
9830 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9831 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9832 the checking function is called with @var{os} argument set to
9835 In addition to this, there are checking built-in functions
9836 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9837 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9838 These have just one additional argument, @var{flag}, right before
9839 format string @var{fmt}. If the compiler is able to optimize them to
9840 @code{fputc} etc.@: functions, it does, otherwise the checking function
9841 is called and the @var{flag} argument passed to it.
9843 @node Pointer Bounds Checker builtins
9844 @section Pointer Bounds Checker Built-in Functions
9845 @cindex Pointer Bounds Checker builtins
9846 @findex __builtin___bnd_set_ptr_bounds
9847 @findex __builtin___bnd_narrow_ptr_bounds
9848 @findex __builtin___bnd_copy_ptr_bounds
9849 @findex __builtin___bnd_init_ptr_bounds
9850 @findex __builtin___bnd_null_ptr_bounds
9851 @findex __builtin___bnd_store_ptr_bounds
9852 @findex __builtin___bnd_chk_ptr_lbounds
9853 @findex __builtin___bnd_chk_ptr_ubounds
9854 @findex __builtin___bnd_chk_ptr_bounds
9855 @findex __builtin___bnd_get_ptr_lbound
9856 @findex __builtin___bnd_get_ptr_ubound
9858 GCC provides a set of built-in functions to control Pointer Bounds Checker
9859 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9860 even if you compile with Pointer Bounds Checker off
9861 (@option{-fno-check-pointer-bounds}).
9862 The behavior may differ in such case as documented below.
9864 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9866 This built-in function returns a new pointer with the value of @var{q}, and
9867 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9868 Bounds Checker off, the built-in function just returns the first argument.
9871 extern void *__wrap_malloc (size_t n)
9873 void *p = (void *)__real_malloc (n);
9874 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9875 return __builtin___bnd_set_ptr_bounds (p, n);
9881 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9883 This built-in function returns a new pointer with the value of @var{p}
9884 and associates it with the narrowed bounds formed by the intersection
9885 of bounds associated with @var{q} and the bounds
9886 [@var{p}, @var{p} + @var{size} - 1].
9887 With Pointer Bounds Checker off, the built-in function just returns the first
9891 void init_objects (object *objs, size_t size)
9894 /* Initialize objects one-by-one passing pointers with bounds of
9895 an object, not the full array of objects. */
9896 for (i = 0; i < size; i++)
9897 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
9904 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
9906 This built-in function returns a new pointer with the value of @var{q},
9907 and associates it with the bounds already associated with pointer @var{r}.
9908 With Pointer Bounds Checker off, the built-in function just returns the first
9912 /* Here is a way to get pointer to object's field but
9913 still with the full object's bounds. */
9914 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
9920 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
9922 This built-in function returns a new pointer with the value of @var{q}, and
9923 associates it with INIT (allowing full memory access) bounds. With Pointer
9924 Bounds Checker off, the built-in function just returns the first argument.
9928 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
9930 This built-in function returns a new pointer with the value of @var{q}, and
9931 associates it with NULL (allowing no memory access) bounds. With Pointer
9932 Bounds Checker off, the built-in function just returns the first argument.
9936 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
9938 This built-in function stores the bounds associated with pointer @var{ptr_val}
9939 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
9940 bounds from legacy code without touching the associated pointer's memory when
9941 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
9942 function call is ignored.
9946 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
9948 This built-in function checks if the pointer @var{q} is within the lower
9949 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
9950 function call is ignored.
9953 extern void *__wrap_memset (void *dst, int c, size_t len)
9957 __builtin___bnd_chk_ptr_lbounds (dst);
9958 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
9959 __real_memset (dst, c, len);
9967 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
9969 This built-in function checks if the pointer @var{q} is within the upper
9970 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
9971 function call is ignored.
9975 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
9977 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
9978 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
9979 off, the built-in function call is ignored.
9982 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
9986 __bnd_chk_ptr_bounds (dst, n);
9987 __bnd_chk_ptr_bounds (src, n);
9988 __real_memcpy (dst, src, n);
9996 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
9998 This built-in function returns the lower bound associated
9999 with the pointer @var{q}, as a pointer value.
10000 This is useful for debugging using @code{printf}.
10001 With Pointer Bounds Checker off, the built-in function returns 0.
10004 void *lb = __builtin___bnd_get_ptr_lbound (q);
10005 void *ub = __builtin___bnd_get_ptr_ubound (q);
10006 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10011 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10013 This built-in function returns the upper bound (which is a pointer) associated
10014 with the pointer @var{q}. With Pointer Bounds Checker off,
10015 the built-in function returns -1.
10019 @node Cilk Plus Builtins
10020 @section Cilk Plus C/C++ Language Extension Built-in Functions
10022 GCC provides support for the following built-in reduction functions if Cilk Plus
10023 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10026 @item @code{__sec_implicit_index}
10027 @item @code{__sec_reduce}
10028 @item @code{__sec_reduce_add}
10029 @item @code{__sec_reduce_all_nonzero}
10030 @item @code{__sec_reduce_all_zero}
10031 @item @code{__sec_reduce_any_nonzero}
10032 @item @code{__sec_reduce_any_zero}
10033 @item @code{__sec_reduce_max}
10034 @item @code{__sec_reduce_min}
10035 @item @code{__sec_reduce_max_ind}
10036 @item @code{__sec_reduce_min_ind}
10037 @item @code{__sec_reduce_mul}
10038 @item @code{__sec_reduce_mutating}
10041 Further details and examples about these built-in functions are described
10042 in the Cilk Plus language manual which can be found at
10043 @uref{http://www.cilkplus.org}.
10045 @node Other Builtins
10046 @section Other Built-in Functions Provided by GCC
10047 @cindex built-in functions
10048 @findex __builtin_call_with_static_chain
10049 @findex __builtin_fpclassify
10050 @findex __builtin_isfinite
10051 @findex __builtin_isnormal
10052 @findex __builtin_isgreater
10053 @findex __builtin_isgreaterequal
10054 @findex __builtin_isinf_sign
10055 @findex __builtin_isless
10056 @findex __builtin_islessequal
10057 @findex __builtin_islessgreater
10058 @findex __builtin_isunordered
10059 @findex __builtin_powi
10060 @findex __builtin_powif
10061 @findex __builtin_powil
10219 @findex fprintf_unlocked
10221 @findex fputs_unlocked
10329 @findex nexttowardf
10330 @findex nexttowardl
10338 @findex printf_unlocked
10368 @findex signbitd128
10369 @findex significand
10370 @findex significandf
10371 @findex significandl
10399 @findex strncasecmp
10442 GCC provides a large number of built-in functions other than the ones
10443 mentioned above. Some of these are for internal use in the processing
10444 of exceptions or variable-length argument lists and are not
10445 documented here because they may change from time to time; we do not
10446 recommend general use of these functions.
10448 The remaining functions are provided for optimization purposes.
10450 With the exception of built-ins that have library equivalents such as
10451 the standard C library functions discussed below, or that expand to
10452 library calls, GCC built-in functions are always expanded inline and
10453 thus do not have corresponding entry points and their address cannot
10454 be obtained. Attempting to use them in an expression other than
10455 a function call results in a compile-time error.
10457 @opindex fno-builtin
10458 GCC includes built-in versions of many of the functions in the standard
10459 C library. These functions come in two forms: one whose names start with
10460 the @code{__builtin_} prefix, and the other without. Both forms have the
10461 same type (including prototype), the same address (when their address is
10462 taken), and the same meaning as the C library functions even if you specify
10463 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10464 functions are only optimized in certain cases; if they are not optimized in
10465 a particular case, a call to the library function is emitted.
10469 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10470 @option{-std=c99} or @option{-std=c11}), the functions
10471 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10472 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10473 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10474 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10475 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10476 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10477 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10478 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10479 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10480 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10481 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10482 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10483 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10484 @code{significandl}, @code{significand}, @code{sincosf},
10485 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10486 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10487 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10488 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10490 may be handled as built-in functions.
10491 All these functions have corresponding versions
10492 prefixed with @code{__builtin_}, which may be used even in strict C90
10495 The ISO C99 functions
10496 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10497 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10498 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10499 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10500 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10501 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10502 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10503 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10504 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10505 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10506 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10507 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10508 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10509 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10510 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10511 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10512 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10513 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10514 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10515 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10516 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10517 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10518 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10519 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10520 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10521 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10522 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10523 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10524 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10525 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10526 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10527 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10528 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10529 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10530 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10531 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10532 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10533 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10534 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10535 are handled as built-in functions
10536 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10538 There are also built-in versions of the ISO C99 functions
10539 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10540 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10541 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10542 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10543 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10544 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10545 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10546 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10547 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10548 that are recognized in any mode since ISO C90 reserves these names for
10549 the purpose to which ISO C99 puts them. All these functions have
10550 corresponding versions prefixed with @code{__builtin_}.
10552 The ISO C94 functions
10553 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10554 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10555 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10557 are handled as built-in functions
10558 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10560 The ISO C90 functions
10561 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10562 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10563 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10564 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10565 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10566 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10567 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10568 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10569 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10570 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10571 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10572 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10573 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10574 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10575 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10576 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10577 are all recognized as built-in functions unless
10578 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10579 is specified for an individual function). All of these functions have
10580 corresponding versions prefixed with @code{__builtin_}.
10582 GCC provides built-in versions of the ISO C99 floating-point comparison
10583 macros that avoid raising exceptions for unordered operands. They have
10584 the same names as the standard macros ( @code{isgreater},
10585 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10586 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10587 prefixed. We intend for a library implementor to be able to simply
10588 @code{#define} each standard macro to its built-in equivalent.
10589 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10590 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10591 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10592 built-in functions appear both with and without the @code{__builtin_} prefix.
10594 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10596 You can use the built-in function @code{__builtin_types_compatible_p} to
10597 determine whether two types are the same.
10599 This built-in function returns 1 if the unqualified versions of the
10600 types @var{type1} and @var{type2} (which are types, not expressions) are
10601 compatible, 0 otherwise. The result of this built-in function can be
10602 used in integer constant expressions.
10604 This built-in function ignores top level qualifiers (e.g., @code{const},
10605 @code{volatile}). For example, @code{int} is equivalent to @code{const
10608 The type @code{int[]} and @code{int[5]} are compatible. On the other
10609 hand, @code{int} and @code{char *} are not compatible, even if the size
10610 of their types, on the particular architecture are the same. Also, the
10611 amount of pointer indirection is taken into account when determining
10612 similarity. Consequently, @code{short *} is not similar to
10613 @code{short **}. Furthermore, two types that are typedefed are
10614 considered compatible if their underlying types are compatible.
10616 An @code{enum} type is not considered to be compatible with another
10617 @code{enum} type even if both are compatible with the same integer
10618 type; this is what the C standard specifies.
10619 For example, @code{enum @{foo, bar@}} is not similar to
10620 @code{enum @{hot, dog@}}.
10622 You typically use this function in code whose execution varies
10623 depending on the arguments' types. For example:
10628 typeof (x) tmp = (x); \
10629 if (__builtin_types_compatible_p (typeof (x), long double)) \
10630 tmp = foo_long_double (tmp); \
10631 else if (__builtin_types_compatible_p (typeof (x), double)) \
10632 tmp = foo_double (tmp); \
10633 else if (__builtin_types_compatible_p (typeof (x), float)) \
10634 tmp = foo_float (tmp); \
10641 @emph{Note:} This construct is only available for C@.
10645 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10647 The @var{call_exp} expression must be a function call, and the
10648 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10649 is passed to the function call in the target's static chain location.
10650 The result of builtin is the result of the function call.
10652 @emph{Note:} This builtin is only available for C@.
10653 This builtin can be used to call Go closures from C.
10657 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10659 You can use the built-in function @code{__builtin_choose_expr} to
10660 evaluate code depending on the value of a constant expression. This
10661 built-in function returns @var{exp1} if @var{const_exp}, which is an
10662 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10664 This built-in function is analogous to the @samp{? :} operator in C,
10665 except that the expression returned has its type unaltered by promotion
10666 rules. Also, the built-in function does not evaluate the expression
10667 that is not chosen. For example, if @var{const_exp} evaluates to true,
10668 @var{exp2} is not evaluated even if it has side-effects.
10670 This built-in function can return an lvalue if the chosen argument is an
10673 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10674 type. Similarly, if @var{exp2} is returned, its return type is the same
10681 __builtin_choose_expr ( \
10682 __builtin_types_compatible_p (typeof (x), double), \
10684 __builtin_choose_expr ( \
10685 __builtin_types_compatible_p (typeof (x), float), \
10687 /* @r{The void expression results in a compile-time error} \
10688 @r{when assigning the result to something.} */ \
10692 @emph{Note:} This construct is only available for C@. Furthermore, the
10693 unused expression (@var{exp1} or @var{exp2} depending on the value of
10694 @var{const_exp}) may still generate syntax errors. This may change in
10699 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10701 The built-in function @code{__builtin_complex} is provided for use in
10702 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10703 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10704 real binary floating-point type, and the result has the corresponding
10705 complex type with real and imaginary parts @var{real} and @var{imag}.
10706 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10707 infinities, NaNs and negative zeros are involved.
10711 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10712 You can use the built-in function @code{__builtin_constant_p} to
10713 determine if a value is known to be constant at compile time and hence
10714 that GCC can perform constant-folding on expressions involving that
10715 value. The argument of the function is the value to test. The function
10716 returns the integer 1 if the argument is known to be a compile-time
10717 constant and 0 if it is not known to be a compile-time constant. A
10718 return of 0 does not indicate that the value is @emph{not} a constant,
10719 but merely that GCC cannot prove it is a constant with the specified
10720 value of the @option{-O} option.
10722 You typically use this function in an embedded application where
10723 memory is a critical resource. If you have some complex calculation,
10724 you may want it to be folded if it involves constants, but need to call
10725 a function if it does not. For example:
10728 #define Scale_Value(X) \
10729 (__builtin_constant_p (X) \
10730 ? ((X) * SCALE + OFFSET) : Scale (X))
10733 You may use this built-in function in either a macro or an inline
10734 function. However, if you use it in an inlined function and pass an
10735 argument of the function as the argument to the built-in, GCC
10736 never returns 1 when you call the inline function with a string constant
10737 or compound literal (@pxref{Compound Literals}) and does not return 1
10738 when you pass a constant numeric value to the inline function unless you
10739 specify the @option{-O} option.
10741 You may also use @code{__builtin_constant_p} in initializers for static
10742 data. For instance, you can write
10745 static const int table[] = @{
10746 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10752 This is an acceptable initializer even if @var{EXPRESSION} is not a
10753 constant expression, including the case where
10754 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10755 folded to a constant but @var{EXPRESSION} contains operands that are
10756 not otherwise permitted in a static initializer (for example,
10757 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10758 built-in in this case, because it has no opportunity to perform
10762 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10763 @opindex fprofile-arcs
10764 You may use @code{__builtin_expect} to provide the compiler with
10765 branch prediction information. In general, you should prefer to
10766 use actual profile feedback for this (@option{-fprofile-arcs}), as
10767 programmers are notoriously bad at predicting how their programs
10768 actually perform. However, there are applications in which this
10769 data is hard to collect.
10771 The return value is the value of @var{exp}, which should be an integral
10772 expression. The semantics of the built-in are that it is expected that
10773 @var{exp} == @var{c}. For example:
10776 if (__builtin_expect (x, 0))
10781 indicates that we do not expect to call @code{foo}, since
10782 we expect @code{x} to be zero. Since you are limited to integral
10783 expressions for @var{exp}, you should use constructions such as
10786 if (__builtin_expect (ptr != NULL, 1))
10791 when testing pointer or floating-point values.
10794 @deftypefn {Built-in Function} void __builtin_trap (void)
10795 This function causes the program to exit abnormally. GCC implements
10796 this function by using a target-dependent mechanism (such as
10797 intentionally executing an illegal instruction) or by calling
10798 @code{abort}. The mechanism used may vary from release to release so
10799 you should not rely on any particular implementation.
10802 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10803 If control flow reaches the point of the @code{__builtin_unreachable},
10804 the program is undefined. It is useful in situations where the
10805 compiler cannot deduce the unreachability of the code.
10807 One such case is immediately following an @code{asm} statement that
10808 either never terminates, or one that transfers control elsewhere
10809 and never returns. In this example, without the
10810 @code{__builtin_unreachable}, GCC issues a warning that control
10811 reaches the end of a non-void function. It also generates code
10812 to return after the @code{asm}.
10815 int f (int c, int v)
10823 asm("jmp error_handler");
10824 __builtin_unreachable ();
10830 Because the @code{asm} statement unconditionally transfers control out
10831 of the function, control never reaches the end of the function
10832 body. The @code{__builtin_unreachable} is in fact unreachable and
10833 communicates this fact to the compiler.
10835 Another use for @code{__builtin_unreachable} is following a call a
10836 function that never returns but that is not declared
10837 @code{__attribute__((noreturn))}, as in this example:
10840 void function_that_never_returns (void);
10850 function_that_never_returns ();
10851 __builtin_unreachable ();
10858 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
10859 This function returns its first argument, and allows the compiler
10860 to assume that the returned pointer is at least @var{align} bytes
10861 aligned. This built-in can have either two or three arguments,
10862 if it has three, the third argument should have integer type, and
10863 if it is nonzero means misalignment offset. For example:
10866 void *x = __builtin_assume_aligned (arg, 16);
10870 means that the compiler can assume @code{x}, set to @code{arg}, is at least
10871 16-byte aligned, while:
10874 void *x = __builtin_assume_aligned (arg, 32, 8);
10878 means that the compiler can assume for @code{x}, set to @code{arg}, that
10879 @code{(char *) x - 8} is 32-byte aligned.
10882 @deftypefn {Built-in Function} int __builtin_LINE ()
10883 This function is the equivalent to the preprocessor @code{__LINE__}
10884 macro and returns the line number of the invocation of the built-in.
10885 In a C++ default argument for a function @var{F}, it gets the line number of
10886 the call to @var{F}.
10889 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
10890 This function is the equivalent to the preprocessor @code{__FUNCTION__}
10891 macro and returns the function name the invocation of the built-in is in.
10894 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
10895 This function is the equivalent to the preprocessor @code{__FILE__}
10896 macro and returns the file name the invocation of the built-in is in.
10897 In a C++ default argument for a function @var{F}, it gets the file name of
10898 the call to @var{F}.
10901 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
10902 This function is used to flush the processor's instruction cache for
10903 the region of memory between @var{begin} inclusive and @var{end}
10904 exclusive. Some targets require that the instruction cache be
10905 flushed, after modifying memory containing code, in order to obtain
10906 deterministic behavior.
10908 If the target does not require instruction cache flushes,
10909 @code{__builtin___clear_cache} has no effect. Otherwise either
10910 instructions are emitted in-line to clear the instruction cache or a
10911 call to the @code{__clear_cache} function in libgcc is made.
10914 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
10915 This function is used to minimize cache-miss latency by moving data into
10916 a cache before it is accessed.
10917 You can insert calls to @code{__builtin_prefetch} into code for which
10918 you know addresses of data in memory that is likely to be accessed soon.
10919 If the target supports them, data prefetch instructions are generated.
10920 If the prefetch is done early enough before the access then the data will
10921 be in the cache by the time it is accessed.
10923 The value of @var{addr} is the address of the memory to prefetch.
10924 There are two optional arguments, @var{rw} and @var{locality}.
10925 The value of @var{rw} is a compile-time constant one or zero; one
10926 means that the prefetch is preparing for a write to the memory address
10927 and zero, the default, means that the prefetch is preparing for a read.
10928 The value @var{locality} must be a compile-time constant integer between
10929 zero and three. A value of zero means that the data has no temporal
10930 locality, so it need not be left in the cache after the access. A value
10931 of three means that the data has a high degree of temporal locality and
10932 should be left in all levels of cache possible. Values of one and two
10933 mean, respectively, a low or moderate degree of temporal locality. The
10937 for (i = 0; i < n; i++)
10939 a[i] = a[i] + b[i];
10940 __builtin_prefetch (&a[i+j], 1, 1);
10941 __builtin_prefetch (&b[i+j], 0, 1);
10946 Data prefetch does not generate faults if @var{addr} is invalid, but
10947 the address expression itself must be valid. For example, a prefetch
10948 of @code{p->next} does not fault if @code{p->next} is not a valid
10949 address, but evaluation faults if @code{p} is not a valid address.
10951 If the target does not support data prefetch, the address expression
10952 is evaluated if it includes side effects but no other code is generated
10953 and GCC does not issue a warning.
10956 @deftypefn {Built-in Function} double __builtin_huge_val (void)
10957 Returns a positive infinity, if supported by the floating-point format,
10958 else @code{DBL_MAX}. This function is suitable for implementing the
10959 ISO C macro @code{HUGE_VAL}.
10962 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
10963 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
10966 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
10967 Similar to @code{__builtin_huge_val}, except the return
10968 type is @code{long double}.
10971 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
10972 This built-in implements the C99 fpclassify functionality. The first
10973 five int arguments should be the target library's notion of the
10974 possible FP classes and are used for return values. They must be
10975 constant values and they must appear in this order: @code{FP_NAN},
10976 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
10977 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
10978 to classify. GCC treats the last argument as type-generic, which
10979 means it does not do default promotion from float to double.
10982 @deftypefn {Built-in Function} double __builtin_inf (void)
10983 Similar to @code{__builtin_huge_val}, except a warning is generated
10984 if the target floating-point format does not support infinities.
10987 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
10988 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
10991 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
10992 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
10995 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
10996 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
10999 @deftypefn {Built-in Function} float __builtin_inff (void)
11000 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11001 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11004 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11005 Similar to @code{__builtin_inf}, except the return
11006 type is @code{long double}.
11009 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11010 Similar to @code{isinf}, except the return value is -1 for
11011 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11012 Note while the parameter list is an
11013 ellipsis, this function only accepts exactly one floating-point
11014 argument. GCC treats this parameter as type-generic, which means it
11015 does not do default promotion from float to double.
11018 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11019 This is an implementation of the ISO C99 function @code{nan}.
11021 Since ISO C99 defines this function in terms of @code{strtod}, which we
11022 do not implement, a description of the parsing is in order. The string
11023 is parsed as by @code{strtol}; that is, the base is recognized by
11024 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11025 in the significand such that the least significant bit of the number
11026 is at the least significant bit of the significand. The number is
11027 truncated to fit the significand field provided. The significand is
11028 forced to be a quiet NaN@.
11030 This function, if given a string literal all of which would have been
11031 consumed by @code{strtol}, is evaluated early enough that it is considered a
11032 compile-time constant.
11035 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11036 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11039 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11040 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11043 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11044 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11047 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11048 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11051 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11052 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11055 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11056 Similar to @code{__builtin_nan}, except the significand is forced
11057 to be a signaling NaN@. The @code{nans} function is proposed by
11058 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11061 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11062 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11065 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11066 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11069 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11070 Returns one plus the index of the least significant 1-bit of @var{x}, or
11071 if @var{x} is zero, returns zero.
11074 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11075 Returns the number of leading 0-bits in @var{x}, starting at the most
11076 significant bit position. If @var{x} is 0, the result is undefined.
11079 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11080 Returns the number of trailing 0-bits in @var{x}, starting at the least
11081 significant bit position. If @var{x} is 0, the result is undefined.
11084 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11085 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11086 number of bits following the most significant bit that are identical
11087 to it. There are no special cases for 0 or other values.
11090 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11091 Returns the number of 1-bits in @var{x}.
11094 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11095 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11099 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11100 Similar to @code{__builtin_ffs}, except the argument type is
11104 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11105 Similar to @code{__builtin_clz}, except the argument type is
11106 @code{unsigned long}.
11109 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11110 Similar to @code{__builtin_ctz}, except the argument type is
11111 @code{unsigned long}.
11114 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11115 Similar to @code{__builtin_clrsb}, except the argument type is
11119 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11120 Similar to @code{__builtin_popcount}, except the argument type is
11121 @code{unsigned long}.
11124 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11125 Similar to @code{__builtin_parity}, except the argument type is
11126 @code{unsigned long}.
11129 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11130 Similar to @code{__builtin_ffs}, except the argument type is
11134 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11135 Similar to @code{__builtin_clz}, except the argument type is
11136 @code{unsigned long long}.
11139 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11140 Similar to @code{__builtin_ctz}, except the argument type is
11141 @code{unsigned long long}.
11144 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11145 Similar to @code{__builtin_clrsb}, except the argument type is
11149 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11150 Similar to @code{__builtin_popcount}, except the argument type is
11151 @code{unsigned long long}.
11154 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11155 Similar to @code{__builtin_parity}, except the argument type is
11156 @code{unsigned long long}.
11159 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11160 Returns the first argument raised to the power of the second. Unlike the
11161 @code{pow} function no guarantees about precision and rounding are made.
11164 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11165 Similar to @code{__builtin_powi}, except the argument and return types
11169 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11170 Similar to @code{__builtin_powi}, except the argument and return types
11171 are @code{long double}.
11174 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11175 Returns @var{x} with the order of the bytes reversed; for example,
11176 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11180 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11181 Similar to @code{__builtin_bswap16}, except the argument and return types
11185 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11186 Similar to @code{__builtin_bswap32}, except the argument and return types
11190 @node Target Builtins
11191 @section Built-in Functions Specific to Particular Target Machines
11193 On some target machines, GCC supports many built-in functions specific
11194 to those machines. Generally these generate calls to specific machine
11195 instructions, but allow the compiler to schedule those calls.
11198 * AArch64 Built-in Functions::
11199 * Alpha Built-in Functions::
11200 * Altera Nios II Built-in Functions::
11201 * ARC Built-in Functions::
11202 * ARC SIMD Built-in Functions::
11203 * ARM iWMMXt Built-in Functions::
11204 * ARM C Language Extensions (ACLE)::
11205 * ARM Floating Point Status and Control Intrinsics::
11206 * AVR Built-in Functions::
11207 * Blackfin Built-in Functions::
11208 * FR-V Built-in Functions::
11209 * MIPS DSP Built-in Functions::
11210 * MIPS Paired-Single Support::
11211 * MIPS Loongson Built-in Functions::
11212 * Other MIPS Built-in Functions::
11213 * MSP430 Built-in Functions::
11214 * NDS32 Built-in Functions::
11215 * picoChip Built-in Functions::
11216 * PowerPC Built-in Functions::
11217 * PowerPC AltiVec/VSX Built-in Functions::
11218 * PowerPC Hardware Transactional Memory Built-in Functions::
11219 * RX Built-in Functions::
11220 * S/390 System z Built-in Functions::
11221 * SH Built-in Functions::
11222 * SPARC VIS Built-in Functions::
11223 * SPU Built-in Functions::
11224 * TI C6X Built-in Functions::
11225 * TILE-Gx Built-in Functions::
11226 * TILEPro Built-in Functions::
11227 * x86 Built-in Functions::
11228 * x86 transactional memory intrinsics::
11231 @node AArch64 Built-in Functions
11232 @subsection AArch64 Built-in Functions
11234 These built-in functions are available for the AArch64 family of
11237 unsigned int __builtin_aarch64_get_fpcr ()
11238 void __builtin_aarch64_set_fpcr (unsigned int)
11239 unsigned int __builtin_aarch64_get_fpsr ()
11240 void __builtin_aarch64_set_fpsr (unsigned int)
11243 @node Alpha Built-in Functions
11244 @subsection Alpha Built-in Functions
11246 These built-in functions are available for the Alpha family of
11247 processors, depending on the command-line switches used.
11249 The following built-in functions are always available. They
11250 all generate the machine instruction that is part of the name.
11253 long __builtin_alpha_implver (void)
11254 long __builtin_alpha_rpcc (void)
11255 long __builtin_alpha_amask (long)
11256 long __builtin_alpha_cmpbge (long, long)
11257 long __builtin_alpha_extbl (long, long)
11258 long __builtin_alpha_extwl (long, long)
11259 long __builtin_alpha_extll (long, long)
11260 long __builtin_alpha_extql (long, long)
11261 long __builtin_alpha_extwh (long, long)
11262 long __builtin_alpha_extlh (long, long)
11263 long __builtin_alpha_extqh (long, long)
11264 long __builtin_alpha_insbl (long, long)
11265 long __builtin_alpha_inswl (long, long)
11266 long __builtin_alpha_insll (long, long)
11267 long __builtin_alpha_insql (long, long)
11268 long __builtin_alpha_inswh (long, long)
11269 long __builtin_alpha_inslh (long, long)
11270 long __builtin_alpha_insqh (long, long)
11271 long __builtin_alpha_mskbl (long, long)
11272 long __builtin_alpha_mskwl (long, long)
11273 long __builtin_alpha_mskll (long, long)
11274 long __builtin_alpha_mskql (long, long)
11275 long __builtin_alpha_mskwh (long, long)
11276 long __builtin_alpha_msklh (long, long)
11277 long __builtin_alpha_mskqh (long, long)
11278 long __builtin_alpha_umulh (long, long)
11279 long __builtin_alpha_zap (long, long)
11280 long __builtin_alpha_zapnot (long, long)
11283 The following built-in functions are always with @option{-mmax}
11284 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11285 later. They all generate the machine instruction that is part
11289 long __builtin_alpha_pklb (long)
11290 long __builtin_alpha_pkwb (long)
11291 long __builtin_alpha_unpkbl (long)
11292 long __builtin_alpha_unpkbw (long)
11293 long __builtin_alpha_minub8 (long, long)
11294 long __builtin_alpha_minsb8 (long, long)
11295 long __builtin_alpha_minuw4 (long, long)
11296 long __builtin_alpha_minsw4 (long, long)
11297 long __builtin_alpha_maxub8 (long, long)
11298 long __builtin_alpha_maxsb8 (long, long)
11299 long __builtin_alpha_maxuw4 (long, long)
11300 long __builtin_alpha_maxsw4 (long, long)
11301 long __builtin_alpha_perr (long, long)
11304 The following built-in functions are always with @option{-mcix}
11305 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11306 later. They all generate the machine instruction that is part
11310 long __builtin_alpha_cttz (long)
11311 long __builtin_alpha_ctlz (long)
11312 long __builtin_alpha_ctpop (long)
11315 The following built-in functions are available on systems that use the OSF/1
11316 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11317 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11318 @code{rdval} and @code{wrval}.
11321 void *__builtin_thread_pointer (void)
11322 void __builtin_set_thread_pointer (void *)
11325 @node Altera Nios II Built-in Functions
11326 @subsection Altera Nios II Built-in Functions
11328 These built-in functions are available for the Altera Nios II
11329 family of processors.
11331 The following built-in functions are always available. They
11332 all generate the machine instruction that is part of the name.
11335 int __builtin_ldbio (volatile const void *)
11336 int __builtin_ldbuio (volatile const void *)
11337 int __builtin_ldhio (volatile const void *)
11338 int __builtin_ldhuio (volatile const void *)
11339 int __builtin_ldwio (volatile const void *)
11340 void __builtin_stbio (volatile void *, int)
11341 void __builtin_sthio (volatile void *, int)
11342 void __builtin_stwio (volatile void *, int)
11343 void __builtin_sync (void)
11344 int __builtin_rdctl (int)
11345 int __builtin_rdprs (int, int)
11346 void __builtin_wrctl (int, int)
11347 void __builtin_flushd (volatile void *)
11348 void __builtin_flushda (volatile void *)
11349 int __builtin_wrpie (int);
11350 void __builtin_eni (int);
11351 int __builtin_ldex (volatile const void *)
11352 int __builtin_stex (volatile void *, int)
11353 int __builtin_ldsex (volatile const void *)
11354 int __builtin_stsex (volatile void *, int)
11357 The following built-in functions are always available. They
11358 all generate a Nios II Custom Instruction. The name of the
11359 function represents the types that the function takes and
11360 returns. The letter before the @code{n} is the return type
11361 or void if absent. The @code{n} represents the first parameter
11362 to all the custom instructions, the custom instruction number.
11363 The two letters after the @code{n} represent the up to two
11364 parameters to the function.
11366 The letters represent the following data types:
11369 @code{void} for return type and no parameter for parameter types.
11372 @code{int} for return type and parameter type
11375 @code{float} for return type and parameter type
11378 @code{void *} for return type and parameter type
11382 And the function names are:
11384 void __builtin_custom_n (void)
11385 void __builtin_custom_ni (int)
11386 void __builtin_custom_nf (float)
11387 void __builtin_custom_np (void *)
11388 void __builtin_custom_nii (int, int)
11389 void __builtin_custom_nif (int, float)
11390 void __builtin_custom_nip (int, void *)
11391 void __builtin_custom_nfi (float, int)
11392 void __builtin_custom_nff (float, float)
11393 void __builtin_custom_nfp (float, void *)
11394 void __builtin_custom_npi (void *, int)
11395 void __builtin_custom_npf (void *, float)
11396 void __builtin_custom_npp (void *, void *)
11397 int __builtin_custom_in (void)
11398 int __builtin_custom_ini (int)
11399 int __builtin_custom_inf (float)
11400 int __builtin_custom_inp (void *)
11401 int __builtin_custom_inii (int, int)
11402 int __builtin_custom_inif (int, float)
11403 int __builtin_custom_inip (int, void *)
11404 int __builtin_custom_infi (float, int)
11405 int __builtin_custom_inff (float, float)
11406 int __builtin_custom_infp (float, void *)
11407 int __builtin_custom_inpi (void *, int)
11408 int __builtin_custom_inpf (void *, float)
11409 int __builtin_custom_inpp (void *, void *)
11410 float __builtin_custom_fn (void)
11411 float __builtin_custom_fni (int)
11412 float __builtin_custom_fnf (float)
11413 float __builtin_custom_fnp (void *)
11414 float __builtin_custom_fnii (int, int)
11415 float __builtin_custom_fnif (int, float)
11416 float __builtin_custom_fnip (int, void *)
11417 float __builtin_custom_fnfi (float, int)
11418 float __builtin_custom_fnff (float, float)
11419 float __builtin_custom_fnfp (float, void *)
11420 float __builtin_custom_fnpi (void *, int)
11421 float __builtin_custom_fnpf (void *, float)
11422 float __builtin_custom_fnpp (void *, void *)
11423 void * __builtin_custom_pn (void)
11424 void * __builtin_custom_pni (int)
11425 void * __builtin_custom_pnf (float)
11426 void * __builtin_custom_pnp (void *)
11427 void * __builtin_custom_pnii (int, int)
11428 void * __builtin_custom_pnif (int, float)
11429 void * __builtin_custom_pnip (int, void *)
11430 void * __builtin_custom_pnfi (float, int)
11431 void * __builtin_custom_pnff (float, float)
11432 void * __builtin_custom_pnfp (float, void *)
11433 void * __builtin_custom_pnpi (void *, int)
11434 void * __builtin_custom_pnpf (void *, float)
11435 void * __builtin_custom_pnpp (void *, void *)
11438 @node ARC Built-in Functions
11439 @subsection ARC Built-in Functions
11441 The following built-in functions are provided for ARC targets. The
11442 built-ins generate the corresponding assembly instructions. In the
11443 examples given below, the generated code often requires an operand or
11444 result to be in a register. Where necessary further code will be
11445 generated to ensure this is true, but for brevity this is not
11446 described in each case.
11448 @emph{Note:} Using a built-in to generate an instruction not supported
11449 by a target may cause problems. At present the compiler is not
11450 guaranteed to detect such misuse, and as a result an internal compiler
11451 error may be generated.
11453 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11454 Return 1 if @var{val} is known to have the byte alignment given
11455 by @var{alignval}, otherwise return 0.
11456 Note that this is different from
11458 __alignof__(*(char *)@var{val}) >= alignval
11460 because __alignof__ sees only the type of the dereference, whereas
11461 __builtin_arc_align uses alignment information from the pointer
11462 as well as from the pointed-to type.
11463 The information available will depend on optimization level.
11466 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11473 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11474 The operand is the number of a register to be read. Generates:
11476 mov @var{dest}, r@var{regno}
11478 where the value in @var{dest} will be the result returned from the
11482 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11483 The first operand is the number of a register to be written, the
11484 second operand is a compile time constant to write into that
11485 register. Generates:
11487 mov r@var{regno}, @var{val}
11491 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11492 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11495 divaw @var{dest}, @var{a}, @var{b}
11497 where the value in @var{dest} will be the result returned from the
11501 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11508 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11509 The operand, @var{auxv}, is the address of an auxiliary register and
11510 must be a compile time constant. Generates:
11512 lr @var{dest}, [@var{auxr}]
11514 Where the value in @var{dest} will be the result returned from the
11518 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11519 Only available with @option{-mmul64}. Generates:
11521 mul64 @var{a}, @var{b}
11525 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11526 Only available with @option{-mmul64}. Generates:
11528 mulu64 @var{a}, @var{b}
11532 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11539 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11540 Only valid if the @samp{norm} instruction is available through the
11541 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11544 norm @var{dest}, @var{src}
11546 Where the value in @var{dest} will be the result returned from the
11550 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11551 Only valid if the @samp{normw} instruction is available through the
11552 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11555 normw @var{dest}, @var{src}
11557 Where the value in @var{dest} will be the result returned from the
11561 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11568 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11575 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11576 The first argument, @var{auxv}, is the address of an auxiliary
11577 register, the second argument, @var{val}, is a compile time constant
11578 to be written to the register. Generates:
11580 sr @var{auxr}, [@var{val}]
11584 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11585 Only valid with @option{-mswap}. Generates:
11587 swap @var{dest}, @var{src}
11589 Where the value in @var{dest} will be the result returned from the
11593 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11600 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11601 Only available with @option{-mcpu=ARC700}. Generates:
11607 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11608 Only available with @option{-mcpu=ARC700}. Generates:
11614 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11615 Only available with @option{-mcpu=ARC700}. Generates:
11621 The instructions generated by the following builtins are not
11622 considered as candidates for scheduling. They are not moved around by
11623 the compiler during scheduling, and thus can be expected to appear
11624 where they are put in the C code:
11626 __builtin_arc_brk()
11627 __builtin_arc_core_read()
11628 __builtin_arc_core_write()
11629 __builtin_arc_flag()
11631 __builtin_arc_sleep()
11633 __builtin_arc_swi()
11636 @node ARC SIMD Built-in Functions
11637 @subsection ARC SIMD Built-in Functions
11639 SIMD builtins provided by the compiler can be used to generate the
11640 vector instructions. This section describes the available builtins
11641 and their usage in programs. With the @option{-msimd} option, the
11642 compiler provides 128-bit vector types, which can be specified using
11643 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11644 can be included to use the following predefined types:
11646 typedef int __v4si __attribute__((vector_size(16)));
11647 typedef short __v8hi __attribute__((vector_size(16)));
11650 These types can be used to define 128-bit variables. The built-in
11651 functions listed in the following section can be used on these
11652 variables to generate the vector operations.
11654 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11655 @file{arc-simd.h} also provides equivalent macros called
11656 @code{_@var{someinsn}} that can be used for programming ease and
11657 improved readability. The following macros for DMA control are also
11660 #define _setup_dma_in_channel_reg _vdiwr
11661 #define _setup_dma_out_channel_reg _vdowr
11664 The following is a complete list of all the SIMD built-ins provided
11665 for ARC, grouped by calling signature.
11667 The following take two @code{__v8hi} arguments and return a
11668 @code{__v8hi} result:
11670 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11671 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11672 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11673 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11674 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11675 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11676 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11677 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11678 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11679 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11680 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11681 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11682 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11683 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11684 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11685 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11686 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11687 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11688 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11689 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11690 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11691 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11692 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11693 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11694 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11695 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11696 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11697 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11698 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11699 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11700 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11701 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11702 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11703 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11704 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11705 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11706 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11707 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11708 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11709 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11710 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11711 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11712 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11713 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11714 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11715 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11716 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11717 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11720 The following take one @code{__v8hi} and one @code{int} argument and return a
11721 @code{__v8hi} result:
11724 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11725 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11726 __v8hi __builtin_arc_vbminw (__v8hi, int)
11727 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11728 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11729 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11730 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11731 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11734 The following take one @code{__v8hi} argument and one @code{int} argument which
11735 must be a 3-bit compile time constant indicating a register number
11736 I0-I7. They return a @code{__v8hi} result.
11738 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11739 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11740 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11743 The following take one @code{__v8hi} argument and one @code{int}
11744 argument which must be a 6-bit compile time constant. They return a
11745 @code{__v8hi} result.
11747 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11748 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11749 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11750 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11751 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11752 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11753 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11756 The following take one @code{__v8hi} argument and one @code{int} argument which
11757 must be a 8-bit compile time constant. They return a @code{__v8hi}
11760 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11761 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11762 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11763 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11766 The following take two @code{int} arguments, the second of which which
11767 must be a 8-bit compile time constant. They return a @code{__v8hi}
11770 __v8hi __builtin_arc_vmovaw (int, const int)
11771 __v8hi __builtin_arc_vmovw (int, const int)
11772 __v8hi __builtin_arc_vmovzw (int, const int)
11775 The following take a single @code{__v8hi} argument and return a
11776 @code{__v8hi} result:
11778 __v8hi __builtin_arc_vabsaw (__v8hi)
11779 __v8hi __builtin_arc_vabsw (__v8hi)
11780 __v8hi __builtin_arc_vaddsuw (__v8hi)
11781 __v8hi __builtin_arc_vexch1 (__v8hi)
11782 __v8hi __builtin_arc_vexch2 (__v8hi)
11783 __v8hi __builtin_arc_vexch4 (__v8hi)
11784 __v8hi __builtin_arc_vsignw (__v8hi)
11785 __v8hi __builtin_arc_vupbaw (__v8hi)
11786 __v8hi __builtin_arc_vupbw (__v8hi)
11787 __v8hi __builtin_arc_vupsbaw (__v8hi)
11788 __v8hi __builtin_arc_vupsbw (__v8hi)
11791 The following take two @code{int} arguments and return no result:
11793 void __builtin_arc_vdirun (int, int)
11794 void __builtin_arc_vdorun (int, int)
11797 The following take two @code{int} arguments and return no result. The
11798 first argument must a 3-bit compile time constant indicating one of
11799 the DR0-DR7 DMA setup channels:
11801 void __builtin_arc_vdiwr (const int, int)
11802 void __builtin_arc_vdowr (const int, int)
11805 The following take an @code{int} argument and return no result:
11807 void __builtin_arc_vendrec (int)
11808 void __builtin_arc_vrec (int)
11809 void __builtin_arc_vrecrun (int)
11810 void __builtin_arc_vrun (int)
11813 The following take a @code{__v8hi} argument and two @code{int}
11814 arguments and return a @code{__v8hi} result. The second argument must
11815 be a 3-bit compile time constants, indicating one the registers I0-I7,
11816 and the third argument must be an 8-bit compile time constant.
11818 @emph{Note:} Although the equivalent hardware instructions do not take
11819 an SIMD register as an operand, these builtins overwrite the relevant
11820 bits of the @code{__v8hi} register provided as the first argument with
11821 the value loaded from the @code{[Ib, u8]} location in the SDM.
11824 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
11825 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
11826 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
11827 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
11830 The following take two @code{int} arguments and return a @code{__v8hi}
11831 result. The first argument must be a 3-bit compile time constants,
11832 indicating one the registers I0-I7, and the second argument must be an
11833 8-bit compile time constant.
11836 __v8hi __builtin_arc_vld128 (const int, const int)
11837 __v8hi __builtin_arc_vld64w (const int, const int)
11840 The following take a @code{__v8hi} argument and two @code{int}
11841 arguments and return no result. The second argument must be a 3-bit
11842 compile time constants, indicating one the registers I0-I7, and the
11843 third argument must be an 8-bit compile time constant.
11846 void __builtin_arc_vst128 (__v8hi, const int, const int)
11847 void __builtin_arc_vst64 (__v8hi, const int, const int)
11850 The following take a @code{__v8hi} argument and three @code{int}
11851 arguments and return no result. The second argument must be a 3-bit
11852 compile-time constant, identifying the 16-bit sub-register to be
11853 stored, the third argument must be a 3-bit compile time constants,
11854 indicating one the registers I0-I7, and the fourth argument must be an
11855 8-bit compile time constant.
11858 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
11859 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
11862 @node ARM iWMMXt Built-in Functions
11863 @subsection ARM iWMMXt Built-in Functions
11865 These built-in functions are available for the ARM family of
11866 processors when the @option{-mcpu=iwmmxt} switch is used:
11869 typedef int v2si __attribute__ ((vector_size (8)));
11870 typedef short v4hi __attribute__ ((vector_size (8)));
11871 typedef char v8qi __attribute__ ((vector_size (8)));
11873 int __builtin_arm_getwcgr0 (void)
11874 void __builtin_arm_setwcgr0 (int)
11875 int __builtin_arm_getwcgr1 (void)
11876 void __builtin_arm_setwcgr1 (int)
11877 int __builtin_arm_getwcgr2 (void)
11878 void __builtin_arm_setwcgr2 (int)
11879 int __builtin_arm_getwcgr3 (void)
11880 void __builtin_arm_setwcgr3 (int)
11881 int __builtin_arm_textrmsb (v8qi, int)
11882 int __builtin_arm_textrmsh (v4hi, int)
11883 int __builtin_arm_textrmsw (v2si, int)
11884 int __builtin_arm_textrmub (v8qi, int)
11885 int __builtin_arm_textrmuh (v4hi, int)
11886 int __builtin_arm_textrmuw (v2si, int)
11887 v8qi __builtin_arm_tinsrb (v8qi, int, int)
11888 v4hi __builtin_arm_tinsrh (v4hi, int, int)
11889 v2si __builtin_arm_tinsrw (v2si, int, int)
11890 long long __builtin_arm_tmia (long long, int, int)
11891 long long __builtin_arm_tmiabb (long long, int, int)
11892 long long __builtin_arm_tmiabt (long long, int, int)
11893 long long __builtin_arm_tmiaph (long long, int, int)
11894 long long __builtin_arm_tmiatb (long long, int, int)
11895 long long __builtin_arm_tmiatt (long long, int, int)
11896 int __builtin_arm_tmovmskb (v8qi)
11897 int __builtin_arm_tmovmskh (v4hi)
11898 int __builtin_arm_tmovmskw (v2si)
11899 long long __builtin_arm_waccb (v8qi)
11900 long long __builtin_arm_wacch (v4hi)
11901 long long __builtin_arm_waccw (v2si)
11902 v8qi __builtin_arm_waddb (v8qi, v8qi)
11903 v8qi __builtin_arm_waddbss (v8qi, v8qi)
11904 v8qi __builtin_arm_waddbus (v8qi, v8qi)
11905 v4hi __builtin_arm_waddh (v4hi, v4hi)
11906 v4hi __builtin_arm_waddhss (v4hi, v4hi)
11907 v4hi __builtin_arm_waddhus (v4hi, v4hi)
11908 v2si __builtin_arm_waddw (v2si, v2si)
11909 v2si __builtin_arm_waddwss (v2si, v2si)
11910 v2si __builtin_arm_waddwus (v2si, v2si)
11911 v8qi __builtin_arm_walign (v8qi, v8qi, int)
11912 long long __builtin_arm_wand(long long, long long)
11913 long long __builtin_arm_wandn (long long, long long)
11914 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
11915 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
11916 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
11917 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
11918 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
11919 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
11920 v2si __builtin_arm_wcmpeqw (v2si, v2si)
11921 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
11922 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
11923 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
11924 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
11925 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
11926 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
11927 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
11928 long long __builtin_arm_wmacsz (v4hi, v4hi)
11929 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
11930 long long __builtin_arm_wmacuz (v4hi, v4hi)
11931 v4hi __builtin_arm_wmadds (v4hi, v4hi)
11932 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
11933 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
11934 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
11935 v2si __builtin_arm_wmaxsw (v2si, v2si)
11936 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
11937 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
11938 v2si __builtin_arm_wmaxuw (v2si, v2si)
11939 v8qi __builtin_arm_wminsb (v8qi, v8qi)
11940 v4hi __builtin_arm_wminsh (v4hi, v4hi)
11941 v2si __builtin_arm_wminsw (v2si, v2si)
11942 v8qi __builtin_arm_wminub (v8qi, v8qi)
11943 v4hi __builtin_arm_wminuh (v4hi, v4hi)
11944 v2si __builtin_arm_wminuw (v2si, v2si)
11945 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
11946 v4hi __builtin_arm_wmulul (v4hi, v4hi)
11947 v4hi __builtin_arm_wmulum (v4hi, v4hi)
11948 long long __builtin_arm_wor (long long, long long)
11949 v2si __builtin_arm_wpackdss (long long, long long)
11950 v2si __builtin_arm_wpackdus (long long, long long)
11951 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
11952 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
11953 v4hi __builtin_arm_wpackwss (v2si, v2si)
11954 v4hi __builtin_arm_wpackwus (v2si, v2si)
11955 long long __builtin_arm_wrord (long long, long long)
11956 long long __builtin_arm_wrordi (long long, int)
11957 v4hi __builtin_arm_wrorh (v4hi, long long)
11958 v4hi __builtin_arm_wrorhi (v4hi, int)
11959 v2si __builtin_arm_wrorw (v2si, long long)
11960 v2si __builtin_arm_wrorwi (v2si, int)
11961 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
11962 v2si __builtin_arm_wsadbz (v8qi, v8qi)
11963 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
11964 v2si __builtin_arm_wsadhz (v4hi, v4hi)
11965 v4hi __builtin_arm_wshufh (v4hi, int)
11966 long long __builtin_arm_wslld (long long, long long)
11967 long long __builtin_arm_wslldi (long long, int)
11968 v4hi __builtin_arm_wsllh (v4hi, long long)
11969 v4hi __builtin_arm_wsllhi (v4hi, int)
11970 v2si __builtin_arm_wsllw (v2si, long long)
11971 v2si __builtin_arm_wsllwi (v2si, int)
11972 long long __builtin_arm_wsrad (long long, long long)
11973 long long __builtin_arm_wsradi (long long, int)
11974 v4hi __builtin_arm_wsrah (v4hi, long long)
11975 v4hi __builtin_arm_wsrahi (v4hi, int)
11976 v2si __builtin_arm_wsraw (v2si, long long)
11977 v2si __builtin_arm_wsrawi (v2si, int)
11978 long long __builtin_arm_wsrld (long long, long long)
11979 long long __builtin_arm_wsrldi (long long, int)
11980 v4hi __builtin_arm_wsrlh (v4hi, long long)
11981 v4hi __builtin_arm_wsrlhi (v4hi, int)
11982 v2si __builtin_arm_wsrlw (v2si, long long)
11983 v2si __builtin_arm_wsrlwi (v2si, int)
11984 v8qi __builtin_arm_wsubb (v8qi, v8qi)
11985 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
11986 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
11987 v4hi __builtin_arm_wsubh (v4hi, v4hi)
11988 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
11989 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
11990 v2si __builtin_arm_wsubw (v2si, v2si)
11991 v2si __builtin_arm_wsubwss (v2si, v2si)
11992 v2si __builtin_arm_wsubwus (v2si, v2si)
11993 v4hi __builtin_arm_wunpckehsb (v8qi)
11994 v2si __builtin_arm_wunpckehsh (v4hi)
11995 long long __builtin_arm_wunpckehsw (v2si)
11996 v4hi __builtin_arm_wunpckehub (v8qi)
11997 v2si __builtin_arm_wunpckehuh (v4hi)
11998 long long __builtin_arm_wunpckehuw (v2si)
11999 v4hi __builtin_arm_wunpckelsb (v8qi)
12000 v2si __builtin_arm_wunpckelsh (v4hi)
12001 long long __builtin_arm_wunpckelsw (v2si)
12002 v4hi __builtin_arm_wunpckelub (v8qi)
12003 v2si __builtin_arm_wunpckeluh (v4hi)
12004 long long __builtin_arm_wunpckeluw (v2si)
12005 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12006 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12007 v2si __builtin_arm_wunpckihw (v2si, v2si)
12008 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12009 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12010 v2si __builtin_arm_wunpckilw (v2si, v2si)
12011 long long __builtin_arm_wxor (long long, long long)
12012 long long __builtin_arm_wzero ()
12016 @node ARM C Language Extensions (ACLE)
12017 @subsection ARM C Language Extensions (ACLE)
12019 GCC implements extensions for C as described in the ARM C Language
12020 Extensions (ACLE) specification, which can be found at
12021 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12023 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12024 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12025 intrinsics can be found at
12026 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12027 The built-in intrinsics for the Advanced SIMD extension are available when
12030 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12031 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12032 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12033 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12036 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12037 availability of extensions.
12039 @node ARM Floating Point Status and Control Intrinsics
12040 @subsection ARM Floating Point Status and Control Intrinsics
12042 These built-in functions are available for the ARM family of
12043 processors with floating-point unit.
12046 unsigned int __builtin_arm_get_fpscr ()
12047 void __builtin_arm_set_fpscr (unsigned int)
12050 @node AVR Built-in Functions
12051 @subsection AVR Built-in Functions
12053 For each built-in function for AVR, there is an equally named,
12054 uppercase built-in macro defined. That way users can easily query if
12055 or if not a specific built-in is implemented or not. For example, if
12056 @code{__builtin_avr_nop} is available the macro
12057 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12059 The following built-in functions map to the respective machine
12060 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12061 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12062 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12063 as library call if no hardware multiplier is available.
12066 void __builtin_avr_nop (void)
12067 void __builtin_avr_sei (void)
12068 void __builtin_avr_cli (void)
12069 void __builtin_avr_sleep (void)
12070 void __builtin_avr_wdr (void)
12071 unsigned char __builtin_avr_swap (unsigned char)
12072 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12073 int __builtin_avr_fmuls (char, char)
12074 int __builtin_avr_fmulsu (char, unsigned char)
12077 In order to delay execution for a specific number of cycles, GCC
12080 void __builtin_avr_delay_cycles (unsigned long ticks)
12084 @code{ticks} is the number of ticks to delay execution. Note that this
12085 built-in does not take into account the effect of interrupts that
12086 might increase delay time. @code{ticks} must be a compile-time
12087 integer constant; delays with a variable number of cycles are not supported.
12090 char __builtin_avr_flash_segment (const __memx void*)
12094 This built-in takes a byte address to the 24-bit
12095 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12096 the number of the flash segment (the 64 KiB chunk) where the address
12097 points to. Counting starts at @code{0}.
12098 If the address does not point to flash memory, return @code{-1}.
12101 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12105 Insert bits from @var{bits} into @var{val} and return the resulting
12106 value. The nibbles of @var{map} determine how the insertion is
12107 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12109 @item If @var{X} is @code{0xf},
12110 then the @var{n}-th bit of @var{val} is returned unaltered.
12112 @item If X is in the range 0@dots{}7,
12113 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12115 @item If X is in the range 8@dots{}@code{0xe},
12116 then the @var{n}-th result bit is undefined.
12120 One typical use case for this built-in is adjusting input and
12121 output values to non-contiguous port layouts. Some examples:
12124 // same as val, bits is unused
12125 __builtin_avr_insert_bits (0xffffffff, bits, val)
12129 // same as bits, val is unused
12130 __builtin_avr_insert_bits (0x76543210, bits, val)
12134 // same as rotating bits by 4
12135 __builtin_avr_insert_bits (0x32107654, bits, 0)
12139 // high nibble of result is the high nibble of val
12140 // low nibble of result is the low nibble of bits
12141 __builtin_avr_insert_bits (0xffff3210, bits, val)
12145 // reverse the bit order of bits
12146 __builtin_avr_insert_bits (0x01234567, bits, 0)
12149 @node Blackfin Built-in Functions
12150 @subsection Blackfin Built-in Functions
12152 Currently, there are two Blackfin-specific built-in functions. These are
12153 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12154 using inline assembly; by using these built-in functions the compiler can
12155 automatically add workarounds for hardware errata involving these
12156 instructions. These functions are named as follows:
12159 void __builtin_bfin_csync (void)
12160 void __builtin_bfin_ssync (void)
12163 @node FR-V Built-in Functions
12164 @subsection FR-V Built-in Functions
12166 GCC provides many FR-V-specific built-in functions. In general,
12167 these functions are intended to be compatible with those described
12168 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12169 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12170 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12171 pointer rather than by value.
12173 Most of the functions are named after specific FR-V instructions.
12174 Such functions are said to be ``directly mapped'' and are summarized
12175 here in tabular form.
12179 * Directly-mapped Integer Functions::
12180 * Directly-mapped Media Functions::
12181 * Raw read/write Functions::
12182 * Other Built-in Functions::
12185 @node Argument Types
12186 @subsubsection Argument Types
12188 The arguments to the built-in functions can be divided into three groups:
12189 register numbers, compile-time constants and run-time values. In order
12190 to make this classification clear at a glance, the arguments and return
12191 values are given the following pseudo types:
12193 @multitable @columnfractions .20 .30 .15 .35
12194 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12195 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12196 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12197 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12198 @item @code{uw2} @tab @code{unsigned long long} @tab No
12199 @tab an unsigned doubleword
12200 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12201 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12202 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12203 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12206 These pseudo types are not defined by GCC, they are simply a notational
12207 convenience used in this manual.
12209 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12210 and @code{sw2} are evaluated at run time. They correspond to
12211 register operands in the underlying FR-V instructions.
12213 @code{const} arguments represent immediate operands in the underlying
12214 FR-V instructions. They must be compile-time constants.
12216 @code{acc} arguments are evaluated at compile time and specify the number
12217 of an accumulator register. For example, an @code{acc} argument of 2
12218 selects the ACC2 register.
12220 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12221 number of an IACC register. See @pxref{Other Built-in Functions}
12224 @node Directly-mapped Integer Functions
12225 @subsubsection Directly-Mapped Integer Functions
12227 The functions listed below map directly to FR-V I-type instructions.
12229 @multitable @columnfractions .45 .32 .23
12230 @item Function prototype @tab Example usage @tab Assembly output
12231 @item @code{sw1 __ADDSS (sw1, sw1)}
12232 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12233 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12234 @item @code{sw1 __SCAN (sw1, sw1)}
12235 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12236 @tab @code{SCAN @var{a},@var{b},@var{c}}
12237 @item @code{sw1 __SCUTSS (sw1)}
12238 @tab @code{@var{b} = __SCUTSS (@var{a})}
12239 @tab @code{SCUTSS @var{a},@var{b}}
12240 @item @code{sw1 __SLASS (sw1, sw1)}
12241 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12242 @tab @code{SLASS @var{a},@var{b},@var{c}}
12243 @item @code{void __SMASS (sw1, sw1)}
12244 @tab @code{__SMASS (@var{a}, @var{b})}
12245 @tab @code{SMASS @var{a},@var{b}}
12246 @item @code{void __SMSSS (sw1, sw1)}
12247 @tab @code{__SMSSS (@var{a}, @var{b})}
12248 @tab @code{SMSSS @var{a},@var{b}}
12249 @item @code{void __SMU (sw1, sw1)}
12250 @tab @code{__SMU (@var{a}, @var{b})}
12251 @tab @code{SMU @var{a},@var{b}}
12252 @item @code{sw2 __SMUL (sw1, sw1)}
12253 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12254 @tab @code{SMUL @var{a},@var{b},@var{c}}
12255 @item @code{sw1 __SUBSS (sw1, sw1)}
12256 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12257 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12258 @item @code{uw2 __UMUL (uw1, uw1)}
12259 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12260 @tab @code{UMUL @var{a},@var{b},@var{c}}
12263 @node Directly-mapped Media Functions
12264 @subsubsection Directly-Mapped Media Functions
12266 The functions listed below map directly to FR-V M-type instructions.
12268 @multitable @columnfractions .45 .32 .23
12269 @item Function prototype @tab Example usage @tab Assembly output
12270 @item @code{uw1 __MABSHS (sw1)}
12271 @tab @code{@var{b} = __MABSHS (@var{a})}
12272 @tab @code{MABSHS @var{a},@var{b}}
12273 @item @code{void __MADDACCS (acc, acc)}
12274 @tab @code{__MADDACCS (@var{b}, @var{a})}
12275 @tab @code{MADDACCS @var{a},@var{b}}
12276 @item @code{sw1 __MADDHSS (sw1, sw1)}
12277 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12278 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12279 @item @code{uw1 __MADDHUS (uw1, uw1)}
12280 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12281 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12282 @item @code{uw1 __MAND (uw1, uw1)}
12283 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12284 @tab @code{MAND @var{a},@var{b},@var{c}}
12285 @item @code{void __MASACCS (acc, acc)}
12286 @tab @code{__MASACCS (@var{b}, @var{a})}
12287 @tab @code{MASACCS @var{a},@var{b}}
12288 @item @code{uw1 __MAVEH (uw1, uw1)}
12289 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12290 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12291 @item @code{uw2 __MBTOH (uw1)}
12292 @tab @code{@var{b} = __MBTOH (@var{a})}
12293 @tab @code{MBTOH @var{a},@var{b}}
12294 @item @code{void __MBTOHE (uw1 *, uw1)}
12295 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12296 @tab @code{MBTOHE @var{a},@var{b}}
12297 @item @code{void __MCLRACC (acc)}
12298 @tab @code{__MCLRACC (@var{a})}
12299 @tab @code{MCLRACC @var{a}}
12300 @item @code{void __MCLRACCA (void)}
12301 @tab @code{__MCLRACCA ()}
12302 @tab @code{MCLRACCA}
12303 @item @code{uw1 __Mcop1 (uw1, uw1)}
12304 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12305 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12306 @item @code{uw1 __Mcop2 (uw1, uw1)}
12307 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12308 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12309 @item @code{uw1 __MCPLHI (uw2, const)}
12310 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12311 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12312 @item @code{uw1 __MCPLI (uw2, const)}
12313 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12314 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12315 @item @code{void __MCPXIS (acc, sw1, sw1)}
12316 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12317 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12318 @item @code{void __MCPXIU (acc, uw1, uw1)}
12319 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12320 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12321 @item @code{void __MCPXRS (acc, sw1, sw1)}
12322 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12323 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12324 @item @code{void __MCPXRU (acc, uw1, uw1)}
12325 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12326 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12327 @item @code{uw1 __MCUT (acc, uw1)}
12328 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12329 @tab @code{MCUT @var{a},@var{b},@var{c}}
12330 @item @code{uw1 __MCUTSS (acc, sw1)}
12331 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12332 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12333 @item @code{void __MDADDACCS (acc, acc)}
12334 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12335 @tab @code{MDADDACCS @var{a},@var{b}}
12336 @item @code{void __MDASACCS (acc, acc)}
12337 @tab @code{__MDASACCS (@var{b}, @var{a})}
12338 @tab @code{MDASACCS @var{a},@var{b}}
12339 @item @code{uw2 __MDCUTSSI (acc, const)}
12340 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12341 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12342 @item @code{uw2 __MDPACKH (uw2, uw2)}
12343 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12344 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12345 @item @code{uw2 __MDROTLI (uw2, const)}
12346 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12347 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12348 @item @code{void __MDSUBACCS (acc, acc)}
12349 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12350 @tab @code{MDSUBACCS @var{a},@var{b}}
12351 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12352 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12353 @tab @code{MDUNPACKH @var{a},@var{b}}
12354 @item @code{uw2 __MEXPDHD (uw1, const)}
12355 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12356 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12357 @item @code{uw1 __MEXPDHW (uw1, const)}
12358 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12359 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12360 @item @code{uw1 __MHDSETH (uw1, const)}
12361 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12362 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12363 @item @code{sw1 __MHDSETS (const)}
12364 @tab @code{@var{b} = __MHDSETS (@var{a})}
12365 @tab @code{MHDSETS #@var{a},@var{b}}
12366 @item @code{uw1 __MHSETHIH (uw1, const)}
12367 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12368 @tab @code{MHSETHIH #@var{a},@var{b}}
12369 @item @code{sw1 __MHSETHIS (sw1, const)}
12370 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12371 @tab @code{MHSETHIS #@var{a},@var{b}}
12372 @item @code{uw1 __MHSETLOH (uw1, const)}
12373 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12374 @tab @code{MHSETLOH #@var{a},@var{b}}
12375 @item @code{sw1 __MHSETLOS (sw1, const)}
12376 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12377 @tab @code{MHSETLOS #@var{a},@var{b}}
12378 @item @code{uw1 __MHTOB (uw2)}
12379 @tab @code{@var{b} = __MHTOB (@var{a})}
12380 @tab @code{MHTOB @var{a},@var{b}}
12381 @item @code{void __MMACHS (acc, sw1, sw1)}
12382 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12383 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12384 @item @code{void __MMACHU (acc, uw1, uw1)}
12385 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12386 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12387 @item @code{void __MMRDHS (acc, sw1, sw1)}
12388 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12389 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12390 @item @code{void __MMRDHU (acc, uw1, uw1)}
12391 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12392 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12393 @item @code{void __MMULHS (acc, sw1, sw1)}
12394 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12395 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12396 @item @code{void __MMULHU (acc, uw1, uw1)}
12397 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12398 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12399 @item @code{void __MMULXHS (acc, sw1, sw1)}
12400 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12401 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12402 @item @code{void __MMULXHU (acc, uw1, uw1)}
12403 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12404 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12405 @item @code{uw1 __MNOT (uw1)}
12406 @tab @code{@var{b} = __MNOT (@var{a})}
12407 @tab @code{MNOT @var{a},@var{b}}
12408 @item @code{uw1 __MOR (uw1, uw1)}
12409 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12410 @tab @code{MOR @var{a},@var{b},@var{c}}
12411 @item @code{uw1 __MPACKH (uh, uh)}
12412 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12413 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12414 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12415 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12416 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12417 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12418 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12419 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12420 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12421 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12422 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12423 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12424 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12425 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12426 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12427 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12428 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12429 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12430 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12431 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12432 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12433 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12434 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12435 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12436 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12437 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12438 @item @code{void __MQMACHS (acc, sw2, sw2)}
12439 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12440 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12441 @item @code{void __MQMACHU (acc, uw2, uw2)}
12442 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12443 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12444 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12445 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12446 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12447 @item @code{void __MQMULHS (acc, sw2, sw2)}
12448 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12449 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12450 @item @code{void __MQMULHU (acc, uw2, uw2)}
12451 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12452 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12453 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12454 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12455 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12456 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12457 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12458 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12459 @item @code{sw2 __MQSATHS (sw2, sw2)}
12460 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12461 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12462 @item @code{uw2 __MQSLLHI (uw2, int)}
12463 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12464 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12465 @item @code{sw2 __MQSRAHI (sw2, int)}
12466 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12467 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12468 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12469 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12470 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12471 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12472 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12473 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12474 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12475 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12476 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12477 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12478 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12479 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12480 @item @code{uw1 __MRDACC (acc)}
12481 @tab @code{@var{b} = __MRDACC (@var{a})}
12482 @tab @code{MRDACC @var{a},@var{b}}
12483 @item @code{uw1 __MRDACCG (acc)}
12484 @tab @code{@var{b} = __MRDACCG (@var{a})}
12485 @tab @code{MRDACCG @var{a},@var{b}}
12486 @item @code{uw1 __MROTLI (uw1, const)}
12487 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12488 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12489 @item @code{uw1 __MROTRI (uw1, const)}
12490 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12491 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12492 @item @code{sw1 __MSATHS (sw1, sw1)}
12493 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12494 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12495 @item @code{uw1 __MSATHU (uw1, uw1)}
12496 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12497 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12498 @item @code{uw1 __MSLLHI (uw1, const)}
12499 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12500 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12501 @item @code{sw1 __MSRAHI (sw1, const)}
12502 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12503 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12504 @item @code{uw1 __MSRLHI (uw1, const)}
12505 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12506 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12507 @item @code{void __MSUBACCS (acc, acc)}
12508 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12509 @tab @code{MSUBACCS @var{a},@var{b}}
12510 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12511 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12512 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12513 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12514 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12515 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12516 @item @code{void __MTRAP (void)}
12517 @tab @code{__MTRAP ()}
12519 @item @code{uw2 __MUNPACKH (uw1)}
12520 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12521 @tab @code{MUNPACKH @var{a},@var{b}}
12522 @item @code{uw1 __MWCUT (uw2, uw1)}
12523 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12524 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12525 @item @code{void __MWTACC (acc, uw1)}
12526 @tab @code{__MWTACC (@var{b}, @var{a})}
12527 @tab @code{MWTACC @var{a},@var{b}}
12528 @item @code{void __MWTACCG (acc, uw1)}
12529 @tab @code{__MWTACCG (@var{b}, @var{a})}
12530 @tab @code{MWTACCG @var{a},@var{b}}
12531 @item @code{uw1 __MXOR (uw1, uw1)}
12532 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12533 @tab @code{MXOR @var{a},@var{b},@var{c}}
12536 @node Raw read/write Functions
12537 @subsubsection Raw Read/Write Functions
12539 This sections describes built-in functions related to read and write
12540 instructions to access memory. These functions generate
12541 @code{membar} instructions to flush the I/O load and stores where
12542 appropriate, as described in Fujitsu's manual described above.
12546 @item unsigned char __builtin_read8 (void *@var{data})
12547 @item unsigned short __builtin_read16 (void *@var{data})
12548 @item unsigned long __builtin_read32 (void *@var{data})
12549 @item unsigned long long __builtin_read64 (void *@var{data})
12551 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12552 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12553 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12554 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12557 @node Other Built-in Functions
12558 @subsubsection Other Built-in Functions
12560 This section describes built-in functions that are not named after
12561 a specific FR-V instruction.
12564 @item sw2 __IACCreadll (iacc @var{reg})
12565 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12566 for future expansion and must be 0.
12568 @item sw1 __IACCreadl (iacc @var{reg})
12569 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12570 Other values of @var{reg} are rejected as invalid.
12572 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12573 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12574 is reserved for future expansion and must be 0.
12576 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12577 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12578 is 1. Other values of @var{reg} are rejected as invalid.
12580 @item void __data_prefetch0 (const void *@var{x})
12581 Use the @code{dcpl} instruction to load the contents of address @var{x}
12582 into the data cache.
12584 @item void __data_prefetch (const void *@var{x})
12585 Use the @code{nldub} instruction to load the contents of address @var{x}
12586 into the data cache. The instruction is issued in slot I1@.
12589 @node MIPS DSP Built-in Functions
12590 @subsection MIPS DSP Built-in Functions
12592 The MIPS DSP Application-Specific Extension (ASE) includes new
12593 instructions that are designed to improve the performance of DSP and
12594 media applications. It provides instructions that operate on packed
12595 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12597 GCC supports MIPS DSP operations using both the generic
12598 vector extensions (@pxref{Vector Extensions}) and a collection of
12599 MIPS-specific built-in functions. Both kinds of support are
12600 enabled by the @option{-mdsp} command-line option.
12602 Revision 2 of the ASE was introduced in the second half of 2006.
12603 This revision adds extra instructions to the original ASE, but is
12604 otherwise backwards-compatible with it. You can select revision 2
12605 using the command-line option @option{-mdspr2}; this option implies
12608 The SCOUNT and POS bits of the DSP control register are global. The
12609 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12610 POS bits. During optimization, the compiler does not delete these
12611 instructions and it does not delete calls to functions containing
12612 these instructions.
12614 At present, GCC only provides support for operations on 32-bit
12615 vectors. The vector type associated with 8-bit integer data is
12616 usually called @code{v4i8}, the vector type associated with Q7
12617 is usually called @code{v4q7}, the vector type associated with 16-bit
12618 integer data is usually called @code{v2i16}, and the vector type
12619 associated with Q15 is usually called @code{v2q15}. They can be
12620 defined in C as follows:
12623 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12624 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12625 typedef short v2i16 __attribute__ ((vector_size(4)));
12626 typedef short v2q15 __attribute__ ((vector_size(4)));
12629 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12630 initialized in the same way as aggregates. For example:
12633 v4i8 a = @{1, 2, 3, 4@};
12635 b = (v4i8) @{5, 6, 7, 8@};
12637 v2q15 c = @{0x0fcb, 0x3a75@};
12639 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12642 @emph{Note:} The CPU's endianness determines the order in which values
12643 are packed. On little-endian targets, the first value is the least
12644 significant and the last value is the most significant. The opposite
12645 order applies to big-endian targets. For example, the code above
12646 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12647 and @code{4} on big-endian targets.
12649 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12650 representation. As shown in this example, the integer representation
12651 of a Q7 value can be obtained by multiplying the fractional value by
12652 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12653 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12656 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12657 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12658 and @code{c} and @code{d} are @code{v2q15} values.
12660 @multitable @columnfractions .50 .50
12661 @item C code @tab MIPS instruction
12662 @item @code{a + b} @tab @code{addu.qb}
12663 @item @code{c + d} @tab @code{addq.ph}
12664 @item @code{a - b} @tab @code{subu.qb}
12665 @item @code{c - d} @tab @code{subq.ph}
12668 The table below lists the @code{v2i16} operation for which
12669 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12670 @code{v2i16} values.
12672 @multitable @columnfractions .50 .50
12673 @item C code @tab MIPS instruction
12674 @item @code{e * f} @tab @code{mul.ph}
12677 It is easier to describe the DSP built-in functions if we first define
12678 the following types:
12683 typedef unsigned int ui32;
12684 typedef long long a64;
12687 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12688 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12689 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12690 @code{long long}, but we use @code{a64} to indicate values that are
12691 placed in one of the four DSP accumulators (@code{$ac0},
12692 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12694 Also, some built-in functions prefer or require immediate numbers as
12695 parameters, because the corresponding DSP instructions accept both immediate
12696 numbers and register operands, or accept immediate numbers only. The
12697 immediate parameters are listed as follows.
12705 imm0_255: 0 to 255.
12706 imm_n32_31: -32 to 31.
12707 imm_n512_511: -512 to 511.
12710 The following built-in functions map directly to a particular MIPS DSP
12711 instruction. Please refer to the architecture specification
12712 for details on what each instruction does.
12715 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12716 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12717 q31 __builtin_mips_addq_s_w (q31, q31)
12718 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12719 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12720 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12721 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12722 q31 __builtin_mips_subq_s_w (q31, q31)
12723 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12724 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12725 i32 __builtin_mips_addsc (i32, i32)
12726 i32 __builtin_mips_addwc (i32, i32)
12727 i32 __builtin_mips_modsub (i32, i32)
12728 i32 __builtin_mips_raddu_w_qb (v4i8)
12729 v2q15 __builtin_mips_absq_s_ph (v2q15)
12730 q31 __builtin_mips_absq_s_w (q31)
12731 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12732 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12733 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12734 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12735 q31 __builtin_mips_preceq_w_phl (v2q15)
12736 q31 __builtin_mips_preceq_w_phr (v2q15)
12737 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12738 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12739 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12740 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12741 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12742 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12743 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12744 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12745 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12746 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12747 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12748 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12749 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12750 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12751 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12752 q31 __builtin_mips_shll_s_w (q31, i32)
12753 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12754 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12755 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12756 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12757 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12758 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12759 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12760 q31 __builtin_mips_shra_r_w (q31, i32)
12761 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12762 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12763 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12764 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12765 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12766 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12767 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12768 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12769 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12770 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12771 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12772 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12773 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12774 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12775 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12776 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12777 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12778 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12779 i32 __builtin_mips_bitrev (i32)
12780 i32 __builtin_mips_insv (i32, i32)
12781 v4i8 __builtin_mips_repl_qb (imm0_255)
12782 v4i8 __builtin_mips_repl_qb (i32)
12783 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12784 v2q15 __builtin_mips_repl_ph (i32)
12785 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12786 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12787 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12788 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12789 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12790 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12791 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12792 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12793 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12794 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12795 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12796 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12797 i32 __builtin_mips_extr_w (a64, imm0_31)
12798 i32 __builtin_mips_extr_w (a64, i32)
12799 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12800 i32 __builtin_mips_extr_s_h (a64, i32)
12801 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12802 i32 __builtin_mips_extr_rs_w (a64, i32)
12803 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12804 i32 __builtin_mips_extr_r_w (a64, i32)
12805 i32 __builtin_mips_extp (a64, imm0_31)
12806 i32 __builtin_mips_extp (a64, i32)
12807 i32 __builtin_mips_extpdp (a64, imm0_31)
12808 i32 __builtin_mips_extpdp (a64, i32)
12809 a64 __builtin_mips_shilo (a64, imm_n32_31)
12810 a64 __builtin_mips_shilo (a64, i32)
12811 a64 __builtin_mips_mthlip (a64, i32)
12812 void __builtin_mips_wrdsp (i32, imm0_63)
12813 i32 __builtin_mips_rddsp (imm0_63)
12814 i32 __builtin_mips_lbux (void *, i32)
12815 i32 __builtin_mips_lhx (void *, i32)
12816 i32 __builtin_mips_lwx (void *, i32)
12817 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12818 i32 __builtin_mips_bposge32 (void)
12819 a64 __builtin_mips_madd (a64, i32, i32);
12820 a64 __builtin_mips_maddu (a64, ui32, ui32);
12821 a64 __builtin_mips_msub (a64, i32, i32);
12822 a64 __builtin_mips_msubu (a64, ui32, ui32);
12823 a64 __builtin_mips_mult (i32, i32);
12824 a64 __builtin_mips_multu (ui32, ui32);
12827 The following built-in functions map directly to a particular MIPS DSP REV 2
12828 instruction. Please refer to the architecture specification
12829 for details on what each instruction does.
12832 v4q7 __builtin_mips_absq_s_qb (v4q7);
12833 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12834 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12835 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12836 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12837 i32 __builtin_mips_append (i32, i32, imm0_31);
12838 i32 __builtin_mips_balign (i32, i32, imm0_3);
12839 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12840 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12841 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12842 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12843 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12844 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12845 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12846 q31 __builtin_mips_mulq_rs_w (q31, q31);
12847 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12848 q31 __builtin_mips_mulq_s_w (q31, q31);
12849 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12850 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12851 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12852 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12853 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12854 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12855 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12856 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12857 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12858 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12859 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12860 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12861 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12862 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12863 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12864 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12865 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12866 q31 __builtin_mips_addqh_w (q31, q31);
12867 q31 __builtin_mips_addqh_r_w (q31, q31);
12868 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12869 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12870 q31 __builtin_mips_subqh_w (q31, q31);
12871 q31 __builtin_mips_subqh_r_w (q31, q31);
12872 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12873 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12874 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12875 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12876 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12877 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12881 @node MIPS Paired-Single Support
12882 @subsection MIPS Paired-Single Support
12884 The MIPS64 architecture includes a number of instructions that
12885 operate on pairs of single-precision floating-point values.
12886 Each pair is packed into a 64-bit floating-point register,
12887 with one element being designated the ``upper half'' and
12888 the other being designated the ``lower half''.
12890 GCC supports paired-single operations using both the generic
12891 vector extensions (@pxref{Vector Extensions}) and a collection of
12892 MIPS-specific built-in functions. Both kinds of support are
12893 enabled by the @option{-mpaired-single} command-line option.
12895 The vector type associated with paired-single values is usually
12896 called @code{v2sf}. It can be defined in C as follows:
12899 typedef float v2sf __attribute__ ((vector_size (8)));
12902 @code{v2sf} values are initialized in the same way as aggregates.
12906 v2sf a = @{1.5, 9.1@};
12909 b = (v2sf) @{e, f@};
12912 @emph{Note:} The CPU's endianness determines which value is stored in
12913 the upper half of a register and which value is stored in the lower half.
12914 On little-endian targets, the first value is the lower one and the second
12915 value is the upper one. The opposite order applies to big-endian targets.
12916 For example, the code above sets the lower half of @code{a} to
12917 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
12919 @node MIPS Loongson Built-in Functions
12920 @subsection MIPS Loongson Built-in Functions
12922 GCC provides intrinsics to access the SIMD instructions provided by the
12923 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
12924 available after inclusion of the @code{loongson.h} header file,
12925 operate on the following 64-bit vector types:
12928 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
12929 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
12930 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
12931 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
12932 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
12933 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
12936 The intrinsics provided are listed below; each is named after the
12937 machine instruction to which it corresponds, with suffixes added as
12938 appropriate to distinguish intrinsics that expand to the same machine
12939 instruction yet have different argument types. Refer to the architecture
12940 documentation for a description of the functionality of each
12944 int16x4_t packsswh (int32x2_t s, int32x2_t t);
12945 int8x8_t packsshb (int16x4_t s, int16x4_t t);
12946 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
12947 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
12948 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
12949 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
12950 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
12951 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
12952 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
12953 uint64_t paddd_u (uint64_t s, uint64_t t);
12954 int64_t paddd_s (int64_t s, int64_t t);
12955 int16x4_t paddsh (int16x4_t s, int16x4_t t);
12956 int8x8_t paddsb (int8x8_t s, int8x8_t t);
12957 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
12958 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
12959 uint64_t pandn_ud (uint64_t s, uint64_t t);
12960 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
12961 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
12962 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
12963 int64_t pandn_sd (int64_t s, int64_t t);
12964 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
12965 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
12966 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
12967 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
12968 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
12969 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
12970 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
12971 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
12972 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
12973 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
12974 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
12975 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
12976 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
12977 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
12978 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
12979 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
12980 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
12981 uint16x4_t pextrh_u (uint16x4_t s, int field);
12982 int16x4_t pextrh_s (int16x4_t s, int field);
12983 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
12984 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
12985 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
12986 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
12987 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
12988 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
12989 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
12990 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
12991 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
12992 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
12993 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
12994 int16x4_t pminsh (int16x4_t s, int16x4_t t);
12995 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
12996 uint8x8_t pmovmskb_u (uint8x8_t s);
12997 int8x8_t pmovmskb_s (int8x8_t s);
12998 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
12999 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13000 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13001 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13002 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13003 uint16x4_t biadd (uint8x8_t s);
13004 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13005 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13006 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13007 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13008 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13009 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13010 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13011 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13012 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13013 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13014 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13015 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13016 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13017 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13018 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13019 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13020 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13021 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13022 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13023 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13024 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13025 uint64_t psubd_u (uint64_t s, uint64_t t);
13026 int64_t psubd_s (int64_t s, int64_t t);
13027 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13028 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13029 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13030 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13031 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13032 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13033 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13034 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13035 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13036 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13037 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13038 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13039 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13040 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13041 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13042 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13046 * Paired-Single Arithmetic::
13047 * Paired-Single Built-in Functions::
13048 * MIPS-3D Built-in Functions::
13051 @node Paired-Single Arithmetic
13052 @subsubsection Paired-Single Arithmetic
13054 The table below lists the @code{v2sf} operations for which hardware
13055 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13056 values and @code{x} is an integral value.
13058 @multitable @columnfractions .50 .50
13059 @item C code @tab MIPS instruction
13060 @item @code{a + b} @tab @code{add.ps}
13061 @item @code{a - b} @tab @code{sub.ps}
13062 @item @code{-a} @tab @code{neg.ps}
13063 @item @code{a * b} @tab @code{mul.ps}
13064 @item @code{a * b + c} @tab @code{madd.ps}
13065 @item @code{a * b - c} @tab @code{msub.ps}
13066 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13067 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13068 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13071 Note that the multiply-accumulate instructions can be disabled
13072 using the command-line option @code{-mno-fused-madd}.
13074 @node Paired-Single Built-in Functions
13075 @subsubsection Paired-Single Built-in Functions
13077 The following paired-single functions map directly to a particular
13078 MIPS instruction. Please refer to the architecture specification
13079 for details on what each instruction does.
13082 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13083 Pair lower lower (@code{pll.ps}).
13085 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13086 Pair upper lower (@code{pul.ps}).
13088 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13089 Pair lower upper (@code{plu.ps}).
13091 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13092 Pair upper upper (@code{puu.ps}).
13094 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13095 Convert pair to paired single (@code{cvt.ps.s}).
13097 @item float __builtin_mips_cvt_s_pl (v2sf)
13098 Convert pair lower to single (@code{cvt.s.pl}).
13100 @item float __builtin_mips_cvt_s_pu (v2sf)
13101 Convert pair upper to single (@code{cvt.s.pu}).
13103 @item v2sf __builtin_mips_abs_ps (v2sf)
13104 Absolute value (@code{abs.ps}).
13106 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13107 Align variable (@code{alnv.ps}).
13109 @emph{Note:} The value of the third parameter must be 0 or 4
13110 modulo 8, otherwise the result is unpredictable. Please read the
13111 instruction description for details.
13114 The following multi-instruction functions are also available.
13115 In each case, @var{cond} can be any of the 16 floating-point conditions:
13116 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13117 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13118 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13121 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13122 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13123 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13124 @code{movt.ps}/@code{movf.ps}).
13126 The @code{movt} functions return the value @var{x} computed by:
13129 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13130 mov.ps @var{x},@var{c}
13131 movt.ps @var{x},@var{d},@var{cc}
13134 The @code{movf} functions are similar but use @code{movf.ps} instead
13137 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13138 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13139 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13140 @code{bc1t}/@code{bc1f}).
13142 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13143 and return either the upper or lower half of the result. For example:
13147 if (__builtin_mips_upper_c_eq_ps (a, b))
13148 upper_halves_are_equal ();
13150 upper_halves_are_unequal ();
13152 if (__builtin_mips_lower_c_eq_ps (a, b))
13153 lower_halves_are_equal ();
13155 lower_halves_are_unequal ();
13159 @node MIPS-3D Built-in Functions
13160 @subsubsection MIPS-3D Built-in Functions
13162 The MIPS-3D Application-Specific Extension (ASE) includes additional
13163 paired-single instructions that are designed to improve the performance
13164 of 3D graphics operations. Support for these instructions is controlled
13165 by the @option{-mips3d} command-line option.
13167 The functions listed below map directly to a particular MIPS-3D
13168 instruction. Please refer to the architecture specification for
13169 more details on what each instruction does.
13172 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13173 Reduction add (@code{addr.ps}).
13175 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13176 Reduction multiply (@code{mulr.ps}).
13178 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13179 Convert paired single to paired word (@code{cvt.pw.ps}).
13181 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13182 Convert paired word to paired single (@code{cvt.ps.pw}).
13184 @item float __builtin_mips_recip1_s (float)
13185 @itemx double __builtin_mips_recip1_d (double)
13186 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13187 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13189 @item float __builtin_mips_recip2_s (float, float)
13190 @itemx double __builtin_mips_recip2_d (double, double)
13191 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13192 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13194 @item float __builtin_mips_rsqrt1_s (float)
13195 @itemx double __builtin_mips_rsqrt1_d (double)
13196 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13197 Reduced-precision reciprocal square root (sequence step 1)
13198 (@code{rsqrt1.@var{fmt}}).
13200 @item float __builtin_mips_rsqrt2_s (float, float)
13201 @itemx double __builtin_mips_rsqrt2_d (double, double)
13202 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13203 Reduced-precision reciprocal square root (sequence step 2)
13204 (@code{rsqrt2.@var{fmt}}).
13207 The following multi-instruction functions are also available.
13208 In each case, @var{cond} can be any of the 16 floating-point conditions:
13209 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13210 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13211 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13214 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13215 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13216 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13217 @code{bc1t}/@code{bc1f}).
13219 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13220 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13225 if (__builtin_mips_cabs_eq_s (a, b))
13231 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13232 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13233 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13234 @code{bc1t}/@code{bc1f}).
13236 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13237 and return either the upper or lower half of the result. For example:
13241 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13242 upper_halves_are_equal ();
13244 upper_halves_are_unequal ();
13246 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13247 lower_halves_are_equal ();
13249 lower_halves_are_unequal ();
13252 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13253 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13254 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13255 @code{movt.ps}/@code{movf.ps}).
13257 The @code{movt} functions return the value @var{x} computed by:
13260 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13261 mov.ps @var{x},@var{c}
13262 movt.ps @var{x},@var{d},@var{cc}
13265 The @code{movf} functions are similar but use @code{movf.ps} instead
13268 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13269 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13270 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13271 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13272 Comparison of two paired-single values
13273 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13274 @code{bc1any2t}/@code{bc1any2f}).
13276 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13277 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13278 result is true and the @code{all} forms return true if both results are true.
13283 if (__builtin_mips_any_c_eq_ps (a, b))
13288 if (__builtin_mips_all_c_eq_ps (a, b))
13294 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13295 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13296 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13297 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13298 Comparison of four paired-single values
13299 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13300 @code{bc1any4t}/@code{bc1any4f}).
13302 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13303 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13304 The @code{any} forms return true if any of the four results are true
13305 and the @code{all} forms return true if all four results are true.
13310 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13315 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13322 @node Other MIPS Built-in Functions
13323 @subsection Other MIPS Built-in Functions
13325 GCC provides other MIPS-specific built-in functions:
13328 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13329 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13330 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13331 when this function is available.
13333 @item unsigned int __builtin_mips_get_fcsr (void)
13334 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13335 Get and set the contents of the floating-point control and status register
13336 (FPU control register 31). These functions are only available in hard-float
13337 code but can be called in both MIPS16 and non-MIPS16 contexts.
13339 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13340 register except the condition codes, which GCC assumes are preserved.
13343 @node MSP430 Built-in Functions
13344 @subsection MSP430 Built-in Functions
13346 GCC provides a couple of special builtin functions to aid in the
13347 writing of interrupt handlers in C.
13350 @item __bic_SR_register_on_exit (int @var{mask})
13351 This clears the indicated bits in the saved copy of the status register
13352 currently residing on the stack. This only works inside interrupt
13353 handlers and the changes to the status register will only take affect
13354 once the handler returns.
13356 @item __bis_SR_register_on_exit (int @var{mask})
13357 This sets the indicated bits in the saved copy of the status register
13358 currently residing on the stack. This only works inside interrupt
13359 handlers and the changes to the status register will only take affect
13360 once the handler returns.
13362 @item __delay_cycles (long long @var{cycles})
13363 This inserts an instruction sequence that takes exactly @var{cycles}
13364 cycles (between 0 and about 17E9) to complete. The inserted sequence
13365 may use jumps, loops, or no-ops, and does not interfere with any other
13366 instructions. Note that @var{cycles} must be a compile-time constant
13367 integer - that is, you must pass a number, not a variable that may be
13368 optimized to a constant later. The number of cycles delayed by this
13372 @node NDS32 Built-in Functions
13373 @subsection NDS32 Built-in Functions
13375 These built-in functions are available for the NDS32 target:
13377 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13378 Insert an ISYNC instruction into the instruction stream where
13379 @var{addr} is an instruction address for serialization.
13382 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13383 Insert an ISB instruction into the instruction stream.
13386 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13387 Return the content of a system register which is mapped by @var{sr}.
13390 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13391 Return the content of a user space register which is mapped by @var{usr}.
13394 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13395 Move the @var{value} to a system register which is mapped by @var{sr}.
13398 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13399 Move the @var{value} to a user space register which is mapped by @var{usr}.
13402 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13403 Enable global interrupt.
13406 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13407 Disable global interrupt.
13410 @node picoChip Built-in Functions
13411 @subsection picoChip Built-in Functions
13413 GCC provides an interface to selected machine instructions from the
13414 picoChip instruction set.
13417 @item int __builtin_sbc (int @var{value})
13418 Sign bit count. Return the number of consecutive bits in @var{value}
13419 that have the same value as the sign bit. The result is the number of
13420 leading sign bits minus one, giving the number of redundant sign bits in
13423 @item int __builtin_byteswap (int @var{value})
13424 Byte swap. Return the result of swapping the upper and lower bytes of
13427 @item int __builtin_brev (int @var{value})
13428 Bit reversal. Return the result of reversing the bits in
13429 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13432 @item int __builtin_adds (int @var{x}, int @var{y})
13433 Saturating addition. Return the result of adding @var{x} and @var{y},
13434 storing the value 32767 if the result overflows.
13436 @item int __builtin_subs (int @var{x}, int @var{y})
13437 Saturating subtraction. Return the result of subtracting @var{y} from
13438 @var{x}, storing the value @minus{}32768 if the result overflows.
13440 @item void __builtin_halt (void)
13441 Halt. The processor stops execution. This built-in is useful for
13442 implementing assertions.
13446 @node PowerPC Built-in Functions
13447 @subsection PowerPC Built-in Functions
13449 These built-in functions are available for the PowerPC family of
13452 float __builtin_recipdivf (float, float);
13453 float __builtin_rsqrtf (float);
13454 double __builtin_recipdiv (double, double);
13455 double __builtin_rsqrt (double);
13456 uint64_t __builtin_ppc_get_timebase ();
13457 unsigned long __builtin_ppc_mftb ();
13458 double __builtin_unpack_longdouble (long double, int);
13459 long double __builtin_pack_longdouble (double, double);
13462 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13463 @code{__builtin_rsqrtf} functions generate multiple instructions to
13464 implement the reciprocal sqrt functionality using reciprocal sqrt
13465 estimate instructions.
13467 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13468 functions generate multiple instructions to implement division using
13469 the reciprocal estimate instructions.
13471 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13472 functions generate instructions to read the Time Base Register. The
13473 @code{__builtin_ppc_get_timebase} function may generate multiple
13474 instructions and always returns the 64 bits of the Time Base Register.
13475 The @code{__builtin_ppc_mftb} function always generates one instruction and
13476 returns the Time Base Register value as an unsigned long, throwing away
13477 the most significant word on 32-bit environments.
13479 The following built-in functions are available for the PowerPC family
13480 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13481 or @option{-mpopcntd}):
13483 long __builtin_bpermd (long, long);
13484 int __builtin_divwe (int, int);
13485 int __builtin_divweo (int, int);
13486 unsigned int __builtin_divweu (unsigned int, unsigned int);
13487 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13488 long __builtin_divde (long, long);
13489 long __builtin_divdeo (long, long);
13490 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13491 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13492 unsigned int cdtbcd (unsigned int);
13493 unsigned int cbcdtd (unsigned int);
13494 unsigned int addg6s (unsigned int, unsigned int);
13497 The @code{__builtin_divde}, @code{__builtin_divdeo},
13498 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13499 64-bit environment support ISA 2.06 or later.
13501 The following built-in functions are available for the PowerPC family
13502 of processors when hardware decimal floating point
13503 (@option{-mhard-dfp}) is available:
13505 _Decimal64 __builtin_dxex (_Decimal64);
13506 _Decimal128 __builtin_dxexq (_Decimal128);
13507 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13508 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13509 _Decimal64 __builtin_denbcd (int, _Decimal64);
13510 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13511 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13512 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13513 _Decimal64 __builtin_dscli (_Decimal64, int);
13514 _Decimal128 __builtin_dscliq (_Decimal128, int);
13515 _Decimal64 __builtin_dscri (_Decimal64, int);
13516 _Decimal128 __builtin_dscriq (_Decimal128, int);
13517 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13518 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13521 The following built-in functions are available for the PowerPC family
13522 of processors when the Vector Scalar (vsx) instruction set is
13525 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13526 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13527 unsigned long long);
13530 @node PowerPC AltiVec/VSX Built-in Functions
13531 @subsection PowerPC AltiVec Built-in Functions
13533 GCC provides an interface for the PowerPC family of processors to access
13534 the AltiVec operations described in Motorola's AltiVec Programming
13535 Interface Manual. The interface is made available by including
13536 @code{<altivec.h>} and using @option{-maltivec} and
13537 @option{-mabi=altivec}. The interface supports the following vector
13541 vector unsigned char
13545 vector unsigned short
13546 vector signed short
13550 vector unsigned int
13556 If @option{-mvsx} is used the following additional vector types are
13560 vector unsigned long
13565 The long types are only implemented for 64-bit code generation, and
13566 the long type is only used in the floating point/integer conversion
13569 GCC's implementation of the high-level language interface available from
13570 C and C++ code differs from Motorola's documentation in several ways.
13575 A vector constant is a list of constant expressions within curly braces.
13578 A vector initializer requires no cast if the vector constant is of the
13579 same type as the variable it is initializing.
13582 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13583 vector type is the default signedness of the base type. The default
13584 varies depending on the operating system, so a portable program should
13585 always specify the signedness.
13588 Compiling with @option{-maltivec} adds keywords @code{__vector},
13589 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13590 @code{bool}. When compiling ISO C, the context-sensitive substitution
13591 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13592 disabled. To use them, you must include @code{<altivec.h>} instead.
13595 GCC allows using a @code{typedef} name as the type specifier for a
13599 For C, overloaded functions are implemented with macros so the following
13603 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13607 Since @code{vec_add} is a macro, the vector constant in the example
13608 is treated as four separate arguments. Wrap the entire argument in
13609 parentheses for this to work.
13612 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13613 Internally, GCC uses built-in functions to achieve the functionality in
13614 the aforementioned header file, but they are not supported and are
13615 subject to change without notice.
13617 The following interfaces are supported for the generic and specific
13618 AltiVec operations and the AltiVec predicates. In cases where there
13619 is a direct mapping between generic and specific operations, only the
13620 generic names are shown here, although the specific operations can also
13623 Arguments that are documented as @code{const int} require literal
13624 integral values within the range required for that operation.
13627 vector signed char vec_abs (vector signed char);
13628 vector signed short vec_abs (vector signed short);
13629 vector signed int vec_abs (vector signed int);
13630 vector float vec_abs (vector float);
13632 vector signed char vec_abss (vector signed char);
13633 vector signed short vec_abss (vector signed short);
13634 vector signed int vec_abss (vector signed int);
13636 vector signed char vec_add (vector bool char, vector signed char);
13637 vector signed char vec_add (vector signed char, vector bool char);
13638 vector signed char vec_add (vector signed char, vector signed char);
13639 vector unsigned char vec_add (vector bool char, vector unsigned char);
13640 vector unsigned char vec_add (vector unsigned char, vector bool char);
13641 vector unsigned char vec_add (vector unsigned char,
13642 vector unsigned char);
13643 vector signed short vec_add (vector bool short, vector signed short);
13644 vector signed short vec_add (vector signed short, vector bool short);
13645 vector signed short vec_add (vector signed short, vector signed short);
13646 vector unsigned short vec_add (vector bool short,
13647 vector unsigned short);
13648 vector unsigned short vec_add (vector unsigned short,
13649 vector bool short);
13650 vector unsigned short vec_add (vector unsigned short,
13651 vector unsigned short);
13652 vector signed int vec_add (vector bool int, vector signed int);
13653 vector signed int vec_add (vector signed int, vector bool int);
13654 vector signed int vec_add (vector signed int, vector signed int);
13655 vector unsigned int vec_add (vector bool int, vector unsigned int);
13656 vector unsigned int vec_add (vector unsigned int, vector bool int);
13657 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
13658 vector float vec_add (vector float, vector float);
13660 vector float vec_vaddfp (vector float, vector float);
13662 vector signed int vec_vadduwm (vector bool int, vector signed int);
13663 vector signed int vec_vadduwm (vector signed int, vector bool int);
13664 vector signed int vec_vadduwm (vector signed int, vector signed int);
13665 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
13666 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
13667 vector unsigned int vec_vadduwm (vector unsigned int,
13668 vector unsigned int);
13670 vector signed short vec_vadduhm (vector bool short,
13671 vector signed short);
13672 vector signed short vec_vadduhm (vector signed short,
13673 vector bool short);
13674 vector signed short vec_vadduhm (vector signed short,
13675 vector signed short);
13676 vector unsigned short vec_vadduhm (vector bool short,
13677 vector unsigned short);
13678 vector unsigned short vec_vadduhm (vector unsigned short,
13679 vector bool short);
13680 vector unsigned short vec_vadduhm (vector unsigned short,
13681 vector unsigned short);
13683 vector signed char vec_vaddubm (vector bool char, vector signed char);
13684 vector signed char vec_vaddubm (vector signed char, vector bool char);
13685 vector signed char vec_vaddubm (vector signed char, vector signed char);
13686 vector unsigned char vec_vaddubm (vector bool char,
13687 vector unsigned char);
13688 vector unsigned char vec_vaddubm (vector unsigned char,
13690 vector unsigned char vec_vaddubm (vector unsigned char,
13691 vector unsigned char);
13693 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
13695 vector unsigned char vec_adds (vector bool char, vector unsigned char);
13696 vector unsigned char vec_adds (vector unsigned char, vector bool char);
13697 vector unsigned char vec_adds (vector unsigned char,
13698 vector unsigned char);
13699 vector signed char vec_adds (vector bool char, vector signed char);
13700 vector signed char vec_adds (vector signed char, vector bool char);
13701 vector signed char vec_adds (vector signed char, vector signed char);
13702 vector unsigned short vec_adds (vector bool short,
13703 vector unsigned short);
13704 vector unsigned short vec_adds (vector unsigned short,
13705 vector bool short);
13706 vector unsigned short vec_adds (vector unsigned short,
13707 vector unsigned short);
13708 vector signed short vec_adds (vector bool short, vector signed short);
13709 vector signed short vec_adds (vector signed short, vector bool short);
13710 vector signed short vec_adds (vector signed short, vector signed short);
13711 vector unsigned int vec_adds (vector bool int, vector unsigned int);
13712 vector unsigned int vec_adds (vector unsigned int, vector bool int);
13713 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
13714 vector signed int vec_adds (vector bool int, vector signed int);
13715 vector signed int vec_adds (vector signed int, vector bool int);
13716 vector signed int vec_adds (vector signed int, vector signed int);
13718 vector signed int vec_vaddsws (vector bool int, vector signed int);
13719 vector signed int vec_vaddsws (vector signed int, vector bool int);
13720 vector signed int vec_vaddsws (vector signed int, vector signed int);
13722 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
13723 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
13724 vector unsigned int vec_vadduws (vector unsigned int,
13725 vector unsigned int);
13727 vector signed short vec_vaddshs (vector bool short,
13728 vector signed short);
13729 vector signed short vec_vaddshs (vector signed short,
13730 vector bool short);
13731 vector signed short vec_vaddshs (vector signed short,
13732 vector signed short);
13734 vector unsigned short vec_vadduhs (vector bool short,
13735 vector unsigned short);
13736 vector unsigned short vec_vadduhs (vector unsigned short,
13737 vector bool short);
13738 vector unsigned short vec_vadduhs (vector unsigned short,
13739 vector unsigned short);
13741 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13742 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13743 vector signed char vec_vaddsbs (vector signed char, vector signed char);
13745 vector unsigned char vec_vaddubs (vector bool char,
13746 vector unsigned char);
13747 vector unsigned char vec_vaddubs (vector unsigned char,
13749 vector unsigned char vec_vaddubs (vector unsigned char,
13750 vector unsigned char);
13752 vector float vec_and (vector float, vector float);
13753 vector float vec_and (vector float, vector bool int);
13754 vector float vec_and (vector bool int, vector float);
13755 vector bool int vec_and (vector bool int, vector bool int);
13756 vector signed int vec_and (vector bool int, vector signed int);
13757 vector signed int vec_and (vector signed int, vector bool int);
13758 vector signed int vec_and (vector signed int, vector signed int);
13759 vector unsigned int vec_and (vector bool int, vector unsigned int);
13760 vector unsigned int vec_and (vector unsigned int, vector bool int);
13761 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
13762 vector bool short vec_and (vector bool short, vector bool short);
13763 vector signed short vec_and (vector bool short, vector signed short);
13764 vector signed short vec_and (vector signed short, vector bool short);
13765 vector signed short vec_and (vector signed short, vector signed short);
13766 vector unsigned short vec_and (vector bool short,
13767 vector unsigned short);
13768 vector unsigned short vec_and (vector unsigned short,
13769 vector bool short);
13770 vector unsigned short vec_and (vector unsigned short,
13771 vector unsigned short);
13772 vector signed char vec_and (vector bool char, vector signed char);
13773 vector bool char vec_and (vector bool char, vector bool char);
13774 vector signed char vec_and (vector signed char, vector bool char);
13775 vector signed char vec_and (vector signed char, vector signed char);
13776 vector unsigned char vec_and (vector bool char, vector unsigned char);
13777 vector unsigned char vec_and (vector unsigned char, vector bool char);
13778 vector unsigned char vec_and (vector unsigned char,
13779 vector unsigned char);
13781 vector float vec_andc (vector float, vector float);
13782 vector float vec_andc (vector float, vector bool int);
13783 vector float vec_andc (vector bool int, vector float);
13784 vector bool int vec_andc (vector bool int, vector bool int);
13785 vector signed int vec_andc (vector bool int, vector signed int);
13786 vector signed int vec_andc (vector signed int, vector bool int);
13787 vector signed int vec_andc (vector signed int, vector signed int);
13788 vector unsigned int vec_andc (vector bool int, vector unsigned int);
13789 vector unsigned int vec_andc (vector unsigned int, vector bool int);
13790 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
13791 vector bool short vec_andc (vector bool short, vector bool short);
13792 vector signed short vec_andc (vector bool short, vector signed short);
13793 vector signed short vec_andc (vector signed short, vector bool short);
13794 vector signed short vec_andc (vector signed short, vector signed short);
13795 vector unsigned short vec_andc (vector bool short,
13796 vector unsigned short);
13797 vector unsigned short vec_andc (vector unsigned short,
13798 vector bool short);
13799 vector unsigned short vec_andc (vector unsigned short,
13800 vector unsigned short);
13801 vector signed char vec_andc (vector bool char, vector signed char);
13802 vector bool char vec_andc (vector bool char, vector bool char);
13803 vector signed char vec_andc (vector signed char, vector bool char);
13804 vector signed char vec_andc (vector signed char, vector signed char);
13805 vector unsigned char vec_andc (vector bool char, vector unsigned char);
13806 vector unsigned char vec_andc (vector unsigned char, vector bool char);
13807 vector unsigned char vec_andc (vector unsigned char,
13808 vector unsigned char);
13810 vector unsigned char vec_avg (vector unsigned char,
13811 vector unsigned char);
13812 vector signed char vec_avg (vector signed char, vector signed char);
13813 vector unsigned short vec_avg (vector unsigned short,
13814 vector unsigned short);
13815 vector signed short vec_avg (vector signed short, vector signed short);
13816 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
13817 vector signed int vec_avg (vector signed int, vector signed int);
13819 vector signed int vec_vavgsw (vector signed int, vector signed int);
13821 vector unsigned int vec_vavguw (vector unsigned int,
13822 vector unsigned int);
13824 vector signed short vec_vavgsh (vector signed short,
13825 vector signed short);
13827 vector unsigned short vec_vavguh (vector unsigned short,
13828 vector unsigned short);
13830 vector signed char vec_vavgsb (vector signed char, vector signed char);
13832 vector unsigned char vec_vavgub (vector unsigned char,
13833 vector unsigned char);
13835 vector float vec_copysign (vector float);
13837 vector float vec_ceil (vector float);
13839 vector signed int vec_cmpb (vector float, vector float);
13841 vector bool char vec_cmpeq (vector signed char, vector signed char);
13842 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
13843 vector bool short vec_cmpeq (vector signed short, vector signed short);
13844 vector bool short vec_cmpeq (vector unsigned short,
13845 vector unsigned short);
13846 vector bool int vec_cmpeq (vector signed int, vector signed int);
13847 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
13848 vector bool int vec_cmpeq (vector float, vector float);
13850 vector bool int vec_vcmpeqfp (vector float, vector float);
13852 vector bool int vec_vcmpequw (vector signed int, vector signed int);
13853 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
13855 vector bool short vec_vcmpequh (vector signed short,
13856 vector signed short);
13857 vector bool short vec_vcmpequh (vector unsigned short,
13858 vector unsigned short);
13860 vector bool char vec_vcmpequb (vector signed char, vector signed char);
13861 vector bool char vec_vcmpequb (vector unsigned char,
13862 vector unsigned char);
13864 vector bool int vec_cmpge (vector float, vector float);
13866 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
13867 vector bool char vec_cmpgt (vector signed char, vector signed char);
13868 vector bool short vec_cmpgt (vector unsigned short,
13869 vector unsigned short);
13870 vector bool short vec_cmpgt (vector signed short, vector signed short);
13871 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
13872 vector bool int vec_cmpgt (vector signed int, vector signed int);
13873 vector bool int vec_cmpgt (vector float, vector float);
13875 vector bool int vec_vcmpgtfp (vector float, vector float);
13877 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
13879 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
13881 vector bool short vec_vcmpgtsh (vector signed short,
13882 vector signed short);
13884 vector bool short vec_vcmpgtuh (vector unsigned short,
13885 vector unsigned short);
13887 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
13889 vector bool char vec_vcmpgtub (vector unsigned char,
13890 vector unsigned char);
13892 vector bool int vec_cmple (vector float, vector float);
13894 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
13895 vector bool char vec_cmplt (vector signed char, vector signed char);
13896 vector bool short vec_cmplt (vector unsigned short,
13897 vector unsigned short);
13898 vector bool short vec_cmplt (vector signed short, vector signed short);
13899 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
13900 vector bool int vec_cmplt (vector signed int, vector signed int);
13901 vector bool int vec_cmplt (vector float, vector float);
13903 vector float vec_cpsgn (vector float, vector float);
13905 vector float vec_ctf (vector unsigned int, const int);
13906 vector float vec_ctf (vector signed int, const int);
13907 vector double vec_ctf (vector unsigned long, const int);
13908 vector double vec_ctf (vector signed long, const int);
13910 vector float vec_vcfsx (vector signed int, const int);
13912 vector float vec_vcfux (vector unsigned int, const int);
13914 vector signed int vec_cts (vector float, const int);
13915 vector signed long vec_cts (vector double, const int);
13917 vector unsigned int vec_ctu (vector float, const int);
13918 vector unsigned long vec_ctu (vector double, const int);
13920 void vec_dss (const int);
13922 void vec_dssall (void);
13924 void vec_dst (const vector unsigned char *, int, const int);
13925 void vec_dst (const vector signed char *, int, const int);
13926 void vec_dst (const vector bool char *, int, const int);
13927 void vec_dst (const vector unsigned short *, int, const int);
13928 void vec_dst (const vector signed short *, int, const int);
13929 void vec_dst (const vector bool short *, int, const int);
13930 void vec_dst (const vector pixel *, int, const int);
13931 void vec_dst (const vector unsigned int *, int, const int);
13932 void vec_dst (const vector signed int *, int, const int);
13933 void vec_dst (const vector bool int *, int, const int);
13934 void vec_dst (const vector float *, int, const int);
13935 void vec_dst (const unsigned char *, int, const int);
13936 void vec_dst (const signed char *, int, const int);
13937 void vec_dst (const unsigned short *, int, const int);
13938 void vec_dst (const short *, int, const int);
13939 void vec_dst (const unsigned int *, int, const int);
13940 void vec_dst (const int *, int, const int);
13941 void vec_dst (const unsigned long *, int, const int);
13942 void vec_dst (const long *, int, const int);
13943 void vec_dst (const float *, int, const int);
13945 void vec_dstst (const vector unsigned char *, int, const int);
13946 void vec_dstst (const vector signed char *, int, const int);
13947 void vec_dstst (const vector bool char *, int, const int);
13948 void vec_dstst (const vector unsigned short *, int, const int);
13949 void vec_dstst (const vector signed short *, int, const int);
13950 void vec_dstst (const vector bool short *, int, const int);
13951 void vec_dstst (const vector pixel *, int, const int);
13952 void vec_dstst (const vector unsigned int *, int, const int);
13953 void vec_dstst (const vector signed int *, int, const int);
13954 void vec_dstst (const vector bool int *, int, const int);
13955 void vec_dstst (const vector float *, int, const int);
13956 void vec_dstst (const unsigned char *, int, const int);
13957 void vec_dstst (const signed char *, int, const int);
13958 void vec_dstst (const unsigned short *, int, const int);
13959 void vec_dstst (const short *, int, const int);
13960 void vec_dstst (const unsigned int *, int, const int);
13961 void vec_dstst (const int *, int, const int);
13962 void vec_dstst (const unsigned long *, int, const int);
13963 void vec_dstst (const long *, int, const int);
13964 void vec_dstst (const float *, int, const int);
13966 void vec_dststt (const vector unsigned char *, int, const int);
13967 void vec_dststt (const vector signed char *, int, const int);
13968 void vec_dststt (const vector bool char *, int, const int);
13969 void vec_dststt (const vector unsigned short *, int, const int);
13970 void vec_dststt (const vector signed short *, int, const int);
13971 void vec_dststt (const vector bool short *, int, const int);
13972 void vec_dststt (const vector pixel *, int, const int);
13973 void vec_dststt (const vector unsigned int *, int, const int);
13974 void vec_dststt (const vector signed int *, int, const int);
13975 void vec_dststt (const vector bool int *, int, const int);
13976 void vec_dststt (const vector float *, int, const int);
13977 void vec_dststt (const unsigned char *, int, const int);
13978 void vec_dststt (const signed char *, int, const int);
13979 void vec_dststt (const unsigned short *, int, const int);
13980 void vec_dststt (const short *, int, const int);
13981 void vec_dststt (const unsigned int *, int, const int);
13982 void vec_dststt (const int *, int, const int);
13983 void vec_dststt (const unsigned long *, int, const int);
13984 void vec_dststt (const long *, int, const int);
13985 void vec_dststt (const float *, int, const int);
13987 void vec_dstt (const vector unsigned char *, int, const int);
13988 void vec_dstt (const vector signed char *, int, const int);
13989 void vec_dstt (const vector bool char *, int, const int);
13990 void vec_dstt (const vector unsigned short *, int, const int);
13991 void vec_dstt (const vector signed short *, int, const int);
13992 void vec_dstt (const vector bool short *, int, const int);
13993 void vec_dstt (const vector pixel *, int, const int);
13994 void vec_dstt (const vector unsigned int *, int, const int);
13995 void vec_dstt (const vector signed int *, int, const int);
13996 void vec_dstt (const vector bool int *, int, const int);
13997 void vec_dstt (const vector float *, int, const int);
13998 void vec_dstt (const unsigned char *, int, const int);
13999 void vec_dstt (const signed char *, int, const int);
14000 void vec_dstt (const unsigned short *, int, const int);
14001 void vec_dstt (const short *, int, const int);
14002 void vec_dstt (const unsigned int *, int, const int);
14003 void vec_dstt (const int *, int, const int);
14004 void vec_dstt (const unsigned long *, int, const int);
14005 void vec_dstt (const long *, int, const int);
14006 void vec_dstt (const float *, int, const int);
14008 vector float vec_expte (vector float);
14010 vector float vec_floor (vector float);
14012 vector float vec_ld (int, const vector float *);
14013 vector float vec_ld (int, const float *);
14014 vector bool int vec_ld (int, const vector bool int *);
14015 vector signed int vec_ld (int, const vector signed int *);
14016 vector signed int vec_ld (int, const int *);
14017 vector signed int vec_ld (int, const long *);
14018 vector unsigned int vec_ld (int, const vector unsigned int *);
14019 vector unsigned int vec_ld (int, const unsigned int *);
14020 vector unsigned int vec_ld (int, const unsigned long *);
14021 vector bool short vec_ld (int, const vector bool short *);
14022 vector pixel vec_ld (int, const vector pixel *);
14023 vector signed short vec_ld (int, const vector signed short *);
14024 vector signed short vec_ld (int, const short *);
14025 vector unsigned short vec_ld (int, const vector unsigned short *);
14026 vector unsigned short vec_ld (int, const unsigned short *);
14027 vector bool char vec_ld (int, const vector bool char *);
14028 vector signed char vec_ld (int, const vector signed char *);
14029 vector signed char vec_ld (int, const signed char *);
14030 vector unsigned char vec_ld (int, const vector unsigned char *);
14031 vector unsigned char vec_ld (int, const unsigned char *);
14033 vector signed char vec_lde (int, const signed char *);
14034 vector unsigned char vec_lde (int, const unsigned char *);
14035 vector signed short vec_lde (int, const short *);
14036 vector unsigned short vec_lde (int, const unsigned short *);
14037 vector float vec_lde (int, const float *);
14038 vector signed int vec_lde (int, const int *);
14039 vector unsigned int vec_lde (int, const unsigned int *);
14040 vector signed int vec_lde (int, const long *);
14041 vector unsigned int vec_lde (int, const unsigned long *);
14043 vector float vec_lvewx (int, float *);
14044 vector signed int vec_lvewx (int, int *);
14045 vector unsigned int vec_lvewx (int, unsigned int *);
14046 vector signed int vec_lvewx (int, long *);
14047 vector unsigned int vec_lvewx (int, unsigned long *);
14049 vector signed short vec_lvehx (int, short *);
14050 vector unsigned short vec_lvehx (int, unsigned short *);
14052 vector signed char vec_lvebx (int, char *);
14053 vector unsigned char vec_lvebx (int, unsigned char *);
14055 vector float vec_ldl (int, const vector float *);
14056 vector float vec_ldl (int, const float *);
14057 vector bool int vec_ldl (int, const vector bool int *);
14058 vector signed int vec_ldl (int, const vector signed int *);
14059 vector signed int vec_ldl (int, const int *);
14060 vector signed int vec_ldl (int, const long *);
14061 vector unsigned int vec_ldl (int, const vector unsigned int *);
14062 vector unsigned int vec_ldl (int, const unsigned int *);
14063 vector unsigned int vec_ldl (int, const unsigned long *);
14064 vector bool short vec_ldl (int, const vector bool short *);
14065 vector pixel vec_ldl (int, const vector pixel *);
14066 vector signed short vec_ldl (int, const vector signed short *);
14067 vector signed short vec_ldl (int, const short *);
14068 vector unsigned short vec_ldl (int, const vector unsigned short *);
14069 vector unsigned short vec_ldl (int, const unsigned short *);
14070 vector bool char vec_ldl (int, const vector bool char *);
14071 vector signed char vec_ldl (int, const vector signed char *);
14072 vector signed char vec_ldl (int, const signed char *);
14073 vector unsigned char vec_ldl (int, const vector unsigned char *);
14074 vector unsigned char vec_ldl (int, const unsigned char *);
14076 vector float vec_loge (vector float);
14078 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
14079 vector unsigned char vec_lvsl (int, const volatile signed char *);
14080 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
14081 vector unsigned char vec_lvsl (int, const volatile short *);
14082 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
14083 vector unsigned char vec_lvsl (int, const volatile int *);
14084 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
14085 vector unsigned char vec_lvsl (int, const volatile long *);
14086 vector unsigned char vec_lvsl (int, const volatile float *);
14088 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
14089 vector unsigned char vec_lvsr (int, const volatile signed char *);
14090 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
14091 vector unsigned char vec_lvsr (int, const volatile short *);
14092 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
14093 vector unsigned char vec_lvsr (int, const volatile int *);
14094 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
14095 vector unsigned char vec_lvsr (int, const volatile long *);
14096 vector unsigned char vec_lvsr (int, const volatile float *);
14098 vector float vec_madd (vector float, vector float, vector float);
14100 vector signed short vec_madds (vector signed short,
14101 vector signed short,
14102 vector signed short);
14104 vector unsigned char vec_max (vector bool char, vector unsigned char);
14105 vector unsigned char vec_max (vector unsigned char, vector bool char);
14106 vector unsigned char vec_max (vector unsigned char,
14107 vector unsigned char);
14108 vector signed char vec_max (vector bool char, vector signed char);
14109 vector signed char vec_max (vector signed char, vector bool char);
14110 vector signed char vec_max (vector signed char, vector signed char);
14111 vector unsigned short vec_max (vector bool short,
14112 vector unsigned short);
14113 vector unsigned short vec_max (vector unsigned short,
14114 vector bool short);
14115 vector unsigned short vec_max (vector unsigned short,
14116 vector unsigned short);
14117 vector signed short vec_max (vector bool short, vector signed short);
14118 vector signed short vec_max (vector signed short, vector bool short);
14119 vector signed short vec_max (vector signed short, vector signed short);
14120 vector unsigned int vec_max (vector bool int, vector unsigned int);
14121 vector unsigned int vec_max (vector unsigned int, vector bool int);
14122 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
14123 vector signed int vec_max (vector bool int, vector signed int);
14124 vector signed int vec_max (vector signed int, vector bool int);
14125 vector signed int vec_max (vector signed int, vector signed int);
14126 vector float vec_max (vector float, vector float);
14128 vector float vec_vmaxfp (vector float, vector float);
14130 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14131 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14132 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14134 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14135 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14136 vector unsigned int vec_vmaxuw (vector unsigned int,
14137 vector unsigned int);
14139 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14140 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14141 vector signed short vec_vmaxsh (vector signed short,
14142 vector signed short);
14144 vector unsigned short vec_vmaxuh (vector bool short,
14145 vector unsigned short);
14146 vector unsigned short vec_vmaxuh (vector unsigned short,
14147 vector bool short);
14148 vector unsigned short vec_vmaxuh (vector unsigned short,
14149 vector unsigned short);
14151 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14152 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14153 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14155 vector unsigned char vec_vmaxub (vector bool char,
14156 vector unsigned char);
14157 vector unsigned char vec_vmaxub (vector unsigned char,
14159 vector unsigned char vec_vmaxub (vector unsigned char,
14160 vector unsigned char);
14162 vector bool char vec_mergeh (vector bool char, vector bool char);
14163 vector signed char vec_mergeh (vector signed char, vector signed char);
14164 vector unsigned char vec_mergeh (vector unsigned char,
14165 vector unsigned char);
14166 vector bool short vec_mergeh (vector bool short, vector bool short);
14167 vector pixel vec_mergeh (vector pixel, vector pixel);
14168 vector signed short vec_mergeh (vector signed short,
14169 vector signed short);
14170 vector unsigned short vec_mergeh (vector unsigned short,
14171 vector unsigned short);
14172 vector float vec_mergeh (vector float, vector float);
14173 vector bool int vec_mergeh (vector bool int, vector bool int);
14174 vector signed int vec_mergeh (vector signed int, vector signed int);
14175 vector unsigned int vec_mergeh (vector unsigned int,
14176 vector unsigned int);
14178 vector float vec_vmrghw (vector float, vector float);
14179 vector bool int vec_vmrghw (vector bool int, vector bool int);
14180 vector signed int vec_vmrghw (vector signed int, vector signed int);
14181 vector unsigned int vec_vmrghw (vector unsigned int,
14182 vector unsigned int);
14184 vector bool short vec_vmrghh (vector bool short, vector bool short);
14185 vector signed short vec_vmrghh (vector signed short,
14186 vector signed short);
14187 vector unsigned short vec_vmrghh (vector unsigned short,
14188 vector unsigned short);
14189 vector pixel vec_vmrghh (vector pixel, vector pixel);
14191 vector bool char vec_vmrghb (vector bool char, vector bool char);
14192 vector signed char vec_vmrghb (vector signed char, vector signed char);
14193 vector unsigned char vec_vmrghb (vector unsigned char,
14194 vector unsigned char);
14196 vector bool char vec_mergel (vector bool char, vector bool char);
14197 vector signed char vec_mergel (vector signed char, vector signed char);
14198 vector unsigned char vec_mergel (vector unsigned char,
14199 vector unsigned char);
14200 vector bool short vec_mergel (vector bool short, vector bool short);
14201 vector pixel vec_mergel (vector pixel, vector pixel);
14202 vector signed short vec_mergel (vector signed short,
14203 vector signed short);
14204 vector unsigned short vec_mergel (vector unsigned short,
14205 vector unsigned short);
14206 vector float vec_mergel (vector float, vector float);
14207 vector bool int vec_mergel (vector bool int, vector bool int);
14208 vector signed int vec_mergel (vector signed int, vector signed int);
14209 vector unsigned int vec_mergel (vector unsigned int,
14210 vector unsigned int);
14212 vector float vec_vmrglw (vector float, vector float);
14213 vector signed int vec_vmrglw (vector signed int, vector signed int);
14214 vector unsigned int vec_vmrglw (vector unsigned int,
14215 vector unsigned int);
14216 vector bool int vec_vmrglw (vector bool int, vector bool int);
14218 vector bool short vec_vmrglh (vector bool short, vector bool short);
14219 vector signed short vec_vmrglh (vector signed short,
14220 vector signed short);
14221 vector unsigned short vec_vmrglh (vector unsigned short,
14222 vector unsigned short);
14223 vector pixel vec_vmrglh (vector pixel, vector pixel);
14225 vector bool char vec_vmrglb (vector bool char, vector bool char);
14226 vector signed char vec_vmrglb (vector signed char, vector signed char);
14227 vector unsigned char vec_vmrglb (vector unsigned char,
14228 vector unsigned char);
14230 vector unsigned short vec_mfvscr (void);
14232 vector unsigned char vec_min (vector bool char, vector unsigned char);
14233 vector unsigned char vec_min (vector unsigned char, vector bool char);
14234 vector unsigned char vec_min (vector unsigned char,
14235 vector unsigned char);
14236 vector signed char vec_min (vector bool char, vector signed char);
14237 vector signed char vec_min (vector signed char, vector bool char);
14238 vector signed char vec_min (vector signed char, vector signed char);
14239 vector unsigned short vec_min (vector bool short,
14240 vector unsigned short);
14241 vector unsigned short vec_min (vector unsigned short,
14242 vector bool short);
14243 vector unsigned short vec_min (vector unsigned short,
14244 vector unsigned short);
14245 vector signed short vec_min (vector bool short, vector signed short);
14246 vector signed short vec_min (vector signed short, vector bool short);
14247 vector signed short vec_min (vector signed short, vector signed short);
14248 vector unsigned int vec_min (vector bool int, vector unsigned int);
14249 vector unsigned int vec_min (vector unsigned int, vector bool int);
14250 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14251 vector signed int vec_min (vector bool int, vector signed int);
14252 vector signed int vec_min (vector signed int, vector bool int);
14253 vector signed int vec_min (vector signed int, vector signed int);
14254 vector float vec_min (vector float, vector float);
14256 vector float vec_vminfp (vector float, vector float);
14258 vector signed int vec_vminsw (vector bool int, vector signed int);
14259 vector signed int vec_vminsw (vector signed int, vector bool int);
14260 vector signed int vec_vminsw (vector signed int, vector signed int);
14262 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14263 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14264 vector unsigned int vec_vminuw (vector unsigned int,
14265 vector unsigned int);
14267 vector signed short vec_vminsh (vector bool short, vector signed short);
14268 vector signed short vec_vminsh (vector signed short, vector bool short);
14269 vector signed short vec_vminsh (vector signed short,
14270 vector signed short);
14272 vector unsigned short vec_vminuh (vector bool short,
14273 vector unsigned short);
14274 vector unsigned short vec_vminuh (vector unsigned short,
14275 vector bool short);
14276 vector unsigned short vec_vminuh (vector unsigned short,
14277 vector unsigned short);
14279 vector signed char vec_vminsb (vector bool char, vector signed char);
14280 vector signed char vec_vminsb (vector signed char, vector bool char);
14281 vector signed char vec_vminsb (vector signed char, vector signed char);
14283 vector unsigned char vec_vminub (vector bool char,
14284 vector unsigned char);
14285 vector unsigned char vec_vminub (vector unsigned char,
14287 vector unsigned char vec_vminub (vector unsigned char,
14288 vector unsigned char);
14290 vector signed short vec_mladd (vector signed short,
14291 vector signed short,
14292 vector signed short);
14293 vector signed short vec_mladd (vector signed short,
14294 vector unsigned short,
14295 vector unsigned short);
14296 vector signed short vec_mladd (vector unsigned short,
14297 vector signed short,
14298 vector signed short);
14299 vector unsigned short vec_mladd (vector unsigned short,
14300 vector unsigned short,
14301 vector unsigned short);
14303 vector signed short vec_mradds (vector signed short,
14304 vector signed short,
14305 vector signed short);
14307 vector unsigned int vec_msum (vector unsigned char,
14308 vector unsigned char,
14309 vector unsigned int);
14310 vector signed int vec_msum (vector signed char,
14311 vector unsigned char,
14312 vector signed int);
14313 vector unsigned int vec_msum (vector unsigned short,
14314 vector unsigned short,
14315 vector unsigned int);
14316 vector signed int vec_msum (vector signed short,
14317 vector signed short,
14318 vector signed int);
14320 vector signed int vec_vmsumshm (vector signed short,
14321 vector signed short,
14322 vector signed int);
14324 vector unsigned int vec_vmsumuhm (vector unsigned short,
14325 vector unsigned short,
14326 vector unsigned int);
14328 vector signed int vec_vmsummbm (vector signed char,
14329 vector unsigned char,
14330 vector signed int);
14332 vector unsigned int vec_vmsumubm (vector unsigned char,
14333 vector unsigned char,
14334 vector unsigned int);
14336 vector unsigned int vec_msums (vector unsigned short,
14337 vector unsigned short,
14338 vector unsigned int);
14339 vector signed int vec_msums (vector signed short,
14340 vector signed short,
14341 vector signed int);
14343 vector signed int vec_vmsumshs (vector signed short,
14344 vector signed short,
14345 vector signed int);
14347 vector unsigned int vec_vmsumuhs (vector unsigned short,
14348 vector unsigned short,
14349 vector unsigned int);
14351 void vec_mtvscr (vector signed int);
14352 void vec_mtvscr (vector unsigned int);
14353 void vec_mtvscr (vector bool int);
14354 void vec_mtvscr (vector signed short);
14355 void vec_mtvscr (vector unsigned short);
14356 void vec_mtvscr (vector bool short);
14357 void vec_mtvscr (vector pixel);
14358 void vec_mtvscr (vector signed char);
14359 void vec_mtvscr (vector unsigned char);
14360 void vec_mtvscr (vector bool char);
14362 vector unsigned short vec_mule (vector unsigned char,
14363 vector unsigned char);
14364 vector signed short vec_mule (vector signed char,
14365 vector signed char);
14366 vector unsigned int vec_mule (vector unsigned short,
14367 vector unsigned short);
14368 vector signed int vec_mule (vector signed short, vector signed short);
14370 vector signed int vec_vmulesh (vector signed short,
14371 vector signed short);
14373 vector unsigned int vec_vmuleuh (vector unsigned short,
14374 vector unsigned short);
14376 vector signed short vec_vmulesb (vector signed char,
14377 vector signed char);
14379 vector unsigned short vec_vmuleub (vector unsigned char,
14380 vector unsigned char);
14382 vector unsigned short vec_mulo (vector unsigned char,
14383 vector unsigned char);
14384 vector signed short vec_mulo (vector signed char, vector signed char);
14385 vector unsigned int vec_mulo (vector unsigned short,
14386 vector unsigned short);
14387 vector signed int vec_mulo (vector signed short, vector signed short);
14389 vector signed int vec_vmulosh (vector signed short,
14390 vector signed short);
14392 vector unsigned int vec_vmulouh (vector unsigned short,
14393 vector unsigned short);
14395 vector signed short vec_vmulosb (vector signed char,
14396 vector signed char);
14398 vector unsigned short vec_vmuloub (vector unsigned char,
14399 vector unsigned char);
14401 vector float vec_nmsub (vector float, vector float, vector float);
14403 vector float vec_nor (vector float, vector float);
14404 vector signed int vec_nor (vector signed int, vector signed int);
14405 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14406 vector bool int vec_nor (vector bool int, vector bool int);
14407 vector signed short vec_nor (vector signed short, vector signed short);
14408 vector unsigned short vec_nor (vector unsigned short,
14409 vector unsigned short);
14410 vector bool short vec_nor (vector bool short, vector bool short);
14411 vector signed char vec_nor (vector signed char, vector signed char);
14412 vector unsigned char vec_nor (vector unsigned char,
14413 vector unsigned char);
14414 vector bool char vec_nor (vector bool char, vector bool char);
14416 vector float vec_or (vector float, vector float);
14417 vector float vec_or (vector float, vector bool int);
14418 vector float vec_or (vector bool int, vector float);
14419 vector bool int vec_or (vector bool int, vector bool int);
14420 vector signed int vec_or (vector bool int, vector signed int);
14421 vector signed int vec_or (vector signed int, vector bool int);
14422 vector signed int vec_or (vector signed int, vector signed int);
14423 vector unsigned int vec_or (vector bool int, vector unsigned int);
14424 vector unsigned int vec_or (vector unsigned int, vector bool int);
14425 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14426 vector bool short vec_or (vector bool short, vector bool short);
14427 vector signed short vec_or (vector bool short, vector signed short);
14428 vector signed short vec_or (vector signed short, vector bool short);
14429 vector signed short vec_or (vector signed short, vector signed short);
14430 vector unsigned short vec_or (vector bool short, vector unsigned short);
14431 vector unsigned short vec_or (vector unsigned short, vector bool short);
14432 vector unsigned short vec_or (vector unsigned short,
14433 vector unsigned short);
14434 vector signed char vec_or (vector bool char, vector signed char);
14435 vector bool char vec_or (vector bool char, vector bool char);
14436 vector signed char vec_or (vector signed char, vector bool char);
14437 vector signed char vec_or (vector signed char, vector signed char);
14438 vector unsigned char vec_or (vector bool char, vector unsigned char);
14439 vector unsigned char vec_or (vector unsigned char, vector bool char);
14440 vector unsigned char vec_or (vector unsigned char,
14441 vector unsigned char);
14443 vector signed char vec_pack (vector signed short, vector signed short);
14444 vector unsigned char vec_pack (vector unsigned short,
14445 vector unsigned short);
14446 vector bool char vec_pack (vector bool short, vector bool short);
14447 vector signed short vec_pack (vector signed int, vector signed int);
14448 vector unsigned short vec_pack (vector unsigned int,
14449 vector unsigned int);
14450 vector bool short vec_pack (vector bool int, vector bool int);
14452 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14453 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14454 vector unsigned short vec_vpkuwum (vector unsigned int,
14455 vector unsigned int);
14457 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14458 vector signed char vec_vpkuhum (vector signed short,
14459 vector signed short);
14460 vector unsigned char vec_vpkuhum (vector unsigned short,
14461 vector unsigned short);
14463 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14465 vector unsigned char vec_packs (vector unsigned short,
14466 vector unsigned short);
14467 vector signed char vec_packs (vector signed short, vector signed short);
14468 vector unsigned short vec_packs (vector unsigned int,
14469 vector unsigned int);
14470 vector signed short vec_packs (vector signed int, vector signed int);
14472 vector signed short vec_vpkswss (vector signed int, vector signed int);
14474 vector unsigned short vec_vpkuwus (vector unsigned int,
14475 vector unsigned int);
14477 vector signed char vec_vpkshss (vector signed short,
14478 vector signed short);
14480 vector unsigned char vec_vpkuhus (vector unsigned short,
14481 vector unsigned short);
14483 vector unsigned char vec_packsu (vector unsigned short,
14484 vector unsigned short);
14485 vector unsigned char vec_packsu (vector signed short,
14486 vector signed short);
14487 vector unsigned short vec_packsu (vector unsigned int,
14488 vector unsigned int);
14489 vector unsigned short vec_packsu (vector signed int, vector signed int);
14491 vector unsigned short vec_vpkswus (vector signed int,
14492 vector signed int);
14494 vector unsigned char vec_vpkshus (vector signed short,
14495 vector signed short);
14497 vector float vec_perm (vector float,
14499 vector unsigned char);
14500 vector signed int vec_perm (vector signed int,
14502 vector unsigned char);
14503 vector unsigned int vec_perm (vector unsigned int,
14504 vector unsigned int,
14505 vector unsigned char);
14506 vector bool int vec_perm (vector bool int,
14508 vector unsigned char);
14509 vector signed short vec_perm (vector signed short,
14510 vector signed short,
14511 vector unsigned char);
14512 vector unsigned short vec_perm (vector unsigned short,
14513 vector unsigned short,
14514 vector unsigned char);
14515 vector bool short vec_perm (vector bool short,
14517 vector unsigned char);
14518 vector pixel vec_perm (vector pixel,
14520 vector unsigned char);
14521 vector signed char vec_perm (vector signed char,
14522 vector signed char,
14523 vector unsigned char);
14524 vector unsigned char vec_perm (vector unsigned char,
14525 vector unsigned char,
14526 vector unsigned char);
14527 vector bool char vec_perm (vector bool char,
14529 vector unsigned char);
14531 vector float vec_re (vector float);
14533 vector signed char vec_rl (vector signed char,
14534 vector unsigned char);
14535 vector unsigned char vec_rl (vector unsigned char,
14536 vector unsigned char);
14537 vector signed short vec_rl (vector signed short, vector unsigned short);
14538 vector unsigned short vec_rl (vector unsigned short,
14539 vector unsigned short);
14540 vector signed int vec_rl (vector signed int, vector unsigned int);
14541 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14543 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14544 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14546 vector signed short vec_vrlh (vector signed short,
14547 vector unsigned short);
14548 vector unsigned short vec_vrlh (vector unsigned short,
14549 vector unsigned short);
14551 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14552 vector unsigned char vec_vrlb (vector unsigned char,
14553 vector unsigned char);
14555 vector float vec_round (vector float);
14557 vector float vec_recip (vector float, vector float);
14559 vector float vec_rsqrt (vector float);
14561 vector float vec_rsqrte (vector float);
14563 vector float vec_sel (vector float, vector float, vector bool int);
14564 vector float vec_sel (vector float, vector float, vector unsigned int);
14565 vector signed int vec_sel (vector signed int,
14568 vector signed int vec_sel (vector signed int,
14570 vector unsigned int);
14571 vector unsigned int vec_sel (vector unsigned int,
14572 vector unsigned int,
14574 vector unsigned int vec_sel (vector unsigned int,
14575 vector unsigned int,
14576 vector unsigned int);
14577 vector bool int vec_sel (vector bool int,
14580 vector bool int vec_sel (vector bool int,
14582 vector unsigned int);
14583 vector signed short vec_sel (vector signed short,
14584 vector signed short,
14585 vector bool short);
14586 vector signed short vec_sel (vector signed short,
14587 vector signed short,
14588 vector unsigned short);
14589 vector unsigned short vec_sel (vector unsigned short,
14590 vector unsigned short,
14591 vector bool short);
14592 vector unsigned short vec_sel (vector unsigned short,
14593 vector unsigned short,
14594 vector unsigned short);
14595 vector bool short vec_sel (vector bool short,
14597 vector bool short);
14598 vector bool short vec_sel (vector bool short,
14600 vector unsigned short);
14601 vector signed char vec_sel (vector signed char,
14602 vector signed char,
14604 vector signed char vec_sel (vector signed char,
14605 vector signed char,
14606 vector unsigned char);
14607 vector unsigned char vec_sel (vector unsigned char,
14608 vector unsigned char,
14610 vector unsigned char vec_sel (vector unsigned char,
14611 vector unsigned char,
14612 vector unsigned char);
14613 vector bool char vec_sel (vector bool char,
14616 vector bool char vec_sel (vector bool char,
14618 vector unsigned char);
14620 vector signed char vec_sl (vector signed char,
14621 vector unsigned char);
14622 vector unsigned char vec_sl (vector unsigned char,
14623 vector unsigned char);
14624 vector signed short vec_sl (vector signed short, vector unsigned short);
14625 vector unsigned short vec_sl (vector unsigned short,
14626 vector unsigned short);
14627 vector signed int vec_sl (vector signed int, vector unsigned int);
14628 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14630 vector signed int vec_vslw (vector signed int, vector unsigned int);
14631 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14633 vector signed short vec_vslh (vector signed short,
14634 vector unsigned short);
14635 vector unsigned short vec_vslh (vector unsigned short,
14636 vector unsigned short);
14638 vector signed char vec_vslb (vector signed char, vector unsigned char);
14639 vector unsigned char vec_vslb (vector unsigned char,
14640 vector unsigned char);
14642 vector float vec_sld (vector float, vector float, const int);
14643 vector signed int vec_sld (vector signed int,
14646 vector unsigned int vec_sld (vector unsigned int,
14647 vector unsigned int,
14649 vector bool int vec_sld (vector bool int,
14652 vector signed short vec_sld (vector signed short,
14653 vector signed short,
14655 vector unsigned short vec_sld (vector unsigned short,
14656 vector unsigned short,
14658 vector bool short vec_sld (vector bool short,
14661 vector pixel vec_sld (vector pixel,
14664 vector signed char vec_sld (vector signed char,
14665 vector signed char,
14667 vector unsigned char vec_sld (vector unsigned char,
14668 vector unsigned char,
14670 vector bool char vec_sld (vector bool char,
14674 vector signed int vec_sll (vector signed int,
14675 vector unsigned int);
14676 vector signed int vec_sll (vector signed int,
14677 vector unsigned short);
14678 vector signed int vec_sll (vector signed int,
14679 vector unsigned char);
14680 vector unsigned int vec_sll (vector unsigned int,
14681 vector unsigned int);
14682 vector unsigned int vec_sll (vector unsigned int,
14683 vector unsigned short);
14684 vector unsigned int vec_sll (vector unsigned int,
14685 vector unsigned char);
14686 vector bool int vec_sll (vector bool int,
14687 vector unsigned int);
14688 vector bool int vec_sll (vector bool int,
14689 vector unsigned short);
14690 vector bool int vec_sll (vector bool int,
14691 vector unsigned char);
14692 vector signed short vec_sll (vector signed short,
14693 vector unsigned int);
14694 vector signed short vec_sll (vector signed short,
14695 vector unsigned short);
14696 vector signed short vec_sll (vector signed short,
14697 vector unsigned char);
14698 vector unsigned short vec_sll (vector unsigned short,
14699 vector unsigned int);
14700 vector unsigned short vec_sll (vector unsigned short,
14701 vector unsigned short);
14702 vector unsigned short vec_sll (vector unsigned short,
14703 vector unsigned char);
14704 vector bool short vec_sll (vector bool short, vector unsigned int);
14705 vector bool short vec_sll (vector bool short, vector unsigned short);
14706 vector bool short vec_sll (vector bool short, vector unsigned char);
14707 vector pixel vec_sll (vector pixel, vector unsigned int);
14708 vector pixel vec_sll (vector pixel, vector unsigned short);
14709 vector pixel vec_sll (vector pixel, vector unsigned char);
14710 vector signed char vec_sll (vector signed char, vector unsigned int);
14711 vector signed char vec_sll (vector signed char, vector unsigned short);
14712 vector signed char vec_sll (vector signed char, vector unsigned char);
14713 vector unsigned char vec_sll (vector unsigned char,
14714 vector unsigned int);
14715 vector unsigned char vec_sll (vector unsigned char,
14716 vector unsigned short);
14717 vector unsigned char vec_sll (vector unsigned char,
14718 vector unsigned char);
14719 vector bool char vec_sll (vector bool char, vector unsigned int);
14720 vector bool char vec_sll (vector bool char, vector unsigned short);
14721 vector bool char vec_sll (vector bool char, vector unsigned char);
14723 vector float vec_slo (vector float, vector signed char);
14724 vector float vec_slo (vector float, vector unsigned char);
14725 vector signed int vec_slo (vector signed int, vector signed char);
14726 vector signed int vec_slo (vector signed int, vector unsigned char);
14727 vector unsigned int vec_slo (vector unsigned int, vector signed char);
14728 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
14729 vector signed short vec_slo (vector signed short, vector signed char);
14730 vector signed short vec_slo (vector signed short, vector unsigned char);
14731 vector unsigned short vec_slo (vector unsigned short,
14732 vector signed char);
14733 vector unsigned short vec_slo (vector unsigned short,
14734 vector unsigned char);
14735 vector pixel vec_slo (vector pixel, vector signed char);
14736 vector pixel vec_slo (vector pixel, vector unsigned char);
14737 vector signed char vec_slo (vector signed char, vector signed char);
14738 vector signed char vec_slo (vector signed char, vector unsigned char);
14739 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14740 vector unsigned char vec_slo (vector unsigned char,
14741 vector unsigned char);
14743 vector signed char vec_splat (vector signed char, const int);
14744 vector unsigned char vec_splat (vector unsigned char, const int);
14745 vector bool char vec_splat (vector bool char, const int);
14746 vector signed short vec_splat (vector signed short, const int);
14747 vector unsigned short vec_splat (vector unsigned short, const int);
14748 vector bool short vec_splat (vector bool short, const int);
14749 vector pixel vec_splat (vector pixel, const int);
14750 vector float vec_splat (vector float, const int);
14751 vector signed int vec_splat (vector signed int, const int);
14752 vector unsigned int vec_splat (vector unsigned int, const int);
14753 vector bool int vec_splat (vector bool int, const int);
14754 vector signed long vec_splat (vector signed long, const int);
14755 vector unsigned long vec_splat (vector unsigned long, const int);
14757 vector signed char vec_splats (signed char);
14758 vector unsigned char vec_splats (unsigned char);
14759 vector signed short vec_splats (signed short);
14760 vector unsigned short vec_splats (unsigned short);
14761 vector signed int vec_splats (signed int);
14762 vector unsigned int vec_splats (unsigned int);
14763 vector float vec_splats (float);
14765 vector float vec_vspltw (vector float, const int);
14766 vector signed int vec_vspltw (vector signed int, const int);
14767 vector unsigned int vec_vspltw (vector unsigned int, const int);
14768 vector bool int vec_vspltw (vector bool int, const int);
14770 vector bool short vec_vsplth (vector bool short, const int);
14771 vector signed short vec_vsplth (vector signed short, const int);
14772 vector unsigned short vec_vsplth (vector unsigned short, const int);
14773 vector pixel vec_vsplth (vector pixel, const int);
14775 vector signed char vec_vspltb (vector signed char, const int);
14776 vector unsigned char vec_vspltb (vector unsigned char, const int);
14777 vector bool char vec_vspltb (vector bool char, const int);
14779 vector signed char vec_splat_s8 (const int);
14781 vector signed short vec_splat_s16 (const int);
14783 vector signed int vec_splat_s32 (const int);
14785 vector unsigned char vec_splat_u8 (const int);
14787 vector unsigned short vec_splat_u16 (const int);
14789 vector unsigned int vec_splat_u32 (const int);
14791 vector signed char vec_sr (vector signed char, vector unsigned char);
14792 vector unsigned char vec_sr (vector unsigned char,
14793 vector unsigned char);
14794 vector signed short vec_sr (vector signed short,
14795 vector unsigned short);
14796 vector unsigned short vec_sr (vector unsigned short,
14797 vector unsigned short);
14798 vector signed int vec_sr (vector signed int, vector unsigned int);
14799 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
14801 vector signed int vec_vsrw (vector signed int, vector unsigned int);
14802 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
14804 vector signed short vec_vsrh (vector signed short,
14805 vector unsigned short);
14806 vector unsigned short vec_vsrh (vector unsigned short,
14807 vector unsigned short);
14809 vector signed char vec_vsrb (vector signed char, vector unsigned char);
14810 vector unsigned char vec_vsrb (vector unsigned char,
14811 vector unsigned char);
14813 vector signed char vec_sra (vector signed char, vector unsigned char);
14814 vector unsigned char vec_sra (vector unsigned char,
14815 vector unsigned char);
14816 vector signed short vec_sra (vector signed short,
14817 vector unsigned short);
14818 vector unsigned short vec_sra (vector unsigned short,
14819 vector unsigned short);
14820 vector signed int vec_sra (vector signed int, vector unsigned int);
14821 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
14823 vector signed int vec_vsraw (vector signed int, vector unsigned int);
14824 vector unsigned int vec_vsraw (vector unsigned int,
14825 vector unsigned int);
14827 vector signed short vec_vsrah (vector signed short,
14828 vector unsigned short);
14829 vector unsigned short vec_vsrah (vector unsigned short,
14830 vector unsigned short);
14832 vector signed char vec_vsrab (vector signed char, vector unsigned char);
14833 vector unsigned char vec_vsrab (vector unsigned char,
14834 vector unsigned char);
14836 vector signed int vec_srl (vector signed int, vector unsigned int);
14837 vector signed int vec_srl (vector signed int, vector unsigned short);
14838 vector signed int vec_srl (vector signed int, vector unsigned char);
14839 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
14840 vector unsigned int vec_srl (vector unsigned int,
14841 vector unsigned short);
14842 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
14843 vector bool int vec_srl (vector bool int, vector unsigned int);
14844 vector bool int vec_srl (vector bool int, vector unsigned short);
14845 vector bool int vec_srl (vector bool int, vector unsigned char);
14846 vector signed short vec_srl (vector signed short, vector unsigned int);
14847 vector signed short vec_srl (vector signed short,
14848 vector unsigned short);
14849 vector signed short vec_srl (vector signed short, vector unsigned char);
14850 vector unsigned short vec_srl (vector unsigned short,
14851 vector unsigned int);
14852 vector unsigned short vec_srl (vector unsigned short,
14853 vector unsigned short);
14854 vector unsigned short vec_srl (vector unsigned short,
14855 vector unsigned char);
14856 vector bool short vec_srl (vector bool short, vector unsigned int);
14857 vector bool short vec_srl (vector bool short, vector unsigned short);
14858 vector bool short vec_srl (vector bool short, vector unsigned char);
14859 vector pixel vec_srl (vector pixel, vector unsigned int);
14860 vector pixel vec_srl (vector pixel, vector unsigned short);
14861 vector pixel vec_srl (vector pixel, vector unsigned char);
14862 vector signed char vec_srl (vector signed char, vector unsigned int);
14863 vector signed char vec_srl (vector signed char, vector unsigned short);
14864 vector signed char vec_srl (vector signed char, vector unsigned char);
14865 vector unsigned char vec_srl (vector unsigned char,
14866 vector unsigned int);
14867 vector unsigned char vec_srl (vector unsigned char,
14868 vector unsigned short);
14869 vector unsigned char vec_srl (vector unsigned char,
14870 vector unsigned char);
14871 vector bool char vec_srl (vector bool char, vector unsigned int);
14872 vector bool char vec_srl (vector bool char, vector unsigned short);
14873 vector bool char vec_srl (vector bool char, vector unsigned char);
14875 vector float vec_sro (vector float, vector signed char);
14876 vector float vec_sro (vector float, vector unsigned char);
14877 vector signed int vec_sro (vector signed int, vector signed char);
14878 vector signed int vec_sro (vector signed int, vector unsigned char);
14879 vector unsigned int vec_sro (vector unsigned int, vector signed char);
14880 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
14881 vector signed short vec_sro (vector signed short, vector signed char);
14882 vector signed short vec_sro (vector signed short, vector unsigned char);
14883 vector unsigned short vec_sro (vector unsigned short,
14884 vector signed char);
14885 vector unsigned short vec_sro (vector unsigned short,
14886 vector unsigned char);
14887 vector pixel vec_sro (vector pixel, vector signed char);
14888 vector pixel vec_sro (vector pixel, vector unsigned char);
14889 vector signed char vec_sro (vector signed char, vector signed char);
14890 vector signed char vec_sro (vector signed char, vector unsigned char);
14891 vector unsigned char vec_sro (vector unsigned char, vector signed char);
14892 vector unsigned char vec_sro (vector unsigned char,
14893 vector unsigned char);
14895 void vec_st (vector float, int, vector float *);
14896 void vec_st (vector float, int, float *);
14897 void vec_st (vector signed int, int, vector signed int *);
14898 void vec_st (vector signed int, int, int *);
14899 void vec_st (vector unsigned int, int, vector unsigned int *);
14900 void vec_st (vector unsigned int, int, unsigned int *);
14901 void vec_st (vector bool int, int, vector bool int *);
14902 void vec_st (vector bool int, int, unsigned int *);
14903 void vec_st (vector bool int, int, int *);
14904 void vec_st (vector signed short, int, vector signed short *);
14905 void vec_st (vector signed short, int, short *);
14906 void vec_st (vector unsigned short, int, vector unsigned short *);
14907 void vec_st (vector unsigned short, int, unsigned short *);
14908 void vec_st (vector bool short, int, vector bool short *);
14909 void vec_st (vector bool short, int, unsigned short *);
14910 void vec_st (vector pixel, int, vector pixel *);
14911 void vec_st (vector pixel, int, unsigned short *);
14912 void vec_st (vector pixel, int, short *);
14913 void vec_st (vector bool short, int, short *);
14914 void vec_st (vector signed char, int, vector signed char *);
14915 void vec_st (vector signed char, int, signed char *);
14916 void vec_st (vector unsigned char, int, vector unsigned char *);
14917 void vec_st (vector unsigned char, int, unsigned char *);
14918 void vec_st (vector bool char, int, vector bool char *);
14919 void vec_st (vector bool char, int, unsigned char *);
14920 void vec_st (vector bool char, int, signed char *);
14922 void vec_ste (vector signed char, int, signed char *);
14923 void vec_ste (vector unsigned char, int, unsigned char *);
14924 void vec_ste (vector bool char, int, signed char *);
14925 void vec_ste (vector bool char, int, unsigned char *);
14926 void vec_ste (vector signed short, int, short *);
14927 void vec_ste (vector unsigned short, int, unsigned short *);
14928 void vec_ste (vector bool short, int, short *);
14929 void vec_ste (vector bool short, int, unsigned short *);
14930 void vec_ste (vector pixel, int, short *);
14931 void vec_ste (vector pixel, int, unsigned short *);
14932 void vec_ste (vector float, int, float *);
14933 void vec_ste (vector signed int, int, int *);
14934 void vec_ste (vector unsigned int, int, unsigned int *);
14935 void vec_ste (vector bool int, int, int *);
14936 void vec_ste (vector bool int, int, unsigned int *);
14938 void vec_stvewx (vector float, int, float *);
14939 void vec_stvewx (vector signed int, int, int *);
14940 void vec_stvewx (vector unsigned int, int, unsigned int *);
14941 void vec_stvewx (vector bool int, int, int *);
14942 void vec_stvewx (vector bool int, int, unsigned int *);
14944 void vec_stvehx (vector signed short, int, short *);
14945 void vec_stvehx (vector unsigned short, int, unsigned short *);
14946 void vec_stvehx (vector bool short, int, short *);
14947 void vec_stvehx (vector bool short, int, unsigned short *);
14948 void vec_stvehx (vector pixel, int, short *);
14949 void vec_stvehx (vector pixel, int, unsigned short *);
14951 void vec_stvebx (vector signed char, int, signed char *);
14952 void vec_stvebx (vector unsigned char, int, unsigned char *);
14953 void vec_stvebx (vector bool char, int, signed char *);
14954 void vec_stvebx (vector bool char, int, unsigned char *);
14956 void vec_stl (vector float, int, vector float *);
14957 void vec_stl (vector float, int, float *);
14958 void vec_stl (vector signed int, int, vector signed int *);
14959 void vec_stl (vector signed int, int, int *);
14960 void vec_stl (vector unsigned int, int, vector unsigned int *);
14961 void vec_stl (vector unsigned int, int, unsigned int *);
14962 void vec_stl (vector bool int, int, vector bool int *);
14963 void vec_stl (vector bool int, int, unsigned int *);
14964 void vec_stl (vector bool int, int, int *);
14965 void vec_stl (vector signed short, int, vector signed short *);
14966 void vec_stl (vector signed short, int, short *);
14967 void vec_stl (vector unsigned short, int, vector unsigned short *);
14968 void vec_stl (vector unsigned short, int, unsigned short *);
14969 void vec_stl (vector bool short, int, vector bool short *);
14970 void vec_stl (vector bool short, int, unsigned short *);
14971 void vec_stl (vector bool short, int, short *);
14972 void vec_stl (vector pixel, int, vector pixel *);
14973 void vec_stl (vector pixel, int, unsigned short *);
14974 void vec_stl (vector pixel, int, short *);
14975 void vec_stl (vector signed char, int, vector signed char *);
14976 void vec_stl (vector signed char, int, signed char *);
14977 void vec_stl (vector unsigned char, int, vector unsigned char *);
14978 void vec_stl (vector unsigned char, int, unsigned char *);
14979 void vec_stl (vector bool char, int, vector bool char *);
14980 void vec_stl (vector bool char, int, unsigned char *);
14981 void vec_stl (vector bool char, int, signed char *);
14983 vector signed char vec_sub (vector bool char, vector signed char);
14984 vector signed char vec_sub (vector signed char, vector bool char);
14985 vector signed char vec_sub (vector signed char, vector signed char);
14986 vector unsigned char vec_sub (vector bool char, vector unsigned char);
14987 vector unsigned char vec_sub (vector unsigned char, vector bool char);
14988 vector unsigned char vec_sub (vector unsigned char,
14989 vector unsigned char);
14990 vector signed short vec_sub (vector bool short, vector signed short);
14991 vector signed short vec_sub (vector signed short, vector bool short);
14992 vector signed short vec_sub (vector signed short, vector signed short);
14993 vector unsigned short vec_sub (vector bool short,
14994 vector unsigned short);
14995 vector unsigned short vec_sub (vector unsigned short,
14996 vector bool short);
14997 vector unsigned short vec_sub (vector unsigned short,
14998 vector unsigned short);
14999 vector signed int vec_sub (vector bool int, vector signed int);
15000 vector signed int vec_sub (vector signed int, vector bool int);
15001 vector signed int vec_sub (vector signed int, vector signed int);
15002 vector unsigned int vec_sub (vector bool int, vector unsigned int);
15003 vector unsigned int vec_sub (vector unsigned int, vector bool int);
15004 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
15005 vector float vec_sub (vector float, vector float);
15007 vector float vec_vsubfp (vector float, vector float);
15009 vector signed int vec_vsubuwm (vector bool int, vector signed int);
15010 vector signed int vec_vsubuwm (vector signed int, vector bool int);
15011 vector signed int vec_vsubuwm (vector signed int, vector signed int);
15012 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
15013 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
15014 vector unsigned int vec_vsubuwm (vector unsigned int,
15015 vector unsigned int);
15017 vector signed short vec_vsubuhm (vector bool short,
15018 vector signed short);
15019 vector signed short vec_vsubuhm (vector signed short,
15020 vector bool short);
15021 vector signed short vec_vsubuhm (vector signed short,
15022 vector signed short);
15023 vector unsigned short vec_vsubuhm (vector bool short,
15024 vector unsigned short);
15025 vector unsigned short vec_vsubuhm (vector unsigned short,
15026 vector bool short);
15027 vector unsigned short vec_vsubuhm (vector unsigned short,
15028 vector unsigned short);
15030 vector signed char vec_vsububm (vector bool char, vector signed char);
15031 vector signed char vec_vsububm (vector signed char, vector bool char);
15032 vector signed char vec_vsububm (vector signed char, vector signed char);
15033 vector unsigned char vec_vsububm (vector bool char,
15034 vector unsigned char);
15035 vector unsigned char vec_vsububm (vector unsigned char,
15037 vector unsigned char vec_vsububm (vector unsigned char,
15038 vector unsigned char);
15040 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
15042 vector unsigned char vec_subs (vector bool char, vector unsigned char);
15043 vector unsigned char vec_subs (vector unsigned char, vector bool char);
15044 vector unsigned char vec_subs (vector unsigned char,
15045 vector unsigned char);
15046 vector signed char vec_subs (vector bool char, vector signed char);
15047 vector signed char vec_subs (vector signed char, vector bool char);
15048 vector signed char vec_subs (vector signed char, vector signed char);
15049 vector unsigned short vec_subs (vector bool short,
15050 vector unsigned short);
15051 vector unsigned short vec_subs (vector unsigned short,
15052 vector bool short);
15053 vector unsigned short vec_subs (vector unsigned short,
15054 vector unsigned short);
15055 vector signed short vec_subs (vector bool short, vector signed short);
15056 vector signed short vec_subs (vector signed short, vector bool short);
15057 vector signed short vec_subs (vector signed short, vector signed short);
15058 vector unsigned int vec_subs (vector bool int, vector unsigned int);
15059 vector unsigned int vec_subs (vector unsigned int, vector bool int);
15060 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
15061 vector signed int vec_subs (vector bool int, vector signed int);
15062 vector signed int vec_subs (vector signed int, vector bool int);
15063 vector signed int vec_subs (vector signed int, vector signed int);
15065 vector signed int vec_vsubsws (vector bool int, vector signed int);
15066 vector signed int vec_vsubsws (vector signed int, vector bool int);
15067 vector signed int vec_vsubsws (vector signed int, vector signed int);
15069 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
15070 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
15071 vector unsigned int vec_vsubuws (vector unsigned int,
15072 vector unsigned int);
15074 vector signed short vec_vsubshs (vector bool short,
15075 vector signed short);
15076 vector signed short vec_vsubshs (vector signed short,
15077 vector bool short);
15078 vector signed short vec_vsubshs (vector signed short,
15079 vector signed short);
15081 vector unsigned short vec_vsubuhs (vector bool short,
15082 vector unsigned short);
15083 vector unsigned short vec_vsubuhs (vector unsigned short,
15084 vector bool short);
15085 vector unsigned short vec_vsubuhs (vector unsigned short,
15086 vector unsigned short);
15088 vector signed char vec_vsubsbs (vector bool char, vector signed char);
15089 vector signed char vec_vsubsbs (vector signed char, vector bool char);
15090 vector signed char vec_vsubsbs (vector signed char, vector signed char);
15092 vector unsigned char vec_vsububs (vector bool char,
15093 vector unsigned char);
15094 vector unsigned char vec_vsububs (vector unsigned char,
15096 vector unsigned char vec_vsububs (vector unsigned char,
15097 vector unsigned char);
15099 vector unsigned int vec_sum4s (vector unsigned char,
15100 vector unsigned int);
15101 vector signed int vec_sum4s (vector signed char, vector signed int);
15102 vector signed int vec_sum4s (vector signed short, vector signed int);
15104 vector signed int vec_vsum4shs (vector signed short, vector signed int);
15106 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
15108 vector unsigned int vec_vsum4ubs (vector unsigned char,
15109 vector unsigned int);
15111 vector signed int vec_sum2s (vector signed int, vector signed int);
15113 vector signed int vec_sums (vector signed int, vector signed int);
15115 vector float vec_trunc (vector float);
15117 vector signed short vec_unpackh (vector signed char);
15118 vector bool short vec_unpackh (vector bool char);
15119 vector signed int vec_unpackh (vector signed short);
15120 vector bool int vec_unpackh (vector bool short);
15121 vector unsigned int vec_unpackh (vector pixel);
15123 vector bool int vec_vupkhsh (vector bool short);
15124 vector signed int vec_vupkhsh (vector signed short);
15126 vector unsigned int vec_vupkhpx (vector pixel);
15128 vector bool short vec_vupkhsb (vector bool char);
15129 vector signed short vec_vupkhsb (vector signed char);
15131 vector signed short vec_unpackl (vector signed char);
15132 vector bool short vec_unpackl (vector bool char);
15133 vector unsigned int vec_unpackl (vector pixel);
15134 vector signed int vec_unpackl (vector signed short);
15135 vector bool int vec_unpackl (vector bool short);
15137 vector unsigned int vec_vupklpx (vector pixel);
15139 vector bool int vec_vupklsh (vector bool short);
15140 vector signed int vec_vupklsh (vector signed short);
15142 vector bool short vec_vupklsb (vector bool char);
15143 vector signed short vec_vupklsb (vector signed char);
15145 vector float vec_xor (vector float, vector float);
15146 vector float vec_xor (vector float, vector bool int);
15147 vector float vec_xor (vector bool int, vector float);
15148 vector bool int vec_xor (vector bool int, vector bool int);
15149 vector signed int vec_xor (vector bool int, vector signed int);
15150 vector signed int vec_xor (vector signed int, vector bool int);
15151 vector signed int vec_xor (vector signed int, vector signed int);
15152 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15153 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15154 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15155 vector bool short vec_xor (vector bool short, vector bool short);
15156 vector signed short vec_xor (vector bool short, vector signed short);
15157 vector signed short vec_xor (vector signed short, vector bool short);
15158 vector signed short vec_xor (vector signed short, vector signed short);
15159 vector unsigned short vec_xor (vector bool short,
15160 vector unsigned short);
15161 vector unsigned short vec_xor (vector unsigned short,
15162 vector bool short);
15163 vector unsigned short vec_xor (vector unsigned short,
15164 vector unsigned short);
15165 vector signed char vec_xor (vector bool char, vector signed char);
15166 vector bool char vec_xor (vector bool char, vector bool char);
15167 vector signed char vec_xor (vector signed char, vector bool char);
15168 vector signed char vec_xor (vector signed char, vector signed char);
15169 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15170 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15171 vector unsigned char vec_xor (vector unsigned char,
15172 vector unsigned char);
15174 int vec_all_eq (vector signed char, vector bool char);
15175 int vec_all_eq (vector signed char, vector signed char);
15176 int vec_all_eq (vector unsigned char, vector bool char);
15177 int vec_all_eq (vector unsigned char, vector unsigned char);
15178 int vec_all_eq (vector bool char, vector bool char);
15179 int vec_all_eq (vector bool char, vector unsigned char);
15180 int vec_all_eq (vector bool char, vector signed char);
15181 int vec_all_eq (vector signed short, vector bool short);
15182 int vec_all_eq (vector signed short, vector signed short);
15183 int vec_all_eq (vector unsigned short, vector bool short);
15184 int vec_all_eq (vector unsigned short, vector unsigned short);
15185 int vec_all_eq (vector bool short, vector bool short);
15186 int vec_all_eq (vector bool short, vector unsigned short);
15187 int vec_all_eq (vector bool short, vector signed short);
15188 int vec_all_eq (vector pixel, vector pixel);
15189 int vec_all_eq (vector signed int, vector bool int);
15190 int vec_all_eq (vector signed int, vector signed int);
15191 int vec_all_eq (vector unsigned int, vector bool int);
15192 int vec_all_eq (vector unsigned int, vector unsigned int);
15193 int vec_all_eq (vector bool int, vector bool int);
15194 int vec_all_eq (vector bool int, vector unsigned int);
15195 int vec_all_eq (vector bool int, vector signed int);
15196 int vec_all_eq (vector float, vector float);
15198 int vec_all_ge (vector bool char, vector unsigned char);
15199 int vec_all_ge (vector unsigned char, vector bool char);
15200 int vec_all_ge (vector unsigned char, vector unsigned char);
15201 int vec_all_ge (vector bool char, vector signed char);
15202 int vec_all_ge (vector signed char, vector bool char);
15203 int vec_all_ge (vector signed char, vector signed char);
15204 int vec_all_ge (vector bool short, vector unsigned short);
15205 int vec_all_ge (vector unsigned short, vector bool short);
15206 int vec_all_ge (vector unsigned short, vector unsigned short);
15207 int vec_all_ge (vector signed short, vector signed short);
15208 int vec_all_ge (vector bool short, vector signed short);
15209 int vec_all_ge (vector signed short, vector bool short);
15210 int vec_all_ge (vector bool int, vector unsigned int);
15211 int vec_all_ge (vector unsigned int, vector bool int);
15212 int vec_all_ge (vector unsigned int, vector unsigned int);
15213 int vec_all_ge (vector bool int, vector signed int);
15214 int vec_all_ge (vector signed int, vector bool int);
15215 int vec_all_ge (vector signed int, vector signed int);
15216 int vec_all_ge (vector float, vector float);
15218 int vec_all_gt (vector bool char, vector unsigned char);
15219 int vec_all_gt (vector unsigned char, vector bool char);
15220 int vec_all_gt (vector unsigned char, vector unsigned char);
15221 int vec_all_gt (vector bool char, vector signed char);
15222 int vec_all_gt (vector signed char, vector bool char);
15223 int vec_all_gt (vector signed char, vector signed char);
15224 int vec_all_gt (vector bool short, vector unsigned short);
15225 int vec_all_gt (vector unsigned short, vector bool short);
15226 int vec_all_gt (vector unsigned short, vector unsigned short);
15227 int vec_all_gt (vector bool short, vector signed short);
15228 int vec_all_gt (vector signed short, vector bool short);
15229 int vec_all_gt (vector signed short, vector signed short);
15230 int vec_all_gt (vector bool int, vector unsigned int);
15231 int vec_all_gt (vector unsigned int, vector bool int);
15232 int vec_all_gt (vector unsigned int, vector unsigned int);
15233 int vec_all_gt (vector bool int, vector signed int);
15234 int vec_all_gt (vector signed int, vector bool int);
15235 int vec_all_gt (vector signed int, vector signed int);
15236 int vec_all_gt (vector float, vector float);
15238 int vec_all_in (vector float, vector float);
15240 int vec_all_le (vector bool char, vector unsigned char);
15241 int vec_all_le (vector unsigned char, vector bool char);
15242 int vec_all_le (vector unsigned char, vector unsigned char);
15243 int vec_all_le (vector bool char, vector signed char);
15244 int vec_all_le (vector signed char, vector bool char);
15245 int vec_all_le (vector signed char, vector signed char);
15246 int vec_all_le (vector bool short, vector unsigned short);
15247 int vec_all_le (vector unsigned short, vector bool short);
15248 int vec_all_le (vector unsigned short, vector unsigned short);
15249 int vec_all_le (vector bool short, vector signed short);
15250 int vec_all_le (vector signed short, vector bool short);
15251 int vec_all_le (vector signed short, vector signed short);
15252 int vec_all_le (vector bool int, vector unsigned int);
15253 int vec_all_le (vector unsigned int, vector bool int);
15254 int vec_all_le (vector unsigned int, vector unsigned int);
15255 int vec_all_le (vector bool int, vector signed int);
15256 int vec_all_le (vector signed int, vector bool int);
15257 int vec_all_le (vector signed int, vector signed int);
15258 int vec_all_le (vector float, vector float);
15260 int vec_all_lt (vector bool char, vector unsigned char);
15261 int vec_all_lt (vector unsigned char, vector bool char);
15262 int vec_all_lt (vector unsigned char, vector unsigned char);
15263 int vec_all_lt (vector bool char, vector signed char);
15264 int vec_all_lt (vector signed char, vector bool char);
15265 int vec_all_lt (vector signed char, vector signed char);
15266 int vec_all_lt (vector bool short, vector unsigned short);
15267 int vec_all_lt (vector unsigned short, vector bool short);
15268 int vec_all_lt (vector unsigned short, vector unsigned short);
15269 int vec_all_lt (vector bool short, vector signed short);
15270 int vec_all_lt (vector signed short, vector bool short);
15271 int vec_all_lt (vector signed short, vector signed short);
15272 int vec_all_lt (vector bool int, vector unsigned int);
15273 int vec_all_lt (vector unsigned int, vector bool int);
15274 int vec_all_lt (vector unsigned int, vector unsigned int);
15275 int vec_all_lt (vector bool int, vector signed int);
15276 int vec_all_lt (vector signed int, vector bool int);
15277 int vec_all_lt (vector signed int, vector signed int);
15278 int vec_all_lt (vector float, vector float);
15280 int vec_all_nan (vector float);
15282 int vec_all_ne (vector signed char, vector bool char);
15283 int vec_all_ne (vector signed char, vector signed char);
15284 int vec_all_ne (vector unsigned char, vector bool char);
15285 int vec_all_ne (vector unsigned char, vector unsigned char);
15286 int vec_all_ne (vector bool char, vector bool char);
15287 int vec_all_ne (vector bool char, vector unsigned char);
15288 int vec_all_ne (vector bool char, vector signed char);
15289 int vec_all_ne (vector signed short, vector bool short);
15290 int vec_all_ne (vector signed short, vector signed short);
15291 int vec_all_ne (vector unsigned short, vector bool short);
15292 int vec_all_ne (vector unsigned short, vector unsigned short);
15293 int vec_all_ne (vector bool short, vector bool short);
15294 int vec_all_ne (vector bool short, vector unsigned short);
15295 int vec_all_ne (vector bool short, vector signed short);
15296 int vec_all_ne (vector pixel, vector pixel);
15297 int vec_all_ne (vector signed int, vector bool int);
15298 int vec_all_ne (vector signed int, vector signed int);
15299 int vec_all_ne (vector unsigned int, vector bool int);
15300 int vec_all_ne (vector unsigned int, vector unsigned int);
15301 int vec_all_ne (vector bool int, vector bool int);
15302 int vec_all_ne (vector bool int, vector unsigned int);
15303 int vec_all_ne (vector bool int, vector signed int);
15304 int vec_all_ne (vector float, vector float);
15306 int vec_all_nge (vector float, vector float);
15308 int vec_all_ngt (vector float, vector float);
15310 int vec_all_nle (vector float, vector float);
15312 int vec_all_nlt (vector float, vector float);
15314 int vec_all_numeric (vector float);
15316 int vec_any_eq (vector signed char, vector bool char);
15317 int vec_any_eq (vector signed char, vector signed char);
15318 int vec_any_eq (vector unsigned char, vector bool char);
15319 int vec_any_eq (vector unsigned char, vector unsigned char);
15320 int vec_any_eq (vector bool char, vector bool char);
15321 int vec_any_eq (vector bool char, vector unsigned char);
15322 int vec_any_eq (vector bool char, vector signed char);
15323 int vec_any_eq (vector signed short, vector bool short);
15324 int vec_any_eq (vector signed short, vector signed short);
15325 int vec_any_eq (vector unsigned short, vector bool short);
15326 int vec_any_eq (vector unsigned short, vector unsigned short);
15327 int vec_any_eq (vector bool short, vector bool short);
15328 int vec_any_eq (vector bool short, vector unsigned short);
15329 int vec_any_eq (vector bool short, vector signed short);
15330 int vec_any_eq (vector pixel, vector pixel);
15331 int vec_any_eq (vector signed int, vector bool int);
15332 int vec_any_eq (vector signed int, vector signed int);
15333 int vec_any_eq (vector unsigned int, vector bool int);
15334 int vec_any_eq (vector unsigned int, vector unsigned int);
15335 int vec_any_eq (vector bool int, vector bool int);
15336 int vec_any_eq (vector bool int, vector unsigned int);
15337 int vec_any_eq (vector bool int, vector signed int);
15338 int vec_any_eq (vector float, vector float);
15340 int vec_any_ge (vector signed char, vector bool char);
15341 int vec_any_ge (vector unsigned char, vector bool char);
15342 int vec_any_ge (vector unsigned char, vector unsigned char);
15343 int vec_any_ge (vector signed char, vector signed char);
15344 int vec_any_ge (vector bool char, vector unsigned char);
15345 int vec_any_ge (vector bool char, vector signed char);
15346 int vec_any_ge (vector unsigned short, vector bool short);
15347 int vec_any_ge (vector unsigned short, vector unsigned short);
15348 int vec_any_ge (vector signed short, vector signed short);
15349 int vec_any_ge (vector signed short, vector bool short);
15350 int vec_any_ge (vector bool short, vector unsigned short);
15351 int vec_any_ge (vector bool short, vector signed short);
15352 int vec_any_ge (vector signed int, vector bool int);
15353 int vec_any_ge (vector unsigned int, vector bool int);
15354 int vec_any_ge (vector unsigned int, vector unsigned int);
15355 int vec_any_ge (vector signed int, vector signed int);
15356 int vec_any_ge (vector bool int, vector unsigned int);
15357 int vec_any_ge (vector bool int, vector signed int);
15358 int vec_any_ge (vector float, vector float);
15360 int vec_any_gt (vector bool char, vector unsigned char);
15361 int vec_any_gt (vector unsigned char, vector bool char);
15362 int vec_any_gt (vector unsigned char, vector unsigned char);
15363 int vec_any_gt (vector bool char, vector signed char);
15364 int vec_any_gt (vector signed char, vector bool char);
15365 int vec_any_gt (vector signed char, vector signed char);
15366 int vec_any_gt (vector bool short, vector unsigned short);
15367 int vec_any_gt (vector unsigned short, vector bool short);
15368 int vec_any_gt (vector unsigned short, vector unsigned short);
15369 int vec_any_gt (vector bool short, vector signed short);
15370 int vec_any_gt (vector signed short, vector bool short);
15371 int vec_any_gt (vector signed short, vector signed short);
15372 int vec_any_gt (vector bool int, vector unsigned int);
15373 int vec_any_gt (vector unsigned int, vector bool int);
15374 int vec_any_gt (vector unsigned int, vector unsigned int);
15375 int vec_any_gt (vector bool int, vector signed int);
15376 int vec_any_gt (vector signed int, vector bool int);
15377 int vec_any_gt (vector signed int, vector signed int);
15378 int vec_any_gt (vector float, vector float);
15380 int vec_any_le (vector bool char, vector unsigned char);
15381 int vec_any_le (vector unsigned char, vector bool char);
15382 int vec_any_le (vector unsigned char, vector unsigned char);
15383 int vec_any_le (vector bool char, vector signed char);
15384 int vec_any_le (vector signed char, vector bool char);
15385 int vec_any_le (vector signed char, vector signed char);
15386 int vec_any_le (vector bool short, vector unsigned short);
15387 int vec_any_le (vector unsigned short, vector bool short);
15388 int vec_any_le (vector unsigned short, vector unsigned short);
15389 int vec_any_le (vector bool short, vector signed short);
15390 int vec_any_le (vector signed short, vector bool short);
15391 int vec_any_le (vector signed short, vector signed short);
15392 int vec_any_le (vector bool int, vector unsigned int);
15393 int vec_any_le (vector unsigned int, vector bool int);
15394 int vec_any_le (vector unsigned int, vector unsigned int);
15395 int vec_any_le (vector bool int, vector signed int);
15396 int vec_any_le (vector signed int, vector bool int);
15397 int vec_any_le (vector signed int, vector signed int);
15398 int vec_any_le (vector float, vector float);
15400 int vec_any_lt (vector bool char, vector unsigned char);
15401 int vec_any_lt (vector unsigned char, vector bool char);
15402 int vec_any_lt (vector unsigned char, vector unsigned char);
15403 int vec_any_lt (vector bool char, vector signed char);
15404 int vec_any_lt (vector signed char, vector bool char);
15405 int vec_any_lt (vector signed char, vector signed char);
15406 int vec_any_lt (vector bool short, vector unsigned short);
15407 int vec_any_lt (vector unsigned short, vector bool short);
15408 int vec_any_lt (vector unsigned short, vector unsigned short);
15409 int vec_any_lt (vector bool short, vector signed short);
15410 int vec_any_lt (vector signed short, vector bool short);
15411 int vec_any_lt (vector signed short, vector signed short);
15412 int vec_any_lt (vector bool int, vector unsigned int);
15413 int vec_any_lt (vector unsigned int, vector bool int);
15414 int vec_any_lt (vector unsigned int, vector unsigned int);
15415 int vec_any_lt (vector bool int, vector signed int);
15416 int vec_any_lt (vector signed int, vector bool int);
15417 int vec_any_lt (vector signed int, vector signed int);
15418 int vec_any_lt (vector float, vector float);
15420 int vec_any_nan (vector float);
15422 int vec_any_ne (vector signed char, vector bool char);
15423 int vec_any_ne (vector signed char, vector signed char);
15424 int vec_any_ne (vector unsigned char, vector bool char);
15425 int vec_any_ne (vector unsigned char, vector unsigned char);
15426 int vec_any_ne (vector bool char, vector bool char);
15427 int vec_any_ne (vector bool char, vector unsigned char);
15428 int vec_any_ne (vector bool char, vector signed char);
15429 int vec_any_ne (vector signed short, vector bool short);
15430 int vec_any_ne (vector signed short, vector signed short);
15431 int vec_any_ne (vector unsigned short, vector bool short);
15432 int vec_any_ne (vector unsigned short, vector unsigned short);
15433 int vec_any_ne (vector bool short, vector bool short);
15434 int vec_any_ne (vector bool short, vector unsigned short);
15435 int vec_any_ne (vector bool short, vector signed short);
15436 int vec_any_ne (vector pixel, vector pixel);
15437 int vec_any_ne (vector signed int, vector bool int);
15438 int vec_any_ne (vector signed int, vector signed int);
15439 int vec_any_ne (vector unsigned int, vector bool int);
15440 int vec_any_ne (vector unsigned int, vector unsigned int);
15441 int vec_any_ne (vector bool int, vector bool int);
15442 int vec_any_ne (vector bool int, vector unsigned int);
15443 int vec_any_ne (vector bool int, vector signed int);
15444 int vec_any_ne (vector float, vector float);
15446 int vec_any_nge (vector float, vector float);
15448 int vec_any_ngt (vector float, vector float);
15450 int vec_any_nle (vector float, vector float);
15452 int vec_any_nlt (vector float, vector float);
15454 int vec_any_numeric (vector float);
15456 int vec_any_out (vector float, vector float);
15459 If the vector/scalar (VSX) instruction set is available, the following
15460 additional functions are available:
15463 vector double vec_abs (vector double);
15464 vector double vec_add (vector double, vector double);
15465 vector double vec_and (vector double, vector double);
15466 vector double vec_and (vector double, vector bool long);
15467 vector double vec_and (vector bool long, vector double);
15468 vector long vec_and (vector long, vector long);
15469 vector long vec_and (vector long, vector bool long);
15470 vector long vec_and (vector bool long, vector long);
15471 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15472 vector unsigned long vec_and (vector unsigned long, vector bool long);
15473 vector unsigned long vec_and (vector bool long, vector unsigned long);
15474 vector double vec_andc (vector double, vector double);
15475 vector double vec_andc (vector double, vector bool long);
15476 vector double vec_andc (vector bool long, vector double);
15477 vector long vec_andc (vector long, vector long);
15478 vector long vec_andc (vector long, vector bool long);
15479 vector long vec_andc (vector bool long, vector long);
15480 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15481 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15482 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15483 vector double vec_ceil (vector double);
15484 vector bool long vec_cmpeq (vector double, vector double);
15485 vector bool long vec_cmpge (vector double, vector double);
15486 vector bool long vec_cmpgt (vector double, vector double);
15487 vector bool long vec_cmple (vector double, vector double);
15488 vector bool long vec_cmplt (vector double, vector double);
15489 vector double vec_cpsgn (vector double, vector double);
15490 vector float vec_div (vector float, vector float);
15491 vector double vec_div (vector double, vector double);
15492 vector long vec_div (vector long, vector long);
15493 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15494 vector double vec_floor (vector double);
15495 vector double vec_ld (int, const vector double *);
15496 vector double vec_ld (int, const double *);
15497 vector double vec_ldl (int, const vector double *);
15498 vector double vec_ldl (int, const double *);
15499 vector unsigned char vec_lvsl (int, const volatile double *);
15500 vector unsigned char vec_lvsr (int, const volatile double *);
15501 vector double vec_madd (vector double, vector double, vector double);
15502 vector double vec_max (vector double, vector double);
15503 vector signed long vec_mergeh (vector signed long, vector signed long);
15504 vector signed long vec_mergeh (vector signed long, vector bool long);
15505 vector signed long vec_mergeh (vector bool long, vector signed long);
15506 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15507 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15508 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15509 vector signed long vec_mergel (vector signed long, vector signed long);
15510 vector signed long vec_mergel (vector signed long, vector bool long);
15511 vector signed long vec_mergel (vector bool long, vector signed long);
15512 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15513 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15514 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15515 vector double vec_min (vector double, vector double);
15516 vector float vec_msub (vector float, vector float, vector float);
15517 vector double vec_msub (vector double, vector double, vector double);
15518 vector float vec_mul (vector float, vector float);
15519 vector double vec_mul (vector double, vector double);
15520 vector long vec_mul (vector long, vector long);
15521 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15522 vector float vec_nearbyint (vector float);
15523 vector double vec_nearbyint (vector double);
15524 vector float vec_nmadd (vector float, vector float, vector float);
15525 vector double vec_nmadd (vector double, vector double, vector double);
15526 vector double vec_nmsub (vector double, vector double, vector double);
15527 vector double vec_nor (vector double, vector double);
15528 vector long vec_nor (vector long, vector long);
15529 vector long vec_nor (vector long, vector bool long);
15530 vector long vec_nor (vector bool long, vector long);
15531 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15532 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15533 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15534 vector double vec_or (vector double, vector double);
15535 vector double vec_or (vector double, vector bool long);
15536 vector double vec_or (vector bool long, vector double);
15537 vector long vec_or (vector long, vector long);
15538 vector long vec_or (vector long, vector bool long);
15539 vector long vec_or (vector bool long, vector long);
15540 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15541 vector unsigned long vec_or (vector unsigned long, vector bool long);
15542 vector unsigned long vec_or (vector bool long, vector unsigned long);
15543 vector double vec_perm (vector double, vector double, vector unsigned char);
15544 vector long vec_perm (vector long, vector long, vector unsigned char);
15545 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15546 vector unsigned char);
15547 vector double vec_rint (vector double);
15548 vector double vec_recip (vector double, vector double);
15549 vector double vec_rsqrt (vector double);
15550 vector double vec_rsqrte (vector double);
15551 vector double vec_sel (vector double, vector double, vector bool long);
15552 vector double vec_sel (vector double, vector double, vector unsigned long);
15553 vector long vec_sel (vector long, vector long, vector long);
15554 vector long vec_sel (vector long, vector long, vector unsigned long);
15555 vector long vec_sel (vector long, vector long, vector bool long);
15556 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15558 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15559 vector unsigned long);
15560 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15562 vector double vec_splats (double);
15563 vector signed long vec_splats (signed long);
15564 vector unsigned long vec_splats (unsigned long);
15565 vector float vec_sqrt (vector float);
15566 vector double vec_sqrt (vector double);
15567 void vec_st (vector double, int, vector double *);
15568 void vec_st (vector double, int, double *);
15569 vector double vec_sub (vector double, vector double);
15570 vector double vec_trunc (vector double);
15571 vector double vec_xor (vector double, vector double);
15572 vector double vec_xor (vector double, vector bool long);
15573 vector double vec_xor (vector bool long, vector double);
15574 vector long vec_xor (vector long, vector long);
15575 vector long vec_xor (vector long, vector bool long);
15576 vector long vec_xor (vector bool long, vector long);
15577 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15578 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15579 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15580 int vec_all_eq (vector double, vector double);
15581 int vec_all_ge (vector double, vector double);
15582 int vec_all_gt (vector double, vector double);
15583 int vec_all_le (vector double, vector double);
15584 int vec_all_lt (vector double, vector double);
15585 int vec_all_nan (vector double);
15586 int vec_all_ne (vector double, vector double);
15587 int vec_all_nge (vector double, vector double);
15588 int vec_all_ngt (vector double, vector double);
15589 int vec_all_nle (vector double, vector double);
15590 int vec_all_nlt (vector double, vector double);
15591 int vec_all_numeric (vector double);
15592 int vec_any_eq (vector double, vector double);
15593 int vec_any_ge (vector double, vector double);
15594 int vec_any_gt (vector double, vector double);
15595 int vec_any_le (vector double, vector double);
15596 int vec_any_lt (vector double, vector double);
15597 int vec_any_nan (vector double);
15598 int vec_any_ne (vector double, vector double);
15599 int vec_any_nge (vector double, vector double);
15600 int vec_any_ngt (vector double, vector double);
15601 int vec_any_nle (vector double, vector double);
15602 int vec_any_nlt (vector double, vector double);
15603 int vec_any_numeric (vector double);
15605 vector double vec_vsx_ld (int, const vector double *);
15606 vector double vec_vsx_ld (int, const double *);
15607 vector float vec_vsx_ld (int, const vector float *);
15608 vector float vec_vsx_ld (int, const float *);
15609 vector bool int vec_vsx_ld (int, const vector bool int *);
15610 vector signed int vec_vsx_ld (int, const vector signed int *);
15611 vector signed int vec_vsx_ld (int, const int *);
15612 vector signed int vec_vsx_ld (int, const long *);
15613 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15614 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15615 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15616 vector bool short vec_vsx_ld (int, const vector bool short *);
15617 vector pixel vec_vsx_ld (int, const vector pixel *);
15618 vector signed short vec_vsx_ld (int, const vector signed short *);
15619 vector signed short vec_vsx_ld (int, const short *);
15620 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15621 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15622 vector bool char vec_vsx_ld (int, const vector bool char *);
15623 vector signed char vec_vsx_ld (int, const vector signed char *);
15624 vector signed char vec_vsx_ld (int, const signed char *);
15625 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15626 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15628 void vec_vsx_st (vector double, int, vector double *);
15629 void vec_vsx_st (vector double, int, double *);
15630 void vec_vsx_st (vector float, int, vector float *);
15631 void vec_vsx_st (vector float, int, float *);
15632 void vec_vsx_st (vector signed int, int, vector signed int *);
15633 void vec_vsx_st (vector signed int, int, int *);
15634 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15635 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15636 void vec_vsx_st (vector bool int, int, vector bool int *);
15637 void vec_vsx_st (vector bool int, int, unsigned int *);
15638 void vec_vsx_st (vector bool int, int, int *);
15639 void vec_vsx_st (vector signed short, int, vector signed short *);
15640 void vec_vsx_st (vector signed short, int, short *);
15641 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15642 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15643 void vec_vsx_st (vector bool short, int, vector bool short *);
15644 void vec_vsx_st (vector bool short, int, unsigned short *);
15645 void vec_vsx_st (vector pixel, int, vector pixel *);
15646 void vec_vsx_st (vector pixel, int, unsigned short *);
15647 void vec_vsx_st (vector pixel, int, short *);
15648 void vec_vsx_st (vector bool short, int, short *);
15649 void vec_vsx_st (vector signed char, int, vector signed char *);
15650 void vec_vsx_st (vector signed char, int, signed char *);
15651 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15652 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15653 void vec_vsx_st (vector bool char, int, vector bool char *);
15654 void vec_vsx_st (vector bool char, int, unsigned char *);
15655 void vec_vsx_st (vector bool char, int, signed char *);
15657 vector double vec_xxpermdi (vector double, vector double, int);
15658 vector float vec_xxpermdi (vector float, vector float, int);
15659 vector long long vec_xxpermdi (vector long long, vector long long, int);
15660 vector unsigned long long vec_xxpermdi (vector unsigned long long,
15661 vector unsigned long long, int);
15662 vector int vec_xxpermdi (vector int, vector int, int);
15663 vector unsigned int vec_xxpermdi (vector unsigned int,
15664 vector unsigned int, int);
15665 vector short vec_xxpermdi (vector short, vector short, int);
15666 vector unsigned short vec_xxpermdi (vector unsigned short,
15667 vector unsigned short, int);
15668 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
15669 vector unsigned char vec_xxpermdi (vector unsigned char,
15670 vector unsigned char, int);
15672 vector double vec_xxsldi (vector double, vector double, int);
15673 vector float vec_xxsldi (vector float, vector float, int);
15674 vector long long vec_xxsldi (vector long long, vector long long, int);
15675 vector unsigned long long vec_xxsldi (vector unsigned long long,
15676 vector unsigned long long, int);
15677 vector int vec_xxsldi (vector int, vector int, int);
15678 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
15679 vector short vec_xxsldi (vector short, vector short, int);
15680 vector unsigned short vec_xxsldi (vector unsigned short,
15681 vector unsigned short, int);
15682 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
15683 vector unsigned char vec_xxsldi (vector unsigned char,
15684 vector unsigned char, int);
15687 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
15688 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
15689 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
15690 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
15691 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
15693 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15694 instruction set is available, the following additional functions are
15695 available for both 32-bit and 64-bit targets. For 64-bit targets, you
15696 can use @var{vector long} instead of @var{vector long long},
15697 @var{vector bool long} instead of @var{vector bool long long}, and
15698 @var{vector unsigned long} instead of @var{vector unsigned long long}.
15701 vector long long vec_abs (vector long long);
15703 vector long long vec_add (vector long long, vector long long);
15704 vector unsigned long long vec_add (vector unsigned long long,
15705 vector unsigned long long);
15707 int vec_all_eq (vector long long, vector long long);
15708 int vec_all_eq (vector unsigned long long, vector unsigned long long);
15709 int vec_all_ge (vector long long, vector long long);
15710 int vec_all_ge (vector unsigned long long, vector unsigned long long);
15711 int vec_all_gt (vector long long, vector long long);
15712 int vec_all_gt (vector unsigned long long, vector unsigned long long);
15713 int vec_all_le (vector long long, vector long long);
15714 int vec_all_le (vector unsigned long long, vector unsigned long long);
15715 int vec_all_lt (vector long long, vector long long);
15716 int vec_all_lt (vector unsigned long long, vector unsigned long long);
15717 int vec_all_ne (vector long long, vector long long);
15718 int vec_all_ne (vector unsigned long long, vector unsigned long long);
15720 int vec_any_eq (vector long long, vector long long);
15721 int vec_any_eq (vector unsigned long long, vector unsigned long long);
15722 int vec_any_ge (vector long long, vector long long);
15723 int vec_any_ge (vector unsigned long long, vector unsigned long long);
15724 int vec_any_gt (vector long long, vector long long);
15725 int vec_any_gt (vector unsigned long long, vector unsigned long long);
15726 int vec_any_le (vector long long, vector long long);
15727 int vec_any_le (vector unsigned long long, vector unsigned long long);
15728 int vec_any_lt (vector long long, vector long long);
15729 int vec_any_lt (vector unsigned long long, vector unsigned long long);
15730 int vec_any_ne (vector long long, vector long long);
15731 int vec_any_ne (vector unsigned long long, vector unsigned long long);
15733 vector long long vec_eqv (vector long long, vector long long);
15734 vector long long vec_eqv (vector bool long long, vector long long);
15735 vector long long vec_eqv (vector long long, vector bool long long);
15736 vector unsigned long long vec_eqv (vector unsigned long long,
15737 vector unsigned long long);
15738 vector unsigned long long vec_eqv (vector bool long long,
15739 vector unsigned long long);
15740 vector unsigned long long vec_eqv (vector unsigned long long,
15741 vector bool long long);
15742 vector int vec_eqv (vector int, vector int);
15743 vector int vec_eqv (vector bool int, vector int);
15744 vector int vec_eqv (vector int, vector bool int);
15745 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
15746 vector unsigned int vec_eqv (vector bool unsigned int,
15747 vector unsigned int);
15748 vector unsigned int vec_eqv (vector unsigned int,
15749 vector bool unsigned int);
15750 vector short vec_eqv (vector short, vector short);
15751 vector short vec_eqv (vector bool short, vector short);
15752 vector short vec_eqv (vector short, vector bool short);
15753 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
15754 vector unsigned short vec_eqv (vector bool unsigned short,
15755 vector unsigned short);
15756 vector unsigned short vec_eqv (vector unsigned short,
15757 vector bool unsigned short);
15758 vector signed char vec_eqv (vector signed char, vector signed char);
15759 vector signed char vec_eqv (vector bool signed char, vector signed char);
15760 vector signed char vec_eqv (vector signed char, vector bool signed char);
15761 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
15762 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
15763 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
15765 vector long long vec_max (vector long long, vector long long);
15766 vector unsigned long long vec_max (vector unsigned long long,
15767 vector unsigned long long);
15769 vector signed int vec_mergee (vector signed int, vector signed int);
15770 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
15771 vector bool int vec_mergee (vector bool int, vector bool int);
15773 vector signed int vec_mergeo (vector signed int, vector signed int);
15774 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
15775 vector bool int vec_mergeo (vector bool int, vector bool int);
15777 vector long long vec_min (vector long long, vector long long);
15778 vector unsigned long long vec_min (vector unsigned long long,
15779 vector unsigned long long);
15781 vector long long vec_nand (vector long long, vector long long);
15782 vector long long vec_nand (vector bool long long, vector long long);
15783 vector long long vec_nand (vector long long, vector bool long long);
15784 vector unsigned long long vec_nand (vector unsigned long long,
15785 vector unsigned long long);
15786 vector unsigned long long vec_nand (vector bool long long,
15787 vector unsigned long long);
15788 vector unsigned long long vec_nand (vector unsigned long long,
15789 vector bool long long);
15790 vector int vec_nand (vector int, vector int);
15791 vector int vec_nand (vector bool int, vector int);
15792 vector int vec_nand (vector int, vector bool int);
15793 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
15794 vector unsigned int vec_nand (vector bool unsigned int,
15795 vector unsigned int);
15796 vector unsigned int vec_nand (vector unsigned int,
15797 vector bool unsigned int);
15798 vector short vec_nand (vector short, vector short);
15799 vector short vec_nand (vector bool short, vector short);
15800 vector short vec_nand (vector short, vector bool short);
15801 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
15802 vector unsigned short vec_nand (vector bool unsigned short,
15803 vector unsigned short);
15804 vector unsigned short vec_nand (vector unsigned short,
15805 vector bool unsigned short);
15806 vector signed char vec_nand (vector signed char, vector signed char);
15807 vector signed char vec_nand (vector bool signed char, vector signed char);
15808 vector signed char vec_nand (vector signed char, vector bool signed char);
15809 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
15810 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
15811 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
15813 vector long long vec_orc (vector long long, vector long long);
15814 vector long long vec_orc (vector bool long long, vector long long);
15815 vector long long vec_orc (vector long long, vector bool long long);
15816 vector unsigned long long vec_orc (vector unsigned long long,
15817 vector unsigned long long);
15818 vector unsigned long long vec_orc (vector bool long long,
15819 vector unsigned long long);
15820 vector unsigned long long vec_orc (vector unsigned long long,
15821 vector bool long long);
15822 vector int vec_orc (vector int, vector int);
15823 vector int vec_orc (vector bool int, vector int);
15824 vector int vec_orc (vector int, vector bool int);
15825 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
15826 vector unsigned int vec_orc (vector bool unsigned int,
15827 vector unsigned int);
15828 vector unsigned int vec_orc (vector unsigned int,
15829 vector bool unsigned int);
15830 vector short vec_orc (vector short, vector short);
15831 vector short vec_orc (vector bool short, vector short);
15832 vector short vec_orc (vector short, vector bool short);
15833 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
15834 vector unsigned short vec_orc (vector bool unsigned short,
15835 vector unsigned short);
15836 vector unsigned short vec_orc (vector unsigned short,
15837 vector bool unsigned short);
15838 vector signed char vec_orc (vector signed char, vector signed char);
15839 vector signed char vec_orc (vector bool signed char, vector signed char);
15840 vector signed char vec_orc (vector signed char, vector bool signed char);
15841 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
15842 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
15843 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
15845 vector int vec_pack (vector long long, vector long long);
15846 vector unsigned int vec_pack (vector unsigned long long,
15847 vector unsigned long long);
15848 vector bool int vec_pack (vector bool long long, vector bool long long);
15850 vector int vec_packs (vector long long, vector long long);
15851 vector unsigned int vec_packs (vector unsigned long long,
15852 vector unsigned long long);
15854 vector unsigned int vec_packsu (vector long long, vector long long);
15855 vector unsigned int vec_packsu (vector unsigned long long,
15856 vector unsigned long long);
15858 vector long long vec_rl (vector long long,
15859 vector unsigned long long);
15860 vector long long vec_rl (vector unsigned long long,
15861 vector unsigned long long);
15863 vector long long vec_sl (vector long long, vector unsigned long long);
15864 vector long long vec_sl (vector unsigned long long,
15865 vector unsigned long long);
15867 vector long long vec_sr (vector long long, vector unsigned long long);
15868 vector unsigned long long char vec_sr (vector unsigned long long,
15869 vector unsigned long long);
15871 vector long long vec_sra (vector long long, vector unsigned long long);
15872 vector unsigned long long vec_sra (vector unsigned long long,
15873 vector unsigned long long);
15875 vector long long vec_sub (vector long long, vector long long);
15876 vector unsigned long long vec_sub (vector unsigned long long,
15877 vector unsigned long long);
15879 vector long long vec_unpackh (vector int);
15880 vector unsigned long long vec_unpackh (vector unsigned int);
15882 vector long long vec_unpackl (vector int);
15883 vector unsigned long long vec_unpackl (vector unsigned int);
15885 vector long long vec_vaddudm (vector long long, vector long long);
15886 vector long long vec_vaddudm (vector bool long long, vector long long);
15887 vector long long vec_vaddudm (vector long long, vector bool long long);
15888 vector unsigned long long vec_vaddudm (vector unsigned long long,
15889 vector unsigned long long);
15890 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
15891 vector unsigned long long);
15892 vector unsigned long long vec_vaddudm (vector unsigned long long,
15893 vector bool unsigned long long);
15895 vector long long vec_vbpermq (vector signed char, vector signed char);
15896 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
15898 vector long long vec_cntlz (vector long long);
15899 vector unsigned long long vec_cntlz (vector unsigned long long);
15900 vector int vec_cntlz (vector int);
15901 vector unsigned int vec_cntlz (vector int);
15902 vector short vec_cntlz (vector short);
15903 vector unsigned short vec_cntlz (vector unsigned short);
15904 vector signed char vec_cntlz (vector signed char);
15905 vector unsigned char vec_cntlz (vector unsigned char);
15907 vector long long vec_vclz (vector long long);
15908 vector unsigned long long vec_vclz (vector unsigned long long);
15909 vector int vec_vclz (vector int);
15910 vector unsigned int vec_vclz (vector int);
15911 vector short vec_vclz (vector short);
15912 vector unsigned short vec_vclz (vector unsigned short);
15913 vector signed char vec_vclz (vector signed char);
15914 vector unsigned char vec_vclz (vector unsigned char);
15916 vector signed char vec_vclzb (vector signed char);
15917 vector unsigned char vec_vclzb (vector unsigned char);
15919 vector long long vec_vclzd (vector long long);
15920 vector unsigned long long vec_vclzd (vector unsigned long long);
15922 vector short vec_vclzh (vector short);
15923 vector unsigned short vec_vclzh (vector unsigned short);
15925 vector int vec_vclzw (vector int);
15926 vector unsigned int vec_vclzw (vector int);
15928 vector signed char vec_vgbbd (vector signed char);
15929 vector unsigned char vec_vgbbd (vector unsigned char);
15931 vector long long vec_vmaxsd (vector long long, vector long long);
15933 vector unsigned long long vec_vmaxud (vector unsigned long long,
15934 unsigned vector long long);
15936 vector long long vec_vminsd (vector long long, vector long long);
15938 vector unsigned long long vec_vminud (vector long long,
15941 vector int vec_vpksdss (vector long long, vector long long);
15942 vector unsigned int vec_vpksdss (vector long long, vector long long);
15944 vector unsigned int vec_vpkudus (vector unsigned long long,
15945 vector unsigned long long);
15947 vector int vec_vpkudum (vector long long, vector long long);
15948 vector unsigned int vec_vpkudum (vector unsigned long long,
15949 vector unsigned long long);
15950 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
15952 vector long long vec_vpopcnt (vector long long);
15953 vector unsigned long long vec_vpopcnt (vector unsigned long long);
15954 vector int vec_vpopcnt (vector int);
15955 vector unsigned int vec_vpopcnt (vector int);
15956 vector short vec_vpopcnt (vector short);
15957 vector unsigned short vec_vpopcnt (vector unsigned short);
15958 vector signed char vec_vpopcnt (vector signed char);
15959 vector unsigned char vec_vpopcnt (vector unsigned char);
15961 vector signed char vec_vpopcntb (vector signed char);
15962 vector unsigned char vec_vpopcntb (vector unsigned char);
15964 vector long long vec_vpopcntd (vector long long);
15965 vector unsigned long long vec_vpopcntd (vector unsigned long long);
15967 vector short vec_vpopcnth (vector short);
15968 vector unsigned short vec_vpopcnth (vector unsigned short);
15970 vector int vec_vpopcntw (vector int);
15971 vector unsigned int vec_vpopcntw (vector int);
15973 vector long long vec_vrld (vector long long, vector unsigned long long);
15974 vector unsigned long long vec_vrld (vector unsigned long long,
15975 vector unsigned long long);
15977 vector long long vec_vsld (vector long long, vector unsigned long long);
15978 vector long long vec_vsld (vector unsigned long long,
15979 vector unsigned long long);
15981 vector long long vec_vsrad (vector long long, vector unsigned long long);
15982 vector unsigned long long vec_vsrad (vector unsigned long long,
15983 vector unsigned long long);
15985 vector long long vec_vsrd (vector long long, vector unsigned long long);
15986 vector unsigned long long char vec_vsrd (vector unsigned long long,
15987 vector unsigned long long);
15989 vector long long vec_vsubudm (vector long long, vector long long);
15990 vector long long vec_vsubudm (vector bool long long, vector long long);
15991 vector long long vec_vsubudm (vector long long, vector bool long long);
15992 vector unsigned long long vec_vsubudm (vector unsigned long long,
15993 vector unsigned long long);
15994 vector unsigned long long vec_vsubudm (vector bool long long,
15995 vector unsigned long long);
15996 vector unsigned long long vec_vsubudm (vector unsigned long long,
15997 vector bool long long);
15999 vector long long vec_vupkhsw (vector int);
16000 vector unsigned long long vec_vupkhsw (vector unsigned int);
16002 vector long long vec_vupklsw (vector int);
16003 vector unsigned long long vec_vupklsw (vector int);
16006 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16007 instruction set is available, the following additional functions are
16008 available for 64-bit targets. New vector types
16009 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
16010 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
16013 The normal vector extract, and set operations work on
16014 @var{vector __int128_t} and @var{vector __uint128_t} types,
16015 but the index value must be 0.
16018 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
16019 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
16021 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
16022 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
16024 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
16025 vector __int128_t);
16026 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
16027 vector __uint128_t);
16029 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
16030 vector __int128_t);
16031 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
16032 vector __uint128_t);
16034 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
16035 vector __int128_t);
16036 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
16037 vector __uint128_t);
16039 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
16040 vector __int128_t);
16041 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
16042 vector __uint128_t);
16044 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
16045 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
16047 __int128_t vec_vsubuqm (__int128_t, __int128_t);
16048 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
16050 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
16051 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
16052 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
16053 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
16054 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
16055 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
16056 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
16057 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
16058 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
16059 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
16062 If the cryptographic instructions are enabled (@option{-mcrypto} or
16063 @option{-mcpu=power8}), the following builtins are enabled.
16066 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
16068 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
16069 vector unsigned long long);
16071 vector unsigned long long __builtin_crypto_vcipherlast
16072 (vector unsigned long long,
16073 vector unsigned long long);
16075 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
16076 vector unsigned long long);
16078 vector unsigned long long __builtin_crypto_vncipherlast
16079 (vector unsigned long long,
16080 vector unsigned long long);
16082 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
16083 vector unsigned char,
16084 vector unsigned char);
16086 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
16087 vector unsigned short,
16088 vector unsigned short);
16090 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
16091 vector unsigned int,
16092 vector unsigned int);
16094 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
16095 vector unsigned long long,
16096 vector unsigned long long);
16098 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
16099 vector unsigned char);
16101 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
16102 vector unsigned short);
16104 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
16105 vector unsigned int);
16107 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
16108 vector unsigned long long);
16110 vector unsigned long long __builtin_crypto_vshasigmad
16111 (vector unsigned long long, int, int);
16113 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
16117 The second argument to the @var{__builtin_crypto_vshasigmad} and
16118 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
16119 integer that is 0 or 1. The third argument to these builtin functions
16120 must be a constant integer in the range of 0 to 15.
16122 @node PowerPC Hardware Transactional Memory Built-in Functions
16123 @subsection PowerPC Hardware Transactional Memory Built-in Functions
16124 GCC provides two interfaces for accessing the Hardware Transactional
16125 Memory (HTM) instructions available on some of the PowerPC family
16126 of processors (eg, POWER8). The two interfaces come in a low level
16127 interface, consisting of built-in functions specific to PowerPC and a
16128 higher level interface consisting of inline functions that are common
16129 between PowerPC and S/390.
16131 @subsubsection PowerPC HTM Low Level Built-in Functions
16133 The following low level built-in functions are available with
16134 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16135 They all generate the machine instruction that is part of the name.
16137 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16138 the full 4-bit condition register value set by their associated hardware
16139 instruction. The header file @code{htmintrin.h} defines some macros that can
16140 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16141 returns a simple true or false value depending on whether a transaction was
16142 successfully started or not. The arguments of the builtins match exactly the
16143 type and order of the associated hardware instruction's operands, except for
16144 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16145 Refer to the ISA manual for a description of each instruction's operands.
16148 unsigned int __builtin_tbegin (unsigned int)
16149 unsigned int __builtin_tend (unsigned int)
16151 unsigned int __builtin_tabort (unsigned int)
16152 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16153 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16154 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16155 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16157 unsigned int __builtin_tcheck (void)
16158 unsigned int __builtin_treclaim (unsigned int)
16159 unsigned int __builtin_trechkpt (void)
16160 unsigned int __builtin_tsr (unsigned int)
16163 In addition to the above HTM built-ins, we have added built-ins for
16164 some common extended mnemonics of the HTM instructions:
16167 unsigned int __builtin_tendall (void)
16168 unsigned int __builtin_tresume (void)
16169 unsigned int __builtin_tsuspend (void)
16172 Note that the semantics of the above HTM builtins are required to mimic
16173 the locking semantics used for critical sections. Builtins that are used
16174 to create a new transaction or restart a suspended transaction must have
16175 lock acquisition like semantics while those builtins that end or suspend a
16176 transaction must have lock release like semantics. Specifically, this must
16177 mimic lock semantics as specified by C++11, for example: Lock acquisition is
16178 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
16179 that returns 0, and lock release is as-if an execution of
16180 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
16181 implicit implementation-defined lock used for all transactions. The HTM
16182 instructions associated with with the builtins inherently provide the
16183 correct acquisition and release hardware barriers required. However,
16184 the compiler must also be prohibited from moving loads and stores across
16185 the builtins in a way that would violate their semantics. This has been
16186 accomplished by adding memory barriers to the associated HTM instructions
16187 (which is a conservative approach to provide acquire and release semantics).
16188 Earlier versions of the compiler did not treat the HTM instructions as
16189 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
16190 be used to determine whether the current compiler treats HTM instructions
16191 as memory barriers or not. This allows the user to explicitly add memory
16192 barriers to their code when using an older version of the compiler.
16194 The following set of built-in functions are available to gain access
16195 to the HTM specific special purpose registers.
16198 unsigned long __builtin_get_texasr (void)
16199 unsigned long __builtin_get_texasru (void)
16200 unsigned long __builtin_get_tfhar (void)
16201 unsigned long __builtin_get_tfiar (void)
16203 void __builtin_set_texasr (unsigned long);
16204 void __builtin_set_texasru (unsigned long);
16205 void __builtin_set_tfhar (unsigned long);
16206 void __builtin_set_tfiar (unsigned long);
16209 Example usage of these low level built-in functions may look like:
16212 #include <htmintrin.h>
16214 int num_retries = 10;
16218 if (__builtin_tbegin (0))
16220 /* Transaction State Initiated. */
16221 if (is_locked (lock))
16222 __builtin_tabort (0);
16223 ... transaction code...
16224 __builtin_tend (0);
16229 /* Transaction State Failed. Use locks if the transaction
16230 failure is "persistent" or we've tried too many times. */
16231 if (num_retries-- <= 0
16232 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16234 acquire_lock (lock);
16235 ... non transactional fallback path...
16236 release_lock (lock);
16243 One final built-in function has been added that returns the value of
16244 the 2-bit Transaction State field of the Machine Status Register (MSR)
16245 as stored in @code{CR0}.
16248 unsigned long __builtin_ttest (void)
16251 This built-in can be used to determine the current transaction state
16252 using the following code example:
16255 #include <htmintrin.h>
16257 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16259 if (tx_state == _HTM_TRANSACTIONAL)
16261 /* Code to use in transactional state. */
16263 else if (tx_state == _HTM_NONTRANSACTIONAL)
16265 /* Code to use in non-transactional state. */
16267 else if (tx_state == _HTM_SUSPENDED)
16269 /* Code to use in transaction suspended state. */
16273 @subsubsection PowerPC HTM High Level Inline Functions
16275 The following high level HTM interface is made available by including
16276 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16277 where CPU is `power8' or later. This interface is common between PowerPC
16278 and S/390, allowing users to write one HTM source implementation that
16279 can be compiled and executed on either system.
16282 long __TM_simple_begin (void)
16283 long __TM_begin (void* const TM_buff)
16284 long __TM_end (void)
16285 void __TM_abort (void)
16286 void __TM_named_abort (unsigned char const code)
16287 void __TM_resume (void)
16288 void __TM_suspend (void)
16290 long __TM_is_user_abort (void* const TM_buff)
16291 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16292 long __TM_is_illegal (void* const TM_buff)
16293 long __TM_is_footprint_exceeded (void* const TM_buff)
16294 long __TM_nesting_depth (void* const TM_buff)
16295 long __TM_is_nested_too_deep(void* const TM_buff)
16296 long __TM_is_conflict(void* const TM_buff)
16297 long __TM_is_failure_persistent(void* const TM_buff)
16298 long __TM_failure_address(void* const TM_buff)
16299 long long __TM_failure_code(void* const TM_buff)
16302 Using these common set of HTM inline functions, we can create
16303 a more portable version of the HTM example in the previous
16304 section that will work on either PowerPC or S/390:
16307 #include <htmxlintrin.h>
16309 int num_retries = 10;
16310 TM_buff_type TM_buff;
16314 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16316 /* Transaction State Initiated. */
16317 if (is_locked (lock))
16319 ... transaction code...
16325 /* Transaction State Failed. Use locks if the transaction
16326 failure is "persistent" or we've tried too many times. */
16327 if (num_retries-- <= 0
16328 || __TM_is_failure_persistent (TM_buff))
16330 acquire_lock (lock);
16331 ... non transactional fallback path...
16332 release_lock (lock);
16339 @node RX Built-in Functions
16340 @subsection RX Built-in Functions
16341 GCC supports some of the RX instructions which cannot be expressed in
16342 the C programming language via the use of built-in functions. The
16343 following functions are supported:
16345 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16346 Generates the @code{brk} machine instruction.
16349 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16350 Generates the @code{clrpsw} machine instruction to clear the specified
16351 bit in the processor status word.
16354 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16355 Generates the @code{int} machine instruction to generate an interrupt
16356 with the specified value.
16359 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16360 Generates the @code{machi} machine instruction to add the result of
16361 multiplying the top 16 bits of the two arguments into the
16365 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16366 Generates the @code{maclo} machine instruction to add the result of
16367 multiplying the bottom 16 bits of the two arguments into the
16371 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16372 Generates the @code{mulhi} machine instruction to place the result of
16373 multiplying the top 16 bits of the two arguments into the
16377 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16378 Generates the @code{mullo} machine instruction to place the result of
16379 multiplying the bottom 16 bits of the two arguments into the
16383 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16384 Generates the @code{mvfachi} machine instruction to read the top
16385 32 bits of the accumulator.
16388 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16389 Generates the @code{mvfacmi} machine instruction to read the middle
16390 32 bits of the accumulator.
16393 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16394 Generates the @code{mvfc} machine instruction which reads the control
16395 register specified in its argument and returns its value.
16398 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16399 Generates the @code{mvtachi} machine instruction to set the top
16400 32 bits of the accumulator.
16403 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16404 Generates the @code{mvtaclo} machine instruction to set the bottom
16405 32 bits of the accumulator.
16408 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16409 Generates the @code{mvtc} machine instruction which sets control
16410 register number @code{reg} to @code{val}.
16413 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16414 Generates the @code{mvtipl} machine instruction set the interrupt
16418 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16419 Generates the @code{racw} machine instruction to round the accumulator
16420 according to the specified mode.
16423 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16424 Generates the @code{revw} machine instruction which swaps the bytes in
16425 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16426 and also bits 16--23 occupy bits 24--31 and vice versa.
16429 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16430 Generates the @code{rmpa} machine instruction which initiates a
16431 repeated multiply and accumulate sequence.
16434 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16435 Generates the @code{round} machine instruction which returns the
16436 floating-point argument rounded according to the current rounding mode
16437 set in the floating-point status word register.
16440 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16441 Generates the @code{sat} machine instruction which returns the
16442 saturated value of the argument.
16445 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16446 Generates the @code{setpsw} machine instruction to set the specified
16447 bit in the processor status word.
16450 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16451 Generates the @code{wait} machine instruction.
16454 @node S/390 System z Built-in Functions
16455 @subsection S/390 System z Built-in Functions
16456 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16457 Generates the @code{tbegin} machine instruction starting a
16458 non-constraint hardware transaction. If the parameter is non-NULL the
16459 memory area is used to store the transaction diagnostic buffer and
16460 will be passed as first operand to @code{tbegin}. This buffer can be
16461 defined using the @code{struct __htm_tdb} C struct defined in
16462 @code{htmintrin.h} and must reside on a double-word boundary. The
16463 second tbegin operand is set to @code{0xff0c}. This enables
16464 save/restore of all GPRs and disables aborts for FPR and AR
16465 manipulations inside the transaction body. The condition code set by
16466 the tbegin instruction is returned as integer value. The tbegin
16467 instruction by definition overwrites the content of all FPRs. The
16468 compiler will generate code which saves and restores the FPRs. For
16469 soft-float code it is recommended to used the @code{*_nofloat}
16470 variant. In order to prevent a TDB from being written it is required
16471 to pass an constant zero value as parameter. Passing the zero value
16472 through a variable is not sufficient. Although modifications of
16473 access registers inside the transaction will not trigger an
16474 transaction abort it is not supported to actually modify them. Access
16475 registers do not get saved when entering a transaction. They will have
16476 undefined state when reaching the abort code.
16479 Macros for the possible return codes of tbegin are defined in the
16480 @code{htmintrin.h} header file:
16483 @item _HTM_TBEGIN_STARTED
16484 @code{tbegin} has been executed as part of normal processing. The
16485 transaction body is supposed to be executed.
16486 @item _HTM_TBEGIN_INDETERMINATE
16487 The transaction was aborted due to an indeterminate condition which
16488 might be persistent.
16489 @item _HTM_TBEGIN_TRANSIENT
16490 The transaction aborted due to a transient failure. The transaction
16491 should be re-executed in that case.
16492 @item _HTM_TBEGIN_PERSISTENT
16493 The transaction aborted due to a persistent failure. Re-execution
16494 under same circumstances will not be productive.
16497 @defmac _HTM_FIRST_USER_ABORT_CODE
16498 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16499 specifies the first abort code which can be used for
16500 @code{__builtin_tabort}. Values below this threshold are reserved for
16504 @deftp {Data type} {struct __htm_tdb}
16505 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16506 the structure of the transaction diagnostic block as specified in the
16507 Principles of Operation manual chapter 5-91.
16510 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16511 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16512 Using this variant in code making use of FPRs will leave the FPRs in
16513 undefined state when entering the transaction abort handler code.
16516 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16517 In addition to @code{__builtin_tbegin} a loop for transient failures
16518 is generated. If tbegin returns a condition code of 2 the transaction
16519 will be retried as often as specified in the second argument. The
16520 perform processor assist instruction is used to tell the CPU about the
16521 number of fails so far.
16524 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16525 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16526 restores. Using this variant in code making use of FPRs will leave
16527 the FPRs in undefined state when entering the transaction abort
16531 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16532 Generates the @code{tbeginc} machine instruction starting a constraint
16533 hardware transaction. The second operand is set to @code{0xff08}.
16536 @deftypefn {Built-in Function} int __builtin_tend (void)
16537 Generates the @code{tend} machine instruction finishing a transaction
16538 and making the changes visible to other threads. The condition code
16539 generated by tend is returned as integer value.
16542 @deftypefn {Built-in Function} void __builtin_tabort (int)
16543 Generates the @code{tabort} machine instruction with the specified
16544 abort code. Abort codes from 0 through 255 are reserved and will
16545 result in an error message.
16548 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16549 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16550 integer parameter is loaded into rX and a value of zero is loaded into
16551 rY. The integer parameter specifies the number of times the
16552 transaction repeatedly aborted.
16555 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16556 Generates the @code{etnd} machine instruction. The current nesting
16557 depth is returned as integer value. For a nesting depth of 0 the code
16558 is not executed as part of an transaction.
16561 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16563 Generates the @code{ntstg} machine instruction. The second argument
16564 is written to the first arguments location. The store operation will
16565 not be rolled-back in case of an transaction abort.
16568 @node SH Built-in Functions
16569 @subsection SH Built-in Functions
16570 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16571 families of processors:
16573 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16574 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16575 used by system code that manages threads and execution contexts. The compiler
16576 normally does not generate code that modifies the contents of @samp{GBR} and
16577 thus the value is preserved across function calls. Changing the @samp{GBR}
16578 value in user code must be done with caution, since the compiler might use
16579 @samp{GBR} in order to access thread local variables.
16583 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16584 Returns the value that is currently set in the @samp{GBR} register.
16585 Memory loads and stores that use the thread pointer as a base address are
16586 turned into @samp{GBR} based displacement loads and stores, if possible.
16594 int get_tcb_value (void)
16596 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16597 return ((my_tcb*)__builtin_thread_pointer ())->c;
16603 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16604 Returns the value that is currently set in the @samp{FPSCR} register.
16607 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16608 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16609 preserving the current values of the FR, SZ and PR bits.
16612 @node SPARC VIS Built-in Functions
16613 @subsection SPARC VIS Built-in Functions
16615 GCC supports SIMD operations on the SPARC using both the generic vector
16616 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16617 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16618 switch, the VIS extension is exposed as the following built-in functions:
16621 typedef int v1si __attribute__ ((vector_size (4)));
16622 typedef int v2si __attribute__ ((vector_size (8)));
16623 typedef short v4hi __attribute__ ((vector_size (8)));
16624 typedef short v2hi __attribute__ ((vector_size (4)));
16625 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16626 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16628 void __builtin_vis_write_gsr (int64_t);
16629 int64_t __builtin_vis_read_gsr (void);
16631 void * __builtin_vis_alignaddr (void *, long);
16632 void * __builtin_vis_alignaddrl (void *, long);
16633 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16634 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16635 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16636 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16638 v4hi __builtin_vis_fexpand (v4qi);
16640 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16641 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16642 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16643 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16644 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16645 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16646 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16648 v4qi __builtin_vis_fpack16 (v4hi);
16649 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16650 v2hi __builtin_vis_fpackfix (v2si);
16651 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16653 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
16655 long __builtin_vis_edge8 (void *, void *);
16656 long __builtin_vis_edge8l (void *, void *);
16657 long __builtin_vis_edge16 (void *, void *);
16658 long __builtin_vis_edge16l (void *, void *);
16659 long __builtin_vis_edge32 (void *, void *);
16660 long __builtin_vis_edge32l (void *, void *);
16662 long __builtin_vis_fcmple16 (v4hi, v4hi);
16663 long __builtin_vis_fcmple32 (v2si, v2si);
16664 long __builtin_vis_fcmpne16 (v4hi, v4hi);
16665 long __builtin_vis_fcmpne32 (v2si, v2si);
16666 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
16667 long __builtin_vis_fcmpgt32 (v2si, v2si);
16668 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
16669 long __builtin_vis_fcmpeq32 (v2si, v2si);
16671 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
16672 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
16673 v2si __builtin_vis_fpadd32 (v2si, v2si);
16674 v1si __builtin_vis_fpadd32s (v1si, v1si);
16675 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
16676 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
16677 v2si __builtin_vis_fpsub32 (v2si, v2si);
16678 v1si __builtin_vis_fpsub32s (v1si, v1si);
16680 long __builtin_vis_array8 (long, long);
16681 long __builtin_vis_array16 (long, long);
16682 long __builtin_vis_array32 (long, long);
16685 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
16686 functions also become available:
16689 long __builtin_vis_bmask (long, long);
16690 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
16691 v2si __builtin_vis_bshufflev2si (v2si, v2si);
16692 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
16693 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
16695 long __builtin_vis_edge8n (void *, void *);
16696 long __builtin_vis_edge8ln (void *, void *);
16697 long __builtin_vis_edge16n (void *, void *);
16698 long __builtin_vis_edge16ln (void *, void *);
16699 long __builtin_vis_edge32n (void *, void *);
16700 long __builtin_vis_edge32ln (void *, void *);
16703 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
16704 functions also become available:
16707 void __builtin_vis_cmask8 (long);
16708 void __builtin_vis_cmask16 (long);
16709 void __builtin_vis_cmask32 (long);
16711 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
16713 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
16714 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
16715 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
16716 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
16717 v2si __builtin_vis_fsll16 (v2si, v2si);
16718 v2si __builtin_vis_fslas16 (v2si, v2si);
16719 v2si __builtin_vis_fsrl16 (v2si, v2si);
16720 v2si __builtin_vis_fsra16 (v2si, v2si);
16722 long __builtin_vis_pdistn (v8qi, v8qi);
16724 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
16726 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
16727 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
16729 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
16730 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
16731 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
16732 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
16733 v2si __builtin_vis_fpadds32 (v2si, v2si);
16734 v1si __builtin_vis_fpadds32s (v1si, v1si);
16735 v2si __builtin_vis_fpsubs32 (v2si, v2si);
16736 v1si __builtin_vis_fpsubs32s (v1si, v1si);
16738 long __builtin_vis_fucmple8 (v8qi, v8qi);
16739 long __builtin_vis_fucmpne8 (v8qi, v8qi);
16740 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
16741 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
16743 float __builtin_vis_fhadds (float, float);
16744 double __builtin_vis_fhaddd (double, double);
16745 float __builtin_vis_fhsubs (float, float);
16746 double __builtin_vis_fhsubd (double, double);
16747 float __builtin_vis_fnhadds (float, float);
16748 double __builtin_vis_fnhaddd (double, double);
16750 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
16751 int64_t __builtin_vis_xmulx (int64_t, int64_t);
16752 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
16755 @node SPU Built-in Functions
16756 @subsection SPU Built-in Functions
16758 GCC provides extensions for the SPU processor as described in the
16759 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
16760 found at @uref{http://cell.scei.co.jp/} or
16761 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
16762 implementation differs in several ways.
16767 The optional extension of specifying vector constants in parentheses is
16771 A vector initializer requires no cast if the vector constant is of the
16772 same type as the variable it is initializing.
16775 If @code{signed} or @code{unsigned} is omitted, the signedness of the
16776 vector type is the default signedness of the base type. The default
16777 varies depending on the operating system, so a portable program should
16778 always specify the signedness.
16781 By default, the keyword @code{__vector} is added. The macro
16782 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
16786 GCC allows using a @code{typedef} name as the type specifier for a
16790 For C, overloaded functions are implemented with macros so the following
16794 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
16798 Since @code{spu_add} is a macro, the vector constant in the example
16799 is treated as four separate arguments. Wrap the entire argument in
16800 parentheses for this to work.
16803 The extended version of @code{__builtin_expect} is not supported.
16807 @emph{Note:} Only the interface described in the aforementioned
16808 specification is supported. Internally, GCC uses built-in functions to
16809 implement the required functionality, but these are not supported and
16810 are subject to change without notice.
16812 @node TI C6X Built-in Functions
16813 @subsection TI C6X Built-in Functions
16815 GCC provides intrinsics to access certain instructions of the TI C6X
16816 processors. These intrinsics, listed below, are available after
16817 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
16818 to C6X instructions.
16822 int _sadd (int, int)
16823 int _ssub (int, int)
16824 int _sadd2 (int, int)
16825 int _ssub2 (int, int)
16826 long long _mpy2 (int, int)
16827 long long _smpy2 (int, int)
16828 int _add4 (int, int)
16829 int _sub4 (int, int)
16830 int _saddu4 (int, int)
16832 int _smpy (int, int)
16833 int _smpyh (int, int)
16834 int _smpyhl (int, int)
16835 int _smpylh (int, int)
16837 int _sshl (int, int)
16838 int _subc (int, int)
16840 int _avg2 (int, int)
16841 int _avgu4 (int, int)
16843 int _clrr (int, int)
16844 int _extr (int, int)
16845 int _extru (int, int)
16851 @node TILE-Gx Built-in Functions
16852 @subsection TILE-Gx Built-in Functions
16854 GCC provides intrinsics to access every instruction of the TILE-Gx
16855 processor. The intrinsics are of the form:
16859 unsigned long long __insn_@var{op} (...)
16863 Where @var{op} is the name of the instruction. Refer to the ISA manual
16864 for the complete list of instructions.
16866 GCC also provides intrinsics to directly access the network registers.
16867 The intrinsics are:
16871 unsigned long long __tile_idn0_receive (void)
16872 unsigned long long __tile_idn1_receive (void)
16873 unsigned long long __tile_udn0_receive (void)
16874 unsigned long long __tile_udn1_receive (void)
16875 unsigned long long __tile_udn2_receive (void)
16876 unsigned long long __tile_udn3_receive (void)
16877 void __tile_idn_send (unsigned long long)
16878 void __tile_udn_send (unsigned long long)
16882 The intrinsic @code{void __tile_network_barrier (void)} is used to
16883 guarantee that no network operations before it are reordered with
16886 @node TILEPro Built-in Functions
16887 @subsection TILEPro Built-in Functions
16889 GCC provides intrinsics to access every instruction of the TILEPro
16890 processor. The intrinsics are of the form:
16894 unsigned __insn_@var{op} (...)
16899 where @var{op} is the name of the instruction. Refer to the ISA manual
16900 for the complete list of instructions.
16902 GCC also provides intrinsics to directly access the network registers.
16903 The intrinsics are:
16907 unsigned __tile_idn0_receive (void)
16908 unsigned __tile_idn1_receive (void)
16909 unsigned __tile_sn_receive (void)
16910 unsigned __tile_udn0_receive (void)
16911 unsigned __tile_udn1_receive (void)
16912 unsigned __tile_udn2_receive (void)
16913 unsigned __tile_udn3_receive (void)
16914 void __tile_idn_send (unsigned)
16915 void __tile_sn_send (unsigned)
16916 void __tile_udn_send (unsigned)
16920 The intrinsic @code{void __tile_network_barrier (void)} is used to
16921 guarantee that no network operations before it are reordered with
16924 @node x86 Built-in Functions
16925 @subsection x86 Built-in Functions
16927 These built-in functions are available for the x86-32 and x86-64 family
16928 of computers, depending on the command-line switches used.
16930 If you specify command-line switches such as @option{-msse},
16931 the compiler could use the extended instruction sets even if the built-ins
16932 are not used explicitly in the program. For this reason, applications
16933 that perform run-time CPU detection must compile separate files for each
16934 supported architecture, using the appropriate flags. In particular,
16935 the file containing the CPU detection code should be compiled without
16938 The following machine modes are available for use with MMX built-in functions
16939 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
16940 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
16941 vector of eight 8-bit integers. Some of the built-in functions operate on
16942 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
16944 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
16945 of two 32-bit floating-point values.
16947 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
16948 floating-point values. Some instructions use a vector of four 32-bit
16949 integers, these use @code{V4SI}. Finally, some instructions operate on an
16950 entire vector register, interpreting it as a 128-bit integer, these use mode
16953 In 64-bit mode, the x86-64 family of processors uses additional built-in
16954 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
16955 floating point and @code{TC} 128-bit complex floating-point values.
16957 The following floating-point built-in functions are available in 64-bit
16958 mode. All of them implement the function that is part of the name.
16961 __float128 __builtin_fabsq (__float128)
16962 __float128 __builtin_copysignq (__float128, __float128)
16965 The following built-in function is always available.
16968 @item void __builtin_ia32_pause (void)
16969 Generates the @code{pause} machine instruction with a compiler memory
16973 The following floating-point built-in functions are made available in the
16977 @item __float128 __builtin_infq (void)
16978 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
16979 @findex __builtin_infq
16981 @item __float128 __builtin_huge_valq (void)
16982 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
16983 @findex __builtin_huge_valq
16986 The following built-in functions are always available and can be used to
16987 check the target platform type.
16989 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
16990 This function runs the CPU detection code to check the type of CPU and the
16991 features supported. This built-in function needs to be invoked along with the built-in functions
16992 to check CPU type and features, @code{__builtin_cpu_is} and
16993 @code{__builtin_cpu_supports}, only when used in a function that is
16994 executed before any constructors are called. The CPU detection code is
16995 automatically executed in a very high priority constructor.
16997 For example, this function has to be used in @code{ifunc} resolvers that
16998 check for CPU type using the built-in functions @code{__builtin_cpu_is}
16999 and @code{__builtin_cpu_supports}, or in constructors on targets that
17000 don't support constructor priority.
17003 static void (*resolve_memcpy (void)) (void)
17005 // ifunc resolvers fire before constructors, explicitly call the init
17007 __builtin_cpu_init ();
17008 if (__builtin_cpu_supports ("ssse3"))
17009 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
17011 return default_memcpy;
17014 void *memcpy (void *, const void *, size_t)
17015 __attribute__ ((ifunc ("resolve_memcpy")));
17020 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
17021 This function returns a positive integer if the run-time CPU
17022 is of type @var{cpuname}
17023 and returns @code{0} otherwise. The following CPU names can be detected:
17039 Intel Core i7 Nehalem CPU.
17042 Intel Core i7 Westmere CPU.
17045 Intel Core i7 Sandy Bridge CPU.
17051 AMD Family 10h CPU.
17054 AMD Family 10h Barcelona CPU.
17057 AMD Family 10h Shanghai CPU.
17060 AMD Family 10h Istanbul CPU.
17063 AMD Family 14h CPU.
17066 AMD Family 15h CPU.
17069 AMD Family 15h Bulldozer version 1.
17072 AMD Family 15h Bulldozer version 2.
17075 AMD Family 15h Bulldozer version 3.
17078 AMD Family 15h Bulldozer version 4.
17081 AMD Family 16h CPU.
17084 AMD Family 17h CPU.
17087 Here is an example:
17089 if (__builtin_cpu_is ("corei7"))
17091 do_corei7 (); // Core i7 specific implementation.
17095 do_generic (); // Generic implementation.
17100 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
17101 This function returns a positive integer if the run-time CPU
17102 supports @var{feature}
17103 and returns @code{0} otherwise. The following features can be detected:
17111 POPCNT instruction.
17119 SSSE3 instructions.
17121 SSE4.1 instructions.
17123 SSE4.2 instructions.
17129 AVX512F instructions.
17132 Here is an example:
17134 if (__builtin_cpu_supports ("popcnt"))
17136 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
17140 count = generic_countbits (n); //generic implementation.
17146 The following built-in functions are made available by @option{-mmmx}.
17147 All of them generate the machine instruction that is part of the name.
17150 v8qi __builtin_ia32_paddb (v8qi, v8qi)
17151 v4hi __builtin_ia32_paddw (v4hi, v4hi)
17152 v2si __builtin_ia32_paddd (v2si, v2si)
17153 v8qi __builtin_ia32_psubb (v8qi, v8qi)
17154 v4hi __builtin_ia32_psubw (v4hi, v4hi)
17155 v2si __builtin_ia32_psubd (v2si, v2si)
17156 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17157 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17158 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17159 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17160 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17161 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17162 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17163 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17164 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17165 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17166 di __builtin_ia32_pand (di, di)
17167 di __builtin_ia32_pandn (di,di)
17168 di __builtin_ia32_por (di, di)
17169 di __builtin_ia32_pxor (di, di)
17170 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17171 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17172 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17173 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17174 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17175 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17176 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17177 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17178 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17179 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17180 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17181 v2si __builtin_ia32_punpckldq (v2si, v2si)
17182 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17183 v4hi __builtin_ia32_packssdw (v2si, v2si)
17184 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17186 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17187 v2si __builtin_ia32_pslld (v2si, v2si)
17188 v1di __builtin_ia32_psllq (v1di, v1di)
17189 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17190 v2si __builtin_ia32_psrld (v2si, v2si)
17191 v1di __builtin_ia32_psrlq (v1di, v1di)
17192 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17193 v2si __builtin_ia32_psrad (v2si, v2si)
17194 v4hi __builtin_ia32_psllwi (v4hi, int)
17195 v2si __builtin_ia32_pslldi (v2si, int)
17196 v1di __builtin_ia32_psllqi (v1di, int)
17197 v4hi __builtin_ia32_psrlwi (v4hi, int)
17198 v2si __builtin_ia32_psrldi (v2si, int)
17199 v1di __builtin_ia32_psrlqi (v1di, int)
17200 v4hi __builtin_ia32_psrawi (v4hi, int)
17201 v2si __builtin_ia32_psradi (v2si, int)
17205 The following built-in functions are made available either with
17206 @option{-msse}, or with a combination of @option{-m3dnow} and
17207 @option{-march=athlon}. All of them generate the machine
17208 instruction that is part of the name.
17211 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17212 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17213 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17214 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17215 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17216 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17217 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17218 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17219 int __builtin_ia32_pmovmskb (v8qi)
17220 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17221 void __builtin_ia32_movntq (di *, di)
17222 void __builtin_ia32_sfence (void)
17225 The following built-in functions are available when @option{-msse} is used.
17226 All of them generate the machine instruction that is part of the name.
17229 int __builtin_ia32_comieq (v4sf, v4sf)
17230 int __builtin_ia32_comineq (v4sf, v4sf)
17231 int __builtin_ia32_comilt (v4sf, v4sf)
17232 int __builtin_ia32_comile (v4sf, v4sf)
17233 int __builtin_ia32_comigt (v4sf, v4sf)
17234 int __builtin_ia32_comige (v4sf, v4sf)
17235 int __builtin_ia32_ucomieq (v4sf, v4sf)
17236 int __builtin_ia32_ucomineq (v4sf, v4sf)
17237 int __builtin_ia32_ucomilt (v4sf, v4sf)
17238 int __builtin_ia32_ucomile (v4sf, v4sf)
17239 int __builtin_ia32_ucomigt (v4sf, v4sf)
17240 int __builtin_ia32_ucomige (v4sf, v4sf)
17241 v4sf __builtin_ia32_addps (v4sf, v4sf)
17242 v4sf __builtin_ia32_subps (v4sf, v4sf)
17243 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17244 v4sf __builtin_ia32_divps (v4sf, v4sf)
17245 v4sf __builtin_ia32_addss (v4sf, v4sf)
17246 v4sf __builtin_ia32_subss (v4sf, v4sf)
17247 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17248 v4sf __builtin_ia32_divss (v4sf, v4sf)
17249 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17250 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17251 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17252 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17253 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17254 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17255 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17256 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17257 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17258 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17259 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17260 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17261 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17262 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17263 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17264 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17265 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17266 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17267 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17268 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17269 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17270 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17271 v4sf __builtin_ia32_minps (v4sf, v4sf)
17272 v4sf __builtin_ia32_minss (v4sf, v4sf)
17273 v4sf __builtin_ia32_andps (v4sf, v4sf)
17274 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17275 v4sf __builtin_ia32_orps (v4sf, v4sf)
17276 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17277 v4sf __builtin_ia32_movss (v4sf, v4sf)
17278 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17279 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17280 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17281 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17282 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17283 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17284 v2si __builtin_ia32_cvtps2pi (v4sf)
17285 int __builtin_ia32_cvtss2si (v4sf)
17286 v2si __builtin_ia32_cvttps2pi (v4sf)
17287 int __builtin_ia32_cvttss2si (v4sf)
17288 v4sf __builtin_ia32_rcpps (v4sf)
17289 v4sf __builtin_ia32_rsqrtps (v4sf)
17290 v4sf __builtin_ia32_sqrtps (v4sf)
17291 v4sf __builtin_ia32_rcpss (v4sf)
17292 v4sf __builtin_ia32_rsqrtss (v4sf)
17293 v4sf __builtin_ia32_sqrtss (v4sf)
17294 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17295 void __builtin_ia32_movntps (float *, v4sf)
17296 int __builtin_ia32_movmskps (v4sf)
17299 The following built-in functions are available when @option{-msse} is used.
17302 @item v4sf __builtin_ia32_loadups (float *)
17303 Generates the @code{movups} machine instruction as a load from memory.
17304 @item void __builtin_ia32_storeups (float *, v4sf)
17305 Generates the @code{movups} machine instruction as a store to memory.
17306 @item v4sf __builtin_ia32_loadss (float *)
17307 Generates the @code{movss} machine instruction as a load from memory.
17308 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17309 Generates the @code{movhps} machine instruction as a load from memory.
17310 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17311 Generates the @code{movlps} machine instruction as a load from memory
17312 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17313 Generates the @code{movhps} machine instruction as a store to memory.
17314 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17315 Generates the @code{movlps} machine instruction as a store to memory.
17318 The following built-in functions are available when @option{-msse2} is used.
17319 All of them generate the machine instruction that is part of the name.
17322 int __builtin_ia32_comisdeq (v2df, v2df)
17323 int __builtin_ia32_comisdlt (v2df, v2df)
17324 int __builtin_ia32_comisdle (v2df, v2df)
17325 int __builtin_ia32_comisdgt (v2df, v2df)
17326 int __builtin_ia32_comisdge (v2df, v2df)
17327 int __builtin_ia32_comisdneq (v2df, v2df)
17328 int __builtin_ia32_ucomisdeq (v2df, v2df)
17329 int __builtin_ia32_ucomisdlt (v2df, v2df)
17330 int __builtin_ia32_ucomisdle (v2df, v2df)
17331 int __builtin_ia32_ucomisdgt (v2df, v2df)
17332 int __builtin_ia32_ucomisdge (v2df, v2df)
17333 int __builtin_ia32_ucomisdneq (v2df, v2df)
17334 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17335 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17336 v2df __builtin_ia32_cmplepd (v2df, v2df)
17337 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17338 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17339 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17340 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17341 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17342 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17343 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17344 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17345 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17346 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17347 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17348 v2df __builtin_ia32_cmplesd (v2df, v2df)
17349 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17350 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17351 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17352 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17353 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17354 v2di __builtin_ia32_paddq (v2di, v2di)
17355 v2di __builtin_ia32_psubq (v2di, v2di)
17356 v2df __builtin_ia32_addpd (v2df, v2df)
17357 v2df __builtin_ia32_subpd (v2df, v2df)
17358 v2df __builtin_ia32_mulpd (v2df, v2df)
17359 v2df __builtin_ia32_divpd (v2df, v2df)
17360 v2df __builtin_ia32_addsd (v2df, v2df)
17361 v2df __builtin_ia32_subsd (v2df, v2df)
17362 v2df __builtin_ia32_mulsd (v2df, v2df)
17363 v2df __builtin_ia32_divsd (v2df, v2df)
17364 v2df __builtin_ia32_minpd (v2df, v2df)
17365 v2df __builtin_ia32_maxpd (v2df, v2df)
17366 v2df __builtin_ia32_minsd (v2df, v2df)
17367 v2df __builtin_ia32_maxsd (v2df, v2df)
17368 v2df __builtin_ia32_andpd (v2df, v2df)
17369 v2df __builtin_ia32_andnpd (v2df, v2df)
17370 v2df __builtin_ia32_orpd (v2df, v2df)
17371 v2df __builtin_ia32_xorpd (v2df, v2df)
17372 v2df __builtin_ia32_movsd (v2df, v2df)
17373 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17374 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17375 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17376 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17377 v4si __builtin_ia32_paddd128 (v4si, v4si)
17378 v2di __builtin_ia32_paddq128 (v2di, v2di)
17379 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17380 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17381 v4si __builtin_ia32_psubd128 (v4si, v4si)
17382 v2di __builtin_ia32_psubq128 (v2di, v2di)
17383 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17384 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17385 v2di __builtin_ia32_pand128 (v2di, v2di)
17386 v2di __builtin_ia32_pandn128 (v2di, v2di)
17387 v2di __builtin_ia32_por128 (v2di, v2di)
17388 v2di __builtin_ia32_pxor128 (v2di, v2di)
17389 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17390 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17391 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17392 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17393 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17394 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17395 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17396 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17397 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17398 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17399 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17400 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17401 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17402 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17403 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17404 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17405 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17406 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17407 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17408 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17409 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17410 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17411 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17412 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17413 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17414 v2df __builtin_ia32_loadupd (double *)
17415 void __builtin_ia32_storeupd (double *, v2df)
17416 v2df __builtin_ia32_loadhpd (v2df, double const *)
17417 v2df __builtin_ia32_loadlpd (v2df, double const *)
17418 int __builtin_ia32_movmskpd (v2df)
17419 int __builtin_ia32_pmovmskb128 (v16qi)
17420 void __builtin_ia32_movnti (int *, int)
17421 void __builtin_ia32_movnti64 (long long int *, long long int)
17422 void __builtin_ia32_movntpd (double *, v2df)
17423 void __builtin_ia32_movntdq (v2df *, v2df)
17424 v4si __builtin_ia32_pshufd (v4si, int)
17425 v8hi __builtin_ia32_pshuflw (v8hi, int)
17426 v8hi __builtin_ia32_pshufhw (v8hi, int)
17427 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17428 v2df __builtin_ia32_sqrtpd (v2df)
17429 v2df __builtin_ia32_sqrtsd (v2df)
17430 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17431 v2df __builtin_ia32_cvtdq2pd (v4si)
17432 v4sf __builtin_ia32_cvtdq2ps (v4si)
17433 v4si __builtin_ia32_cvtpd2dq (v2df)
17434 v2si __builtin_ia32_cvtpd2pi (v2df)
17435 v4sf __builtin_ia32_cvtpd2ps (v2df)
17436 v4si __builtin_ia32_cvttpd2dq (v2df)
17437 v2si __builtin_ia32_cvttpd2pi (v2df)
17438 v2df __builtin_ia32_cvtpi2pd (v2si)
17439 int __builtin_ia32_cvtsd2si (v2df)
17440 int __builtin_ia32_cvttsd2si (v2df)
17441 long long __builtin_ia32_cvtsd2si64 (v2df)
17442 long long __builtin_ia32_cvttsd2si64 (v2df)
17443 v4si __builtin_ia32_cvtps2dq (v4sf)
17444 v2df __builtin_ia32_cvtps2pd (v4sf)
17445 v4si __builtin_ia32_cvttps2dq (v4sf)
17446 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17447 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17448 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17449 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17450 void __builtin_ia32_clflush (const void *)
17451 void __builtin_ia32_lfence (void)
17452 void __builtin_ia32_mfence (void)
17453 v16qi __builtin_ia32_loaddqu (const char *)
17454 void __builtin_ia32_storedqu (char *, v16qi)
17455 v1di __builtin_ia32_pmuludq (v2si, v2si)
17456 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17457 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17458 v4si __builtin_ia32_pslld128 (v4si, v4si)
17459 v2di __builtin_ia32_psllq128 (v2di, v2di)
17460 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17461 v4si __builtin_ia32_psrld128 (v4si, v4si)
17462 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17463 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17464 v4si __builtin_ia32_psrad128 (v4si, v4si)
17465 v2di __builtin_ia32_pslldqi128 (v2di, int)
17466 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17467 v4si __builtin_ia32_pslldi128 (v4si, int)
17468 v2di __builtin_ia32_psllqi128 (v2di, int)
17469 v2di __builtin_ia32_psrldqi128 (v2di, int)
17470 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17471 v4si __builtin_ia32_psrldi128 (v4si, int)
17472 v2di __builtin_ia32_psrlqi128 (v2di, int)
17473 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17474 v4si __builtin_ia32_psradi128 (v4si, int)
17475 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17476 v2di __builtin_ia32_movq128 (v2di)
17479 The following built-in functions are available when @option{-msse3} is used.
17480 All of them generate the machine instruction that is part of the name.
17483 v2df __builtin_ia32_addsubpd (v2df, v2df)
17484 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17485 v2df __builtin_ia32_haddpd (v2df, v2df)
17486 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17487 v2df __builtin_ia32_hsubpd (v2df, v2df)
17488 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17489 v16qi __builtin_ia32_lddqu (char const *)
17490 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17491 v4sf __builtin_ia32_movshdup (v4sf)
17492 v4sf __builtin_ia32_movsldup (v4sf)
17493 void __builtin_ia32_mwait (unsigned int, unsigned int)
17496 The following built-in functions are available when @option{-mssse3} is used.
17497 All of them generate the machine instruction that is part of the name.
17500 v2si __builtin_ia32_phaddd (v2si, v2si)
17501 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17502 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17503 v2si __builtin_ia32_phsubd (v2si, v2si)
17504 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17505 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17506 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17507 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17508 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17509 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17510 v2si __builtin_ia32_psignd (v2si, v2si)
17511 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17512 v1di __builtin_ia32_palignr (v1di, v1di, int)
17513 v8qi __builtin_ia32_pabsb (v8qi)
17514 v2si __builtin_ia32_pabsd (v2si)
17515 v4hi __builtin_ia32_pabsw (v4hi)
17518 The following built-in functions are available when @option{-mssse3} is used.
17519 All of them generate the machine instruction that is part of the name.
17522 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17523 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17524 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17525 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17526 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17527 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17528 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17529 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17530 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17531 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17532 v4si __builtin_ia32_psignd128 (v4si, v4si)
17533 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17534 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17535 v16qi __builtin_ia32_pabsb128 (v16qi)
17536 v4si __builtin_ia32_pabsd128 (v4si)
17537 v8hi __builtin_ia32_pabsw128 (v8hi)
17540 The following built-in functions are available when @option{-msse4.1} is
17541 used. All of them generate the machine instruction that is part of the
17545 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17546 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17547 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17548 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17549 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17550 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17551 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17552 v2di __builtin_ia32_movntdqa (v2di *);
17553 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17554 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17555 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17556 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17557 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17558 v8hi __builtin_ia32_phminposuw128 (v8hi)
17559 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17560 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17561 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17562 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17563 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17564 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17565 v4si __builtin_ia32_pminud128 (v4si, v4si)
17566 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17567 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17568 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17569 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17570 v2di __builtin_ia32_pmovsxdq128 (v4si)
17571 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17572 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17573 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17574 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17575 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17576 v2di __builtin_ia32_pmovzxdq128 (v4si)
17577 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17578 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17579 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17580 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17581 int __builtin_ia32_ptestc128 (v2di, v2di)
17582 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17583 int __builtin_ia32_ptestz128 (v2di, v2di)
17584 v2df __builtin_ia32_roundpd (v2df, const int)
17585 v4sf __builtin_ia32_roundps (v4sf, const int)
17586 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17587 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17590 The following built-in functions are available when @option{-msse4.1} is
17594 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17595 Generates the @code{insertps} machine instruction.
17596 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17597 Generates the @code{pextrb} machine instruction.
17598 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17599 Generates the @code{pinsrb} machine instruction.
17600 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17601 Generates the @code{pinsrd} machine instruction.
17602 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17603 Generates the @code{pinsrq} machine instruction in 64bit mode.
17606 The following built-in functions are changed to generate new SSE4.1
17607 instructions when @option{-msse4.1} is used.
17610 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17611 Generates the @code{extractps} machine instruction.
17612 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17613 Generates the @code{pextrd} machine instruction.
17614 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17615 Generates the @code{pextrq} machine instruction in 64bit mode.
17618 The following built-in functions are available when @option{-msse4.2} is
17619 used. All of them generate the machine instruction that is part of the
17623 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17624 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17625 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17626 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17627 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17628 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17629 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17630 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17631 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17632 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17633 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17634 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17635 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17636 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17637 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17640 The following built-in functions are available when @option{-msse4.2} is
17644 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17645 Generates the @code{crc32b} machine instruction.
17646 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17647 Generates the @code{crc32w} machine instruction.
17648 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17649 Generates the @code{crc32l} machine instruction.
17650 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17651 Generates the @code{crc32q} machine instruction.
17654 The following built-in functions are changed to generate new SSE4.2
17655 instructions when @option{-msse4.2} is used.
17658 @item int __builtin_popcount (unsigned int)
17659 Generates the @code{popcntl} machine instruction.
17660 @item int __builtin_popcountl (unsigned long)
17661 Generates the @code{popcntl} or @code{popcntq} machine instruction,
17662 depending on the size of @code{unsigned long}.
17663 @item int __builtin_popcountll (unsigned long long)
17664 Generates the @code{popcntq} machine instruction.
17667 The following built-in functions are available when @option{-mavx} is
17668 used. All of them generate the machine instruction that is part of the
17672 v4df __builtin_ia32_addpd256 (v4df,v4df)
17673 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
17674 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
17675 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
17676 v4df __builtin_ia32_andnpd256 (v4df,v4df)
17677 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
17678 v4df __builtin_ia32_andpd256 (v4df,v4df)
17679 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
17680 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
17681 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
17682 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
17683 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
17684 v2df __builtin_ia32_cmppd (v2df,v2df,int)
17685 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
17686 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
17687 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
17688 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
17689 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
17690 v4df __builtin_ia32_cvtdq2pd256 (v4si)
17691 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
17692 v4si __builtin_ia32_cvtpd2dq256 (v4df)
17693 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
17694 v8si __builtin_ia32_cvtps2dq256 (v8sf)
17695 v4df __builtin_ia32_cvtps2pd256 (v4sf)
17696 v4si __builtin_ia32_cvttpd2dq256 (v4df)
17697 v8si __builtin_ia32_cvttps2dq256 (v8sf)
17698 v4df __builtin_ia32_divpd256 (v4df,v4df)
17699 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
17700 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
17701 v4df __builtin_ia32_haddpd256 (v4df,v4df)
17702 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
17703 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
17704 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
17705 v32qi __builtin_ia32_lddqu256 (pcchar)
17706 v32qi __builtin_ia32_loaddqu256 (pcchar)
17707 v4df __builtin_ia32_loadupd256 (pcdouble)
17708 v8sf __builtin_ia32_loadups256 (pcfloat)
17709 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
17710 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
17711 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
17712 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
17713 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
17714 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
17715 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
17716 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
17717 v4df __builtin_ia32_maxpd256 (v4df,v4df)
17718 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
17719 v4df __builtin_ia32_minpd256 (v4df,v4df)
17720 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
17721 v4df __builtin_ia32_movddup256 (v4df)
17722 int __builtin_ia32_movmskpd256 (v4df)
17723 int __builtin_ia32_movmskps256 (v8sf)
17724 v8sf __builtin_ia32_movshdup256 (v8sf)
17725 v8sf __builtin_ia32_movsldup256 (v8sf)
17726 v4df __builtin_ia32_mulpd256 (v4df,v4df)
17727 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
17728 v4df __builtin_ia32_orpd256 (v4df,v4df)
17729 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
17730 v2df __builtin_ia32_pd_pd256 (v4df)
17731 v4df __builtin_ia32_pd256_pd (v2df)
17732 v4sf __builtin_ia32_ps_ps256 (v8sf)
17733 v8sf __builtin_ia32_ps256_ps (v4sf)
17734 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
17735 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
17736 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
17737 v8sf __builtin_ia32_rcpps256 (v8sf)
17738 v4df __builtin_ia32_roundpd256 (v4df,int)
17739 v8sf __builtin_ia32_roundps256 (v8sf,int)
17740 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
17741 v8sf __builtin_ia32_rsqrtps256 (v8sf)
17742 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
17743 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
17744 v4si __builtin_ia32_si_si256 (v8si)
17745 v8si __builtin_ia32_si256_si (v4si)
17746 v4df __builtin_ia32_sqrtpd256 (v4df)
17747 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
17748 v8sf __builtin_ia32_sqrtps256 (v8sf)
17749 void __builtin_ia32_storedqu256 (pchar,v32qi)
17750 void __builtin_ia32_storeupd256 (pdouble,v4df)
17751 void __builtin_ia32_storeups256 (pfloat,v8sf)
17752 v4df __builtin_ia32_subpd256 (v4df,v4df)
17753 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
17754 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
17755 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
17756 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
17757 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
17758 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
17759 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
17760 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
17761 v4sf __builtin_ia32_vbroadcastss (pcfloat)
17762 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
17763 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
17764 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
17765 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
17766 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
17767 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
17768 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
17769 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
17770 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
17771 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
17772 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
17773 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
17774 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
17775 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
17776 v2df __builtin_ia32_vpermilpd (v2df,int)
17777 v4df __builtin_ia32_vpermilpd256 (v4df,int)
17778 v4sf __builtin_ia32_vpermilps (v4sf,int)
17779 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
17780 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
17781 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
17782 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
17783 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
17784 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
17785 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
17786 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
17787 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
17788 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
17789 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
17790 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
17791 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
17792 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
17793 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
17794 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
17795 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
17796 void __builtin_ia32_vzeroall (void)
17797 void __builtin_ia32_vzeroupper (void)
17798 v4df __builtin_ia32_xorpd256 (v4df,v4df)
17799 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
17802 The following built-in functions are available when @option{-mavx2} is
17803 used. All of them generate the machine instruction that is part of the
17807 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
17808 v32qi __builtin_ia32_pabsb256 (v32qi)
17809 v16hi __builtin_ia32_pabsw256 (v16hi)
17810 v8si __builtin_ia32_pabsd256 (v8si)
17811 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
17812 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
17813 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
17814 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
17815 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
17816 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
17817 v8si __builtin_ia32_paddd256 (v8si,v8si)
17818 v4di __builtin_ia32_paddq256 (v4di,v4di)
17819 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
17820 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
17821 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
17822 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
17823 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
17824 v4di __builtin_ia32_andsi256 (v4di,v4di)
17825 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
17826 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
17827 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
17828 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
17829 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
17830 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
17831 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
17832 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
17833 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
17834 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
17835 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
17836 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
17837 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
17838 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
17839 v8si __builtin_ia32_phaddd256 (v8si,v8si)
17840 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
17841 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
17842 v8si __builtin_ia32_phsubd256 (v8si,v8si)
17843 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
17844 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
17845 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
17846 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
17847 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
17848 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
17849 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
17850 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
17851 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
17852 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
17853 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
17854 v8si __builtin_ia32_pminsd256 (v8si,v8si)
17855 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
17856 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
17857 v8si __builtin_ia32_pminud256 (v8si,v8si)
17858 int __builtin_ia32_pmovmskb256 (v32qi)
17859 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
17860 v8si __builtin_ia32_pmovsxbd256 (v16qi)
17861 v4di __builtin_ia32_pmovsxbq256 (v16qi)
17862 v8si __builtin_ia32_pmovsxwd256 (v8hi)
17863 v4di __builtin_ia32_pmovsxwq256 (v8hi)
17864 v4di __builtin_ia32_pmovsxdq256 (v4si)
17865 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
17866 v8si __builtin_ia32_pmovzxbd256 (v16qi)
17867 v4di __builtin_ia32_pmovzxbq256 (v16qi)
17868 v8si __builtin_ia32_pmovzxwd256 (v8hi)
17869 v4di __builtin_ia32_pmovzxwq256 (v8hi)
17870 v4di __builtin_ia32_pmovzxdq256 (v4si)
17871 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
17872 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
17873 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
17874 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
17875 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
17876 v8si __builtin_ia32_pmulld256 (v8si,v8si)
17877 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
17878 v4di __builtin_ia32_por256 (v4di,v4di)
17879 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
17880 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
17881 v8si __builtin_ia32_pshufd256 (v8si,int)
17882 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
17883 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
17884 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
17885 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
17886 v8si __builtin_ia32_psignd256 (v8si,v8si)
17887 v4di __builtin_ia32_pslldqi256 (v4di,int)
17888 v16hi __builtin_ia32_psllwi256 (16hi,int)
17889 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
17890 v8si __builtin_ia32_pslldi256 (v8si,int)
17891 v8si __builtin_ia32_pslld256(v8si,v4si)
17892 v4di __builtin_ia32_psllqi256 (v4di,int)
17893 v4di __builtin_ia32_psllq256(v4di,v2di)
17894 v16hi __builtin_ia32_psrawi256 (v16hi,int)
17895 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
17896 v8si __builtin_ia32_psradi256 (v8si,int)
17897 v8si __builtin_ia32_psrad256 (v8si,v4si)
17898 v4di __builtin_ia32_psrldqi256 (v4di, int)
17899 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
17900 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
17901 v8si __builtin_ia32_psrldi256 (v8si,int)
17902 v8si __builtin_ia32_psrld256 (v8si,v4si)
17903 v4di __builtin_ia32_psrlqi256 (v4di,int)
17904 v4di __builtin_ia32_psrlq256(v4di,v2di)
17905 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
17906 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
17907 v8si __builtin_ia32_psubd256 (v8si,v8si)
17908 v4di __builtin_ia32_psubq256 (v4di,v4di)
17909 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
17910 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
17911 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
17912 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
17913 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
17914 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
17915 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
17916 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
17917 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
17918 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
17919 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
17920 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
17921 v4di __builtin_ia32_pxor256 (v4di,v4di)
17922 v4di __builtin_ia32_movntdqa256 (pv4di)
17923 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
17924 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
17925 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
17926 v4di __builtin_ia32_vbroadcastsi256 (v2di)
17927 v4si __builtin_ia32_pblendd128 (v4si,v4si)
17928 v8si __builtin_ia32_pblendd256 (v8si,v8si)
17929 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
17930 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
17931 v8si __builtin_ia32_pbroadcastd256 (v4si)
17932 v4di __builtin_ia32_pbroadcastq256 (v2di)
17933 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
17934 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
17935 v4si __builtin_ia32_pbroadcastd128 (v4si)
17936 v2di __builtin_ia32_pbroadcastq128 (v2di)
17937 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
17938 v4df __builtin_ia32_permdf256 (v4df,int)
17939 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
17940 v4di __builtin_ia32_permdi256 (v4di,int)
17941 v4di __builtin_ia32_permti256 (v4di,v4di,int)
17942 v4di __builtin_ia32_extract128i256 (v4di,int)
17943 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
17944 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
17945 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
17946 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
17947 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
17948 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
17949 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
17950 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
17951 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
17952 v8si __builtin_ia32_psllv8si (v8si,v8si)
17953 v4si __builtin_ia32_psllv4si (v4si,v4si)
17954 v4di __builtin_ia32_psllv4di (v4di,v4di)
17955 v2di __builtin_ia32_psllv2di (v2di,v2di)
17956 v8si __builtin_ia32_psrav8si (v8si,v8si)
17957 v4si __builtin_ia32_psrav4si (v4si,v4si)
17958 v8si __builtin_ia32_psrlv8si (v8si,v8si)
17959 v4si __builtin_ia32_psrlv4si (v4si,v4si)
17960 v4di __builtin_ia32_psrlv4di (v4di,v4di)
17961 v2di __builtin_ia32_psrlv2di (v2di,v2di)
17962 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
17963 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
17964 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
17965 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
17966 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
17967 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
17968 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
17969 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
17970 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
17971 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
17972 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
17973 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
17974 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
17975 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
17976 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
17977 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
17980 The following built-in functions are available when @option{-maes} is
17981 used. All of them generate the machine instruction that is part of the
17985 v2di __builtin_ia32_aesenc128 (v2di, v2di)
17986 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
17987 v2di __builtin_ia32_aesdec128 (v2di, v2di)
17988 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
17989 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
17990 v2di __builtin_ia32_aesimc128 (v2di)
17993 The following built-in function is available when @option{-mpclmul} is
17997 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
17998 Generates the @code{pclmulqdq} machine instruction.
18001 The following built-in function is available when @option{-mfsgsbase} is
18002 used. All of them generate the machine instruction that is part of the
18006 unsigned int __builtin_ia32_rdfsbase32 (void)
18007 unsigned long long __builtin_ia32_rdfsbase64 (void)
18008 unsigned int __builtin_ia32_rdgsbase32 (void)
18009 unsigned long long __builtin_ia32_rdgsbase64 (void)
18010 void _writefsbase_u32 (unsigned int)
18011 void _writefsbase_u64 (unsigned long long)
18012 void _writegsbase_u32 (unsigned int)
18013 void _writegsbase_u64 (unsigned long long)
18016 The following built-in function is available when @option{-mrdrnd} is
18017 used. All of them generate the machine instruction that is part of the
18021 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
18022 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
18023 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
18026 The following built-in functions are available when @option{-msse4a} is used.
18027 All of them generate the machine instruction that is part of the name.
18030 void __builtin_ia32_movntsd (double *, v2df)
18031 void __builtin_ia32_movntss (float *, v4sf)
18032 v2di __builtin_ia32_extrq (v2di, v16qi)
18033 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
18034 v2di __builtin_ia32_insertq (v2di, v2di)
18035 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
18038 The following built-in functions are available when @option{-mxop} is used.
18040 v2df __builtin_ia32_vfrczpd (v2df)
18041 v4sf __builtin_ia32_vfrczps (v4sf)
18042 v2df __builtin_ia32_vfrczsd (v2df)
18043 v4sf __builtin_ia32_vfrczss (v4sf)
18044 v4df __builtin_ia32_vfrczpd256 (v4df)
18045 v8sf __builtin_ia32_vfrczps256 (v8sf)
18046 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
18047 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
18048 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
18049 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
18050 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
18051 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
18052 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
18053 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
18054 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
18055 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
18056 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
18057 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
18058 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
18059 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
18060 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18061 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
18062 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
18063 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
18064 v4si __builtin_ia32_vpcomequd (v4si, v4si)
18065 v2di __builtin_ia32_vpcomequq (v2di, v2di)
18066 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
18067 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18068 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
18069 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
18070 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
18071 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
18072 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
18073 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
18074 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
18075 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
18076 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
18077 v4si __builtin_ia32_vpcomged (v4si, v4si)
18078 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
18079 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
18080 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
18081 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
18082 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
18083 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
18084 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
18085 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
18086 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
18087 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
18088 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
18089 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
18090 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
18091 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
18092 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
18093 v4si __builtin_ia32_vpcomled (v4si, v4si)
18094 v2di __builtin_ia32_vpcomleq (v2di, v2di)
18095 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
18096 v4si __builtin_ia32_vpcomleud (v4si, v4si)
18097 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
18098 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
18099 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
18100 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
18101 v4si __builtin_ia32_vpcomltd (v4si, v4si)
18102 v2di __builtin_ia32_vpcomltq (v2di, v2di)
18103 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
18104 v4si __builtin_ia32_vpcomltud (v4si, v4si)
18105 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
18106 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
18107 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
18108 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
18109 v4si __builtin_ia32_vpcomned (v4si, v4si)
18110 v2di __builtin_ia32_vpcomneq (v2di, v2di)
18111 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
18112 v4si __builtin_ia32_vpcomneud (v4si, v4si)
18113 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
18114 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
18115 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
18116 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
18117 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
18118 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
18119 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
18120 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
18121 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
18122 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
18123 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
18124 v4si __builtin_ia32_vphaddbd (v16qi)
18125 v2di __builtin_ia32_vphaddbq (v16qi)
18126 v8hi __builtin_ia32_vphaddbw (v16qi)
18127 v2di __builtin_ia32_vphadddq (v4si)
18128 v4si __builtin_ia32_vphaddubd (v16qi)
18129 v2di __builtin_ia32_vphaddubq (v16qi)
18130 v8hi __builtin_ia32_vphaddubw (v16qi)
18131 v2di __builtin_ia32_vphaddudq (v4si)
18132 v4si __builtin_ia32_vphadduwd (v8hi)
18133 v2di __builtin_ia32_vphadduwq (v8hi)
18134 v4si __builtin_ia32_vphaddwd (v8hi)
18135 v2di __builtin_ia32_vphaddwq (v8hi)
18136 v8hi __builtin_ia32_vphsubbw (v16qi)
18137 v2di __builtin_ia32_vphsubdq (v4si)
18138 v4si __builtin_ia32_vphsubwd (v8hi)
18139 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
18140 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
18141 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
18142 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
18143 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
18144 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
18145 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
18146 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
18147 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
18148 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
18149 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
18150 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
18151 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
18152 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
18153 v4si __builtin_ia32_vprotd (v4si, v4si)
18154 v2di __builtin_ia32_vprotq (v2di, v2di)
18155 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18156 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18157 v4si __builtin_ia32_vpshad (v4si, v4si)
18158 v2di __builtin_ia32_vpshaq (v2di, v2di)
18159 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18160 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18161 v4si __builtin_ia32_vpshld (v4si, v4si)
18162 v2di __builtin_ia32_vpshlq (v2di, v2di)
18163 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18166 The following built-in functions are available when @option{-mfma4} is used.
18167 All of them generate the machine instruction that is part of the name.
18170 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18171 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18172 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18173 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18174 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18175 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18176 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18177 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18178 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18179 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18180 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18181 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18182 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18183 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18184 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18185 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18186 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18187 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18188 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18189 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18190 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18191 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18192 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18193 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18194 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18195 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18196 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18197 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18198 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18199 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18200 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18201 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18205 The following built-in functions are available when @option{-mlwp} is used.
18208 void __builtin_ia32_llwpcb16 (void *);
18209 void __builtin_ia32_llwpcb32 (void *);
18210 void __builtin_ia32_llwpcb64 (void *);
18211 void * __builtin_ia32_llwpcb16 (void);
18212 void * __builtin_ia32_llwpcb32 (void);
18213 void * __builtin_ia32_llwpcb64 (void);
18214 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18215 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18216 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18217 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18218 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18219 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18222 The following built-in functions are available when @option{-mbmi} is used.
18223 All of them generate the machine instruction that is part of the name.
18225 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18226 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18229 The following built-in functions are available when @option{-mbmi2} is used.
18230 All of them generate the machine instruction that is part of the name.
18232 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18233 unsigned int _pdep_u32 (unsigned int, unsigned int)
18234 unsigned int _pext_u32 (unsigned int, unsigned int)
18235 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18236 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18237 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18240 The following built-in functions are available when @option{-mlzcnt} is used.
18241 All of them generate the machine instruction that is part of the name.
18243 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18244 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18245 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18248 The following built-in functions are available when @option{-mfxsr} is used.
18249 All of them generate the machine instruction that is part of the name.
18251 void __builtin_ia32_fxsave (void *)
18252 void __builtin_ia32_fxrstor (void *)
18253 void __builtin_ia32_fxsave64 (void *)
18254 void __builtin_ia32_fxrstor64 (void *)
18257 The following built-in functions are available when @option{-mxsave} is used.
18258 All of them generate the machine instruction that is part of the name.
18260 void __builtin_ia32_xsave (void *, long long)
18261 void __builtin_ia32_xrstor (void *, long long)
18262 void __builtin_ia32_xsave64 (void *, long long)
18263 void __builtin_ia32_xrstor64 (void *, long long)
18266 The following built-in functions are available when @option{-mxsaveopt} is used.
18267 All of them generate the machine instruction that is part of the name.
18269 void __builtin_ia32_xsaveopt (void *, long long)
18270 void __builtin_ia32_xsaveopt64 (void *, long long)
18273 The following built-in functions are available when @option{-mtbm} is used.
18274 Both of them generate the immediate form of the bextr machine instruction.
18276 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18277 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18281 The following built-in functions are available when @option{-m3dnow} is used.
18282 All of them generate the machine instruction that is part of the name.
18285 void __builtin_ia32_femms (void)
18286 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18287 v2si __builtin_ia32_pf2id (v2sf)
18288 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18289 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18290 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18291 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18292 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18293 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18294 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18295 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18296 v2sf __builtin_ia32_pfrcp (v2sf)
18297 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18298 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18299 v2sf __builtin_ia32_pfrsqrt (v2sf)
18300 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18301 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18302 v2sf __builtin_ia32_pi2fd (v2si)
18303 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18306 The following built-in functions are available when both @option{-m3dnow}
18307 and @option{-march=athlon} are used. All of them generate the machine
18308 instruction that is part of the name.
18311 v2si __builtin_ia32_pf2iw (v2sf)
18312 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18313 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18314 v2sf __builtin_ia32_pi2fw (v2si)
18315 v2sf __builtin_ia32_pswapdsf (v2sf)
18316 v2si __builtin_ia32_pswapdsi (v2si)
18319 The following built-in functions are available when @option{-mrtm} is used
18320 They are used for restricted transactional memory. These are the internal
18321 low level functions. Normally the functions in
18322 @ref{x86 transactional memory intrinsics} should be used instead.
18325 int __builtin_ia32_xbegin ()
18326 void __builtin_ia32_xend ()
18327 void __builtin_ia32_xabort (status)
18328 int __builtin_ia32_xtest ()
18331 The following built-in functions are available when @option{-mmwaitx} is used.
18332 All of them generate the machine instruction that is part of the name.
18334 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18335 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18338 The following built-in functions are available when @option{-mclzero} is used.
18339 All of them generate the machine instruction that is part of the name.
18341 void __builtin_i32_clzero (void *)
18344 The following built-in functions are available when @option{-mpku} is used.
18345 They generate reads and writes to PKRU.
18347 void __builtin_ia32_wrpkru (unsigned int)
18348 unsigned int __builtin_ia32_rdpkru ()
18351 @node x86 transactional memory intrinsics
18352 @subsection x86 Transactional Memory Intrinsics
18354 These hardware transactional memory intrinsics for x86 allow you to use
18355 memory transactions with RTM (Restricted Transactional Memory).
18356 This support is enabled with the @option{-mrtm} option.
18357 For using HLE (Hardware Lock Elision) see
18358 @ref{x86 specific memory model extensions for transactional memory} instead.
18360 A memory transaction commits all changes to memory in an atomic way,
18361 as visible to other threads. If the transaction fails it is rolled back
18362 and all side effects discarded.
18364 Generally there is no guarantee that a memory transaction ever succeeds
18365 and suitable fallback code always needs to be supplied.
18367 @deftypefn {RTM Function} {unsigned} _xbegin ()
18368 Start a RTM (Restricted Transactional Memory) transaction.
18369 Returns @code{_XBEGIN_STARTED} when the transaction
18370 started successfully (note this is not 0, so the constant has to be
18371 explicitly tested).
18373 If the transaction aborts, all side-effects
18374 are undone and an abort code encoded as a bit mask is returned.
18375 The following macros are defined:
18378 @item _XABORT_EXPLICIT
18379 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18380 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18381 @item _XABORT_RETRY
18382 Transaction retry is possible.
18383 @item _XABORT_CONFLICT
18384 Transaction abort due to a memory conflict with another thread.
18385 @item _XABORT_CAPACITY
18386 Transaction abort due to the transaction using too much memory.
18387 @item _XABORT_DEBUG
18388 Transaction abort due to a debug trap.
18389 @item _XABORT_NESTED
18390 Transaction abort in an inner nested transaction.
18393 There is no guarantee
18394 any transaction ever succeeds, so there always needs to be a valid
18398 @deftypefn {RTM Function} {void} _xend ()
18399 Commit the current transaction. When no transaction is active this faults.
18400 All memory side-effects of the transaction become visible
18401 to other threads in an atomic manner.
18404 @deftypefn {RTM Function} {int} _xtest ()
18405 Return a nonzero value if a transaction is currently active, otherwise 0.
18408 @deftypefn {RTM Function} {void} _xabort (status)
18409 Abort the current transaction. When no transaction is active this is a no-op.
18410 The @var{status} is an 8-bit constant; its value is encoded in the return
18411 value from @code{_xbegin}.
18414 Here is an example showing handling for @code{_XABORT_RETRY}
18415 and a fallback path for other failures:
18418 #include <immintrin.h>
18420 int n_tries, max_tries;
18421 unsigned status = _XABORT_EXPLICIT;
18424 for (n_tries = 0; n_tries < max_tries; n_tries++)
18426 status = _xbegin ();
18427 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18430 if (status == _XBEGIN_STARTED)
18432 ... transaction code...
18437 ... non-transactional fallback path...
18442 Note that, in most cases, the transactional and non-transactional code
18443 must synchronize together to ensure consistency.
18445 @node Target Format Checks
18446 @section Format Checks Specific to Particular Target Machines
18448 For some target machines, GCC supports additional options to the
18450 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18453 * Solaris Format Checks::
18454 * Darwin Format Checks::
18457 @node Solaris Format Checks
18458 @subsection Solaris Format Checks
18460 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18461 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18462 conversions, and the two-argument @code{%b} conversion for displaying
18463 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18465 @node Darwin Format Checks
18466 @subsection Darwin Format Checks
18468 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18469 attribute context. Declarations made with such attribution are parsed for correct syntax
18470 and format argument types. However, parsing of the format string itself is currently undefined
18471 and is not carried out by this version of the compiler.
18473 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18474 also be used as format arguments. Note that the relevant headers are only likely to be
18475 available on Darwin (OSX) installations. On such installations, the XCode and system
18476 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18477 associated functions.
18480 @section Pragmas Accepted by GCC
18482 @cindex @code{#pragma}
18484 GCC supports several types of pragmas, primarily in order to compile
18485 code originally written for other compilers. Note that in general
18486 we do not recommend the use of pragmas; @xref{Function Attributes},
18487 for further explanation.
18490 * AArch64 Pragmas::
18494 * RS/6000 and PowerPC Pragmas::
18497 * Solaris Pragmas::
18498 * Symbol-Renaming Pragmas::
18499 * Structure-Layout Pragmas::
18501 * Diagnostic Pragmas::
18502 * Visibility Pragmas::
18503 * Push/Pop Macro Pragmas::
18504 * Function Specific Option Pragmas::
18505 * Loop-Specific Pragmas::
18508 @node AArch64 Pragmas
18509 @subsection AArch64 Pragmas
18511 The pragmas defined by the AArch64 target correspond to the AArch64
18512 target function attributes. They can be specified as below:
18514 #pragma GCC target("string")
18517 where @code{@var{string}} can be any string accepted as an AArch64 target
18518 attribute. @xref{AArch64 Function Attributes}, for more details
18519 on the permissible values of @code{string}.
18522 @subsection ARM Pragmas
18524 The ARM target defines pragmas for controlling the default addition of
18525 @code{long_call} and @code{short_call} attributes to functions.
18526 @xref{Function Attributes}, for information about the effects of these
18531 @cindex pragma, long_calls
18532 Set all subsequent functions to have the @code{long_call} attribute.
18534 @item no_long_calls
18535 @cindex pragma, no_long_calls
18536 Set all subsequent functions to have the @code{short_call} attribute.
18538 @item long_calls_off
18539 @cindex pragma, long_calls_off
18540 Do not affect the @code{long_call} or @code{short_call} attributes of
18541 subsequent functions.
18545 @subsection M32C Pragmas
18548 @item GCC memregs @var{number}
18549 @cindex pragma, memregs
18550 Overrides the command-line option @code{-memregs=} for the current
18551 file. Use with care! This pragma must be before any function in the
18552 file, and mixing different memregs values in different objects may
18553 make them incompatible. This pragma is useful when a
18554 performance-critical function uses a memreg for temporary values,
18555 as it may allow you to reduce the number of memregs used.
18557 @item ADDRESS @var{name} @var{address}
18558 @cindex pragma, address
18559 For any declared symbols matching @var{name}, this does three things
18560 to that symbol: it forces the symbol to be located at the given
18561 address (a number), it forces the symbol to be volatile, and it
18562 changes the symbol's scope to be static. This pragma exists for
18563 compatibility with other compilers, but note that the common
18564 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18568 #pragma ADDRESS port3 0x103
18575 @subsection MeP Pragmas
18579 @item custom io_volatile (on|off)
18580 @cindex pragma, custom io_volatile
18581 Overrides the command-line option @code{-mio-volatile} for the current
18582 file. Note that for compatibility with future GCC releases, this
18583 option should only be used once before any @code{io} variables in each
18586 @item GCC coprocessor available @var{registers}
18587 @cindex pragma, coprocessor available
18588 Specifies which coprocessor registers are available to the register
18589 allocator. @var{registers} may be a single register, register range
18590 separated by ellipses, or comma-separated list of those. Example:
18593 #pragma GCC coprocessor available $c0...$c10, $c28
18596 @item GCC coprocessor call_saved @var{registers}
18597 @cindex pragma, coprocessor call_saved
18598 Specifies which coprocessor registers are to be saved and restored by
18599 any function using them. @var{registers} may be a single register,
18600 register range separated by ellipses, or comma-separated list of
18604 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18607 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18608 @cindex pragma, coprocessor subclass
18609 Creates and defines a register class. These register classes can be
18610 used by inline @code{asm} constructs. @var{registers} may be a single
18611 register, register range separated by ellipses, or comma-separated
18612 list of those. Example:
18615 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18617 asm ("cpfoo %0" : "=B" (x));
18620 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18621 @cindex pragma, disinterrupt
18622 For the named functions, the compiler adds code to disable interrupts
18623 for the duration of those functions. If any functions so named
18624 are not encountered in the source, a warning is emitted that the pragma is
18625 not used. Examples:
18628 #pragma disinterrupt foo
18629 #pragma disinterrupt bar, grill
18630 int foo () @{ @dots{} @}
18633 @item GCC call @var{name} , @var{name} @dots{}
18634 @cindex pragma, call
18635 For the named functions, the compiler always uses a register-indirect
18636 call model when calling the named functions. Examples:
18645 @node RS/6000 and PowerPC Pragmas
18646 @subsection RS/6000 and PowerPC Pragmas
18648 The RS/6000 and PowerPC targets define one pragma for controlling
18649 whether or not the @code{longcall} attribute is added to function
18650 declarations by default. This pragma overrides the @option{-mlongcall}
18651 option, but not the @code{longcall} and @code{shortcall} attributes.
18652 @xref{RS/6000 and PowerPC Options}, for more information about when long
18653 calls are and are not necessary.
18657 @cindex pragma, longcall
18658 Apply the @code{longcall} attribute to all subsequent function
18662 Do not apply the @code{longcall} attribute to subsequent function
18666 @c Describe h8300 pragmas here.
18667 @c Describe sh pragmas here.
18668 @c Describe v850 pragmas here.
18670 @node S/390 Pragmas
18671 @subsection S/390 Pragmas
18673 The pragmas defined by the S/390 target correspond to the S/390
18674 target function attributes and some the additional options:
18681 Note that options of the pragma, unlike options of the target
18682 attribute, do change the value of preprocessor macros like
18683 @code{__VEC__}. They can be specified as below:
18686 #pragma GCC target("string[,string]...")
18687 #pragma GCC target("string"[,"string"]...)
18690 @node Darwin Pragmas
18691 @subsection Darwin Pragmas
18693 The following pragmas are available for all architectures running the
18694 Darwin operating system. These are useful for compatibility with other
18698 @item mark @var{tokens}@dots{}
18699 @cindex pragma, mark
18700 This pragma is accepted, but has no effect.
18702 @item options align=@var{alignment}
18703 @cindex pragma, options align
18704 This pragma sets the alignment of fields in structures. The values of
18705 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
18706 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
18707 properly; to restore the previous setting, use @code{reset} for the
18710 @item segment @var{tokens}@dots{}
18711 @cindex pragma, segment
18712 This pragma is accepted, but has no effect.
18714 @item unused (@var{var} [, @var{var}]@dots{})
18715 @cindex pragma, unused
18716 This pragma declares variables to be possibly unused. GCC does not
18717 produce warnings for the listed variables. The effect is similar to
18718 that of the @code{unused} attribute, except that this pragma may appear
18719 anywhere within the variables' scopes.
18722 @node Solaris Pragmas
18723 @subsection Solaris Pragmas
18725 The Solaris target supports @code{#pragma redefine_extname}
18726 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
18727 @code{#pragma} directives for compatibility with the system compiler.
18730 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
18731 @cindex pragma, align
18733 Increase the minimum alignment of each @var{variable} to @var{alignment}.
18734 This is the same as GCC's @code{aligned} attribute @pxref{Variable
18735 Attributes}). Macro expansion occurs on the arguments to this pragma
18736 when compiling C and Objective-C@. It does not currently occur when
18737 compiling C++, but this is a bug which may be fixed in a future
18740 @item fini (@var{function} [, @var{function}]...)
18741 @cindex pragma, fini
18743 This pragma causes each listed @var{function} to be called after
18744 main, or during shared module unloading, by adding a call to the
18745 @code{.fini} section.
18747 @item init (@var{function} [, @var{function}]...)
18748 @cindex pragma, init
18750 This pragma causes each listed @var{function} to be called during
18751 initialization (before @code{main}) or during shared module loading, by
18752 adding a call to the @code{.init} section.
18756 @node Symbol-Renaming Pragmas
18757 @subsection Symbol-Renaming Pragmas
18759 GCC supports a @code{#pragma} directive that changes the name used in
18760 assembly for a given declaration. While this pragma is supported on all
18761 platforms, it is intended primarily to provide compatibility with the
18762 Solaris system headers. This effect can also be achieved using the asm
18763 labels extension (@pxref{Asm Labels}).
18766 @item redefine_extname @var{oldname} @var{newname}
18767 @cindex pragma, redefine_extname
18769 This pragma gives the C function @var{oldname} the assembly symbol
18770 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
18771 is defined if this pragma is available (currently on all platforms).
18774 This pragma and the asm labels extension interact in a complicated
18775 manner. Here are some corner cases you may want to be aware of:
18778 @item This pragma silently applies only to declarations with external
18779 linkage. Asm labels do not have this restriction.
18781 @item In C++, this pragma silently applies only to declarations with
18782 ``C'' linkage. Again, asm labels do not have this restriction.
18784 @item If either of the ways of changing the assembly name of a
18785 declaration are applied to a declaration whose assembly name has
18786 already been determined (either by a previous use of one of these
18787 features, or because the compiler needed the assembly name in order to
18788 generate code), and the new name is different, a warning issues and
18789 the name does not change.
18791 @item The @var{oldname} used by @code{#pragma redefine_extname} is
18792 always the C-language name.
18795 @node Structure-Layout Pragmas
18796 @subsection Structure-Layout Pragmas
18798 For compatibility with Microsoft Windows compilers, GCC supports a
18799 set of @code{#pragma} directives that change the maximum alignment of
18800 members of structures (other than zero-width bit-fields), unions, and
18801 classes subsequently defined. The @var{n} value below always is required
18802 to be a small power of two and specifies the new alignment in bytes.
18805 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
18806 @item @code{#pragma pack()} sets the alignment to the one that was in
18807 effect when compilation started (see also command-line option
18808 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
18809 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
18810 setting on an internal stack and then optionally sets the new alignment.
18811 @item @code{#pragma pack(pop)} restores the alignment setting to the one
18812 saved at the top of the internal stack (and removes that stack entry).
18813 Note that @code{#pragma pack([@var{n}])} does not influence this internal
18814 stack; thus it is possible to have @code{#pragma pack(push)} followed by
18815 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
18816 @code{#pragma pack(pop)}.
18819 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
18820 directive which lays out structures and unions subsequently defined as the
18821 documented @code{__attribute__ ((ms_struct))}.
18824 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
18825 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
18826 @item @code{#pragma ms_struct reset} goes back to the default layout.
18829 Most targets also support the @code{#pragma scalar_storage_order} directive
18830 which lays out structures and unions subsequently defined as the documented
18831 @code{__attribute__ ((scalar_storage_order))}.
18834 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
18835 of the scalar fields to big-endian.
18836 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
18837 of the scalar fields to little-endian.
18838 @item @code{#pragma scalar_storage_order default} goes back to the endianness
18839 that was in effect when compilation started (see also command-line option
18840 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
18844 @subsection Weak Pragmas
18846 For compatibility with SVR4, GCC supports a set of @code{#pragma}
18847 directives for declaring symbols to be weak, and defining weak
18851 @item #pragma weak @var{symbol}
18852 @cindex pragma, weak
18853 This pragma declares @var{symbol} to be weak, as if the declaration
18854 had the attribute of the same name. The pragma may appear before
18855 or after the declaration of @var{symbol}. It is not an error for
18856 @var{symbol} to never be defined at all.
18858 @item #pragma weak @var{symbol1} = @var{symbol2}
18859 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
18860 It is an error if @var{symbol2} is not defined in the current
18864 @node Diagnostic Pragmas
18865 @subsection Diagnostic Pragmas
18867 GCC allows the user to selectively enable or disable certain types of
18868 diagnostics, and change the kind of the diagnostic. For example, a
18869 project's policy might require that all sources compile with
18870 @option{-Werror} but certain files might have exceptions allowing
18871 specific types of warnings. Or, a project might selectively enable
18872 diagnostics and treat them as errors depending on which preprocessor
18873 macros are defined.
18876 @item #pragma GCC diagnostic @var{kind} @var{option}
18877 @cindex pragma, diagnostic
18879 Modifies the disposition of a diagnostic. Note that not all
18880 diagnostics are modifiable; at the moment only warnings (normally
18881 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
18882 Use @option{-fdiagnostics-show-option} to determine which diagnostics
18883 are controllable and which option controls them.
18885 @var{kind} is @samp{error} to treat this diagnostic as an error,
18886 @samp{warning} to treat it like a warning (even if @option{-Werror} is
18887 in effect), or @samp{ignored} if the diagnostic is to be ignored.
18888 @var{option} is a double quoted string that matches the command-line
18892 #pragma GCC diagnostic warning "-Wformat"
18893 #pragma GCC diagnostic error "-Wformat"
18894 #pragma GCC diagnostic ignored "-Wformat"
18897 Note that these pragmas override any command-line options. GCC keeps
18898 track of the location of each pragma, and issues diagnostics according
18899 to the state as of that point in the source file. Thus, pragmas occurring
18900 after a line do not affect diagnostics caused by that line.
18902 @item #pragma GCC diagnostic push
18903 @itemx #pragma GCC diagnostic pop
18905 Causes GCC to remember the state of the diagnostics as of each
18906 @code{push}, and restore to that point at each @code{pop}. If a
18907 @code{pop} has no matching @code{push}, the command-line options are
18911 #pragma GCC diagnostic error "-Wuninitialized"
18912 foo(a); /* error is given for this one */
18913 #pragma GCC diagnostic push
18914 #pragma GCC diagnostic ignored "-Wuninitialized"
18915 foo(b); /* no diagnostic for this one */
18916 #pragma GCC diagnostic pop
18917 foo(c); /* error is given for this one */
18918 #pragma GCC diagnostic pop
18919 foo(d); /* depends on command-line options */
18924 GCC also offers a simple mechanism for printing messages during
18928 @item #pragma message @var{string}
18929 @cindex pragma, diagnostic
18931 Prints @var{string} as a compiler message on compilation. The message
18932 is informational only, and is neither a compilation warning nor an error.
18935 #pragma message "Compiling " __FILE__ "..."
18938 @var{string} may be parenthesized, and is printed with location
18939 information. For example,
18942 #define DO_PRAGMA(x) _Pragma (#x)
18943 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
18945 TODO(Remember to fix this)
18949 prints @samp{/tmp/file.c:4: note: #pragma message:
18950 TODO - Remember to fix this}.
18954 @node Visibility Pragmas
18955 @subsection Visibility Pragmas
18958 @item #pragma GCC visibility push(@var{visibility})
18959 @itemx #pragma GCC visibility pop
18960 @cindex pragma, visibility
18962 This pragma allows the user to set the visibility for multiple
18963 declarations without having to give each a visibility attribute
18964 (@pxref{Function Attributes}).
18966 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
18967 declarations. Class members and template specializations are not
18968 affected; if you want to override the visibility for a particular
18969 member or instantiation, you must use an attribute.
18974 @node Push/Pop Macro Pragmas
18975 @subsection Push/Pop Macro Pragmas
18977 For compatibility with Microsoft Windows compilers, GCC supports
18978 @samp{#pragma push_macro(@var{"macro_name"})}
18979 and @samp{#pragma pop_macro(@var{"macro_name"})}.
18982 @item #pragma push_macro(@var{"macro_name"})
18983 @cindex pragma, push_macro
18984 This pragma saves the value of the macro named as @var{macro_name} to
18985 the top of the stack for this macro.
18987 @item #pragma pop_macro(@var{"macro_name"})
18988 @cindex pragma, pop_macro
18989 This pragma sets the value of the macro named as @var{macro_name} to
18990 the value on top of the stack for this macro. If the stack for
18991 @var{macro_name} is empty, the value of the macro remains unchanged.
18998 #pragma push_macro("X")
19001 #pragma pop_macro("X")
19006 In this example, the definition of X as 1 is saved by @code{#pragma
19007 push_macro} and restored by @code{#pragma pop_macro}.
19009 @node Function Specific Option Pragmas
19010 @subsection Function Specific Option Pragmas
19013 @item #pragma GCC target (@var{"string"}...)
19014 @cindex pragma GCC target
19016 This pragma allows you to set target specific options for functions
19017 defined later in the source file. One or more strings can be
19018 specified. Each function that is defined after this point is as
19019 if @code{attribute((target("STRING")))} was specified for that
19020 function. The parenthesis around the options is optional.
19021 @xref{Function Attributes}, for more information about the
19022 @code{target} attribute and the attribute syntax.
19024 The @code{#pragma GCC target} pragma is presently implemented for
19025 x86, PowerPC, and Nios II targets only.
19029 @item #pragma GCC optimize (@var{"string"}...)
19030 @cindex pragma GCC optimize
19032 This pragma allows you to set global optimization options for functions
19033 defined later in the source file. One or more strings can be
19034 specified. Each function that is defined after this point is as
19035 if @code{attribute((optimize("STRING")))} was specified for that
19036 function. The parenthesis around the options is optional.
19037 @xref{Function Attributes}, for more information about the
19038 @code{optimize} attribute and the attribute syntax.
19042 @item #pragma GCC push_options
19043 @itemx #pragma GCC pop_options
19044 @cindex pragma GCC push_options
19045 @cindex pragma GCC pop_options
19047 These pragmas maintain a stack of the current target and optimization
19048 options. It is intended for include files where you temporarily want
19049 to switch to using a different @samp{#pragma GCC target} or
19050 @samp{#pragma GCC optimize} and then to pop back to the previous
19055 @item #pragma GCC reset_options
19056 @cindex pragma GCC reset_options
19058 This pragma clears the current @code{#pragma GCC target} and
19059 @code{#pragma GCC optimize} to use the default switches as specified
19060 on the command line.
19063 @node Loop-Specific Pragmas
19064 @subsection Loop-Specific Pragmas
19067 @item #pragma GCC ivdep
19068 @cindex pragma GCC ivdep
19071 With this pragma, the programmer asserts that there are no loop-carried
19072 dependencies which would prevent consecutive iterations of
19073 the following loop from executing concurrently with SIMD
19074 (single instruction multiple data) instructions.
19076 For example, the compiler can only unconditionally vectorize the following
19077 loop with the pragma:
19080 void foo (int n, int *a, int *b, int *c)
19084 for (i = 0; i < n; ++i)
19085 a[i] = b[i] + c[i];
19090 In this example, using the @code{restrict} qualifier had the same
19091 effect. In the following example, that would not be possible. Assume
19092 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
19093 that it can unconditionally vectorize the following loop:
19096 void ignore_vec_dep (int *a, int k, int c, int m)
19099 for (int i = 0; i < m; i++)
19100 a[i] = a[i + k] * c;
19105 @node Unnamed Fields
19106 @section Unnamed Structure and Union Fields
19107 @cindex @code{struct}
19108 @cindex @code{union}
19110 As permitted by ISO C11 and for compatibility with other compilers,
19111 GCC allows you to define
19112 a structure or union that contains, as fields, structures and unions
19113 without names. For example:
19127 In this example, you are able to access members of the unnamed
19128 union with code like @samp{foo.b}. Note that only unnamed structs and
19129 unions are allowed, you may not have, for example, an unnamed
19132 You must never create such structures that cause ambiguous field definitions.
19133 For example, in this structure:
19145 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
19146 The compiler gives errors for such constructs.
19148 @opindex fms-extensions
19149 Unless @option{-fms-extensions} is used, the unnamed field must be a
19150 structure or union definition without a tag (for example, @samp{struct
19151 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
19152 also be a definition with a tag such as @samp{struct foo @{ int a;
19153 @};}, a reference to a previously defined structure or union such as
19154 @samp{struct foo;}, or a reference to a @code{typedef} name for a
19155 previously defined structure or union type.
19157 @opindex fplan9-extensions
19158 The option @option{-fplan9-extensions} enables
19159 @option{-fms-extensions} as well as two other extensions. First, a
19160 pointer to a structure is automatically converted to a pointer to an
19161 anonymous field for assignments and function calls. For example:
19164 struct s1 @{ int a; @};
19165 struct s2 @{ struct s1; @};
19166 extern void f1 (struct s1 *);
19167 void f2 (struct s2 *p) @{ f1 (p); @}
19171 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
19172 converted into a pointer to the anonymous field.
19174 Second, when the type of an anonymous field is a @code{typedef} for a
19175 @code{struct} or @code{union}, code may refer to the field using the
19176 name of the @code{typedef}.
19179 typedef struct @{ int a; @} s1;
19180 struct s2 @{ s1; @};
19181 s1 f1 (struct s2 *p) @{ return p->s1; @}
19184 These usages are only permitted when they are not ambiguous.
19187 @section Thread-Local Storage
19188 @cindex Thread-Local Storage
19189 @cindex @acronym{TLS}
19190 @cindex @code{__thread}
19192 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
19193 are allocated such that there is one instance of the variable per extant
19194 thread. The runtime model GCC uses to implement this originates
19195 in the IA-64 processor-specific ABI, but has since been migrated
19196 to other processors as well. It requires significant support from
19197 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
19198 system libraries (@file{libc.so} and @file{libpthread.so}), so it
19199 is not available everywhere.
19201 At the user level, the extension is visible with a new storage
19202 class keyword: @code{__thread}. For example:
19206 extern __thread struct state s;
19207 static __thread char *p;
19210 The @code{__thread} specifier may be used alone, with the @code{extern}
19211 or @code{static} specifiers, but with no other storage class specifier.
19212 When used with @code{extern} or @code{static}, @code{__thread} must appear
19213 immediately after the other storage class specifier.
19215 The @code{__thread} specifier may be applied to any global, file-scoped
19216 static, function-scoped static, or static data member of a class. It may
19217 not be applied to block-scoped automatic or non-static data member.
19219 When the address-of operator is applied to a thread-local variable, it is
19220 evaluated at run time and returns the address of the current thread's
19221 instance of that variable. An address so obtained may be used by any
19222 thread. When a thread terminates, any pointers to thread-local variables
19223 in that thread become invalid.
19225 No static initialization may refer to the address of a thread-local variable.
19227 In C++, if an initializer is present for a thread-local variable, it must
19228 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19231 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19232 ELF Handling For Thread-Local Storage} for a detailed explanation of
19233 the four thread-local storage addressing models, and how the runtime
19234 is expected to function.
19237 * C99 Thread-Local Edits::
19238 * C++98 Thread-Local Edits::
19241 @node C99 Thread-Local Edits
19242 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19244 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19245 that document the exact semantics of the language extension.
19249 @cite{5.1.2 Execution environments}
19251 Add new text after paragraph 1
19254 Within either execution environment, a @dfn{thread} is a flow of
19255 control within a program. It is implementation defined whether
19256 or not there may be more than one thread associated with a program.
19257 It is implementation defined how threads beyond the first are
19258 created, the name and type of the function called at thread
19259 startup, and how threads may be terminated. However, objects
19260 with thread storage duration shall be initialized before thread
19265 @cite{6.2.4 Storage durations of objects}
19267 Add new text before paragraph 3
19270 An object whose identifier is declared with the storage-class
19271 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19272 Its lifetime is the entire execution of the thread, and its
19273 stored value is initialized only once, prior to thread startup.
19277 @cite{6.4.1 Keywords}
19279 Add @code{__thread}.
19282 @cite{6.7.1 Storage-class specifiers}
19284 Add @code{__thread} to the list of storage class specifiers in
19287 Change paragraph 2 to
19290 With the exception of @code{__thread}, at most one storage-class
19291 specifier may be given [@dots{}]. The @code{__thread} specifier may
19292 be used alone, or immediately following @code{extern} or
19296 Add new text after paragraph 6
19299 The declaration of an identifier for a variable that has
19300 block scope that specifies @code{__thread} shall also
19301 specify either @code{extern} or @code{static}.
19303 The @code{__thread} specifier shall be used only with
19308 @node C++98 Thread-Local Edits
19309 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19311 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19312 that document the exact semantics of the language extension.
19316 @b{[intro.execution]}
19318 New text after paragraph 4
19321 A @dfn{thread} is a flow of control within the abstract machine.
19322 It is implementation defined whether or not there may be more than
19326 New text after paragraph 7
19329 It is unspecified whether additional action must be taken to
19330 ensure when and whether side effects are visible to other threads.
19336 Add @code{__thread}.
19339 @b{[basic.start.main]}
19341 Add after paragraph 5
19344 The thread that begins execution at the @code{main} function is called
19345 the @dfn{main thread}. It is implementation defined how functions
19346 beginning threads other than the main thread are designated or typed.
19347 A function so designated, as well as the @code{main} function, is called
19348 a @dfn{thread startup function}. It is implementation defined what
19349 happens if a thread startup function returns. It is implementation
19350 defined what happens to other threads when any thread calls @code{exit}.
19354 @b{[basic.start.init]}
19356 Add after paragraph 4
19359 The storage for an object of thread storage duration shall be
19360 statically initialized before the first statement of the thread startup
19361 function. An object of thread storage duration shall not require
19362 dynamic initialization.
19366 @b{[basic.start.term]}
19368 Add after paragraph 3
19371 The type of an object with thread storage duration shall not have a
19372 non-trivial destructor, nor shall it be an array type whose elements
19373 (directly or indirectly) have non-trivial destructors.
19379 Add ``thread storage duration'' to the list in paragraph 1.
19384 Thread, static, and automatic storage durations are associated with
19385 objects introduced by declarations [@dots{}].
19388 Add @code{__thread} to the list of specifiers in paragraph 3.
19391 @b{[basic.stc.thread]}
19393 New section before @b{[basic.stc.static]}
19396 The keyword @code{__thread} applied to a non-local object gives the
19397 object thread storage duration.
19399 A local variable or class data member declared both @code{static}
19400 and @code{__thread} gives the variable or member thread storage
19405 @b{[basic.stc.static]}
19410 All objects that have neither thread storage duration, dynamic
19411 storage duration nor are local [@dots{}].
19417 Add @code{__thread} to the list in paragraph 1.
19422 With the exception of @code{__thread}, at most one
19423 @var{storage-class-specifier} shall appear in a given
19424 @var{decl-specifier-seq}. The @code{__thread} specifier may
19425 be used alone, or immediately following the @code{extern} or
19426 @code{static} specifiers. [@dots{}]
19429 Add after paragraph 5
19432 The @code{__thread} specifier can be applied only to the names of objects
19433 and to anonymous unions.
19439 Add after paragraph 6
19442 Non-@code{static} members shall not be @code{__thread}.
19446 @node Binary constants
19447 @section Binary Constants using the @samp{0b} Prefix
19448 @cindex Binary constants using the @samp{0b} prefix
19450 Integer constants can be written as binary constants, consisting of a
19451 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19452 @samp{0B}. This is particularly useful in environments that operate a
19453 lot on the bit level (like microcontrollers).
19455 The following statements are identical:
19464 The type of these constants follows the same rules as for octal or
19465 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19468 @node C++ Extensions
19469 @chapter Extensions to the C++ Language
19470 @cindex extensions, C++ language
19471 @cindex C++ language extensions
19473 The GNU compiler provides these extensions to the C++ language (and you
19474 can also use most of the C language extensions in your C++ programs). If you
19475 want to write code that checks whether these features are available, you can
19476 test for the GNU compiler the same way as for C programs: check for a
19477 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19478 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19479 Predefined Macros,cpp,The GNU C Preprocessor}).
19482 * C++ Volatiles:: What constitutes an access to a volatile object.
19483 * Restricted Pointers:: C99 restricted pointers and references.
19484 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19485 * C++ Interface:: You can use a single C++ header file for both
19486 declarations and definitions.
19487 * Template Instantiation:: Methods for ensuring that exactly one copy of
19488 each needed template instantiation is emitted.
19489 * Bound member functions:: You can extract a function pointer to the
19490 method denoted by a @samp{->*} or @samp{.*} expression.
19491 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19492 * Function Multiversioning:: Declaring multiple function versions.
19493 * Namespace Association:: Strong using-directives for namespace association.
19494 * Type Traits:: Compiler support for type traits.
19495 * C++ Concepts:: Improved support for generic programming.
19496 * Java Exceptions:: Tweaking exception handling to work with Java.
19497 * Deprecated Features:: Things will disappear from G++.
19498 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19501 @node C++ Volatiles
19502 @section When is a Volatile C++ Object Accessed?
19503 @cindex accessing volatiles
19504 @cindex volatile read
19505 @cindex volatile write
19506 @cindex volatile access
19508 The C++ standard differs from the C standard in its treatment of
19509 volatile objects. It fails to specify what constitutes a volatile
19510 access, except to say that C++ should behave in a similar manner to C
19511 with respect to volatiles, where possible. However, the different
19512 lvalueness of expressions between C and C++ complicate the behavior.
19513 G++ behaves the same as GCC for volatile access, @xref{C
19514 Extensions,,Volatiles}, for a description of GCC's behavior.
19516 The C and C++ language specifications differ when an object is
19517 accessed in a void context:
19520 volatile int *src = @var{somevalue};
19524 The C++ standard specifies that such expressions do not undergo lvalue
19525 to rvalue conversion, and that the type of the dereferenced object may
19526 be incomplete. The C++ standard does not specify explicitly that it
19527 is lvalue to rvalue conversion that is responsible for causing an
19528 access. There is reason to believe that it is, because otherwise
19529 certain simple expressions become undefined. However, because it
19530 would surprise most programmers, G++ treats dereferencing a pointer to
19531 volatile object of complete type as GCC would do for an equivalent
19532 type in C@. When the object has incomplete type, G++ issues a
19533 warning; if you wish to force an error, you must force a conversion to
19534 rvalue with, for instance, a static cast.
19536 When using a reference to volatile, G++ does not treat equivalent
19537 expressions as accesses to volatiles, but instead issues a warning that
19538 no volatile is accessed. The rationale for this is that otherwise it
19539 becomes difficult to determine where volatile access occur, and not
19540 possible to ignore the return value from functions returning volatile
19541 references. Again, if you wish to force a read, cast the reference to
19544 G++ implements the same behavior as GCC does when assigning to a
19545 volatile object---there is no reread of the assigned-to object, the
19546 assigned rvalue is reused. Note that in C++ assignment expressions
19547 are lvalues, and if used as an lvalue, the volatile object is
19548 referred to. For instance, @var{vref} refers to @var{vobj}, as
19549 expected, in the following example:
19553 volatile int &vref = vobj = @var{something};
19556 @node Restricted Pointers
19557 @section Restricting Pointer Aliasing
19558 @cindex restricted pointers
19559 @cindex restricted references
19560 @cindex restricted this pointer
19562 As with the C front end, G++ understands the C99 feature of restricted pointers,
19563 specified with the @code{__restrict__}, or @code{__restrict} type
19564 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19565 language flag, @code{restrict} is not a keyword in C++.
19567 In addition to allowing restricted pointers, you can specify restricted
19568 references, which indicate that the reference is not aliased in the local
19572 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19579 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19580 @var{rref} refers to a (different) unaliased integer.
19582 You may also specify whether a member function's @var{this} pointer is
19583 unaliased by using @code{__restrict__} as a member function qualifier.
19586 void T::fn () __restrict__
19593 Within the body of @code{T::fn}, @var{this} has the effective
19594 definition @code{T *__restrict__ const this}. Notice that the
19595 interpretation of a @code{__restrict__} member function qualifier is
19596 different to that of @code{const} or @code{volatile} qualifier, in that it
19597 is applied to the pointer rather than the object. This is consistent with
19598 other compilers that implement restricted pointers.
19600 As with all outermost parameter qualifiers, @code{__restrict__} is
19601 ignored in function definition matching. This means you only need to
19602 specify @code{__restrict__} in a function definition, rather than
19603 in a function prototype as well.
19605 @node Vague Linkage
19606 @section Vague Linkage
19607 @cindex vague linkage
19609 There are several constructs in C++ that require space in the object
19610 file but are not clearly tied to a single translation unit. We say that
19611 these constructs have ``vague linkage''. Typically such constructs are
19612 emitted wherever they are needed, though sometimes we can be more
19616 @item Inline Functions
19617 Inline functions are typically defined in a header file which can be
19618 included in many different compilations. Hopefully they can usually be
19619 inlined, but sometimes an out-of-line copy is necessary, if the address
19620 of the function is taken or if inlining fails. In general, we emit an
19621 out-of-line copy in all translation units where one is needed. As an
19622 exception, we only emit inline virtual functions with the vtable, since
19623 it always requires a copy.
19625 Local static variables and string constants used in an inline function
19626 are also considered to have vague linkage, since they must be shared
19627 between all inlined and out-of-line instances of the function.
19631 C++ virtual functions are implemented in most compilers using a lookup
19632 table, known as a vtable. The vtable contains pointers to the virtual
19633 functions provided by a class, and each object of the class contains a
19634 pointer to its vtable (or vtables, in some multiple-inheritance
19635 situations). If the class declares any non-inline, non-pure virtual
19636 functions, the first one is chosen as the ``key method'' for the class,
19637 and the vtable is only emitted in the translation unit where the key
19640 @emph{Note:} If the chosen key method is later defined as inline, the
19641 vtable is still emitted in every translation unit that defines it.
19642 Make sure that any inline virtuals are declared inline in the class
19643 body, even if they are not defined there.
19645 @item @code{type_info} objects
19646 @cindex @code{type_info}
19648 C++ requires information about types to be written out in order to
19649 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19650 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19651 object is written out along with the vtable so that @samp{dynamic_cast}
19652 can determine the dynamic type of a class object at run time. For all
19653 other types, we write out the @samp{type_info} object when it is used: when
19654 applying @samp{typeid} to an expression, throwing an object, or
19655 referring to a type in a catch clause or exception specification.
19657 @item Template Instantiations
19658 Most everything in this section also applies to template instantiations,
19659 but there are other options as well.
19660 @xref{Template Instantiation,,Where's the Template?}.
19664 When used with GNU ld version 2.8 or later on an ELF system such as
19665 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
19666 these constructs will be discarded at link time. This is known as
19669 On targets that don't support COMDAT, but do support weak symbols, GCC
19670 uses them. This way one copy overrides all the others, but
19671 the unused copies still take up space in the executable.
19673 For targets that do not support either COMDAT or weak symbols,
19674 most entities with vague linkage are emitted as local symbols to
19675 avoid duplicate definition errors from the linker. This does not happen
19676 for local statics in inlines, however, as having multiple copies
19677 almost certainly breaks things.
19679 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
19680 another way to control placement of these constructs.
19682 @node C++ Interface
19683 @section C++ Interface and Implementation Pragmas
19685 @cindex interface and implementation headers, C++
19686 @cindex C++ interface and implementation headers
19687 @cindex pragmas, interface and implementation
19689 @code{#pragma interface} and @code{#pragma implementation} provide the
19690 user with a way of explicitly directing the compiler to emit entities
19691 with vague linkage (and debugging information) in a particular
19694 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
19695 by COMDAT support and the ``key method'' heuristic
19696 mentioned in @ref{Vague Linkage}. Using them can actually cause your
19697 program to grow due to unnecessary out-of-line copies of inline
19701 @item #pragma interface
19702 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
19703 @kindex #pragma interface
19704 Use this directive in @emph{header files} that define object classes, to save
19705 space in most of the object files that use those classes. Normally,
19706 local copies of certain information (backup copies of inline member
19707 functions, debugging information, and the internal tables that implement
19708 virtual functions) must be kept in each object file that includes class
19709 definitions. You can use this pragma to avoid such duplication. When a
19710 header file containing @samp{#pragma interface} is included in a
19711 compilation, this auxiliary information is not generated (unless
19712 the main input source file itself uses @samp{#pragma implementation}).
19713 Instead, the object files contain references to be resolved at link
19716 The second form of this directive is useful for the case where you have
19717 multiple headers with the same name in different directories. If you
19718 use this form, you must specify the same string to @samp{#pragma
19721 @item #pragma implementation
19722 @itemx #pragma implementation "@var{objects}.h"
19723 @kindex #pragma implementation
19724 Use this pragma in a @emph{main input file}, when you want full output from
19725 included header files to be generated (and made globally visible). The
19726 included header file, in turn, should use @samp{#pragma interface}.
19727 Backup copies of inline member functions, debugging information, and the
19728 internal tables used to implement virtual functions are all generated in
19729 implementation files.
19731 @cindex implied @code{#pragma implementation}
19732 @cindex @code{#pragma implementation}, implied
19733 @cindex naming convention, implementation headers
19734 If you use @samp{#pragma implementation} with no argument, it applies to
19735 an include file with the same basename@footnote{A file's @dfn{basename}
19736 is the name stripped of all leading path information and of trailing
19737 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
19738 file. For example, in @file{allclass.cc}, giving just
19739 @samp{#pragma implementation}
19740 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
19742 Use the string argument if you want a single implementation file to
19743 include code from multiple header files. (You must also use
19744 @samp{#include} to include the header file; @samp{#pragma
19745 implementation} only specifies how to use the file---it doesn't actually
19748 There is no way to split up the contents of a single header file into
19749 multiple implementation files.
19752 @cindex inlining and C++ pragmas
19753 @cindex C++ pragmas, effect on inlining
19754 @cindex pragmas in C++, effect on inlining
19755 @samp{#pragma implementation} and @samp{#pragma interface} also have an
19756 effect on function inlining.
19758 If you define a class in a header file marked with @samp{#pragma
19759 interface}, the effect on an inline function defined in that class is
19760 similar to an explicit @code{extern} declaration---the compiler emits
19761 no code at all to define an independent version of the function. Its
19762 definition is used only for inlining with its callers.
19764 @opindex fno-implement-inlines
19765 Conversely, when you include the same header file in a main source file
19766 that declares it as @samp{#pragma implementation}, the compiler emits
19767 code for the function itself; this defines a version of the function
19768 that can be found via pointers (or by callers compiled without
19769 inlining). If all calls to the function can be inlined, you can avoid
19770 emitting the function by compiling with @option{-fno-implement-inlines}.
19771 If any calls are not inlined, you will get linker errors.
19773 @node Template Instantiation
19774 @section Where's the Template?
19775 @cindex template instantiation
19777 C++ templates were the first language feature to require more
19778 intelligence from the environment than was traditionally found on a UNIX
19779 system. Somehow the compiler and linker have to make sure that each
19780 template instance occurs exactly once in the executable if it is needed,
19781 and not at all otherwise. There are two basic approaches to this
19782 problem, which are referred to as the Borland model and the Cfront model.
19785 @item Borland model
19786 Borland C++ solved the template instantiation problem by adding the code
19787 equivalent of common blocks to their linker; the compiler emits template
19788 instances in each translation unit that uses them, and the linker
19789 collapses them together. The advantage of this model is that the linker
19790 only has to consider the object files themselves; there is no external
19791 complexity to worry about. The disadvantage is that compilation time
19792 is increased because the template code is being compiled repeatedly.
19793 Code written for this model tends to include definitions of all
19794 templates in the header file, since they must be seen to be
19798 The AT&T C++ translator, Cfront, solved the template instantiation
19799 problem by creating the notion of a template repository, an
19800 automatically maintained place where template instances are stored. A
19801 more modern version of the repository works as follows: As individual
19802 object files are built, the compiler places any template definitions and
19803 instantiations encountered in the repository. At link time, the link
19804 wrapper adds in the objects in the repository and compiles any needed
19805 instances that were not previously emitted. The advantages of this
19806 model are more optimal compilation speed and the ability to use the
19807 system linker; to implement the Borland model a compiler vendor also
19808 needs to replace the linker. The disadvantages are vastly increased
19809 complexity, and thus potential for error; for some code this can be
19810 just as transparent, but in practice it can been very difficult to build
19811 multiple programs in one directory and one program in multiple
19812 directories. Code written for this model tends to separate definitions
19813 of non-inline member templates into a separate file, which should be
19814 compiled separately.
19817 G++ implements the Borland model on targets where the linker supports it,
19818 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
19819 Otherwise G++ implements neither automatic model.
19821 You have the following options for dealing with template instantiations:
19825 Do nothing. Code written for the Borland model works fine, but
19826 each translation unit contains instances of each of the templates it
19827 uses. The duplicate instances will be discarded by the linker, but in
19828 a large program, this can lead to an unacceptable amount of code
19829 duplication in object files or shared libraries.
19831 Duplicate instances of a template can be avoided by defining an explicit
19832 instantiation in one object file, and preventing the compiler from doing
19833 implicit instantiations in any other object files by using an explicit
19834 instantiation declaration, using the @code{extern template} syntax:
19837 extern template int max (int, int);
19840 This syntax is defined in the C++ 2011 standard, but has been supported by
19841 G++ and other compilers since well before 2011.
19843 Explicit instantiations can be used for the largest or most frequently
19844 duplicated instances, without having to know exactly which other instances
19845 are used in the rest of the program. You can scatter the explicit
19846 instantiations throughout your program, perhaps putting them in the
19847 translation units where the instances are used or the translation units
19848 that define the templates themselves; you can put all of the explicit
19849 instantiations you need into one big file; or you can create small files
19856 template class Foo<int>;
19857 template ostream& operator <<
19858 (ostream&, const Foo<int>&);
19862 for each of the instances you need, and create a template instantiation
19863 library from those.
19865 This is the simplest option, but also offers flexibility and
19866 fine-grained control when necessary. It is also the most portable
19867 alternative and programs using this approach will work with most modern
19872 Compile your template-using code with @option{-frepo}. The compiler
19873 generates files with the extension @samp{.rpo} listing all of the
19874 template instantiations used in the corresponding object files that
19875 could be instantiated there; the link wrapper, @samp{collect2},
19876 then updates the @samp{.rpo} files to tell the compiler where to place
19877 those instantiations and rebuild any affected object files. The
19878 link-time overhead is negligible after the first pass, as the compiler
19879 continues to place the instantiations in the same files.
19881 This can be a suitable option for application code written for the Borland
19882 model, as it usually just works. Code written for the Cfront model
19883 needs to be modified so that the template definitions are available at
19884 one or more points of instantiation; usually this is as simple as adding
19885 @code{#include <tmethods.cc>} to the end of each template header.
19887 For library code, if you want the library to provide all of the template
19888 instantiations it needs, just try to link all of its object files
19889 together; the link will fail, but cause the instantiations to be
19890 generated as a side effect. Be warned, however, that this may cause
19891 conflicts if multiple libraries try to provide the same instantiations.
19892 For greater control, use explicit instantiation as described in the next
19896 @opindex fno-implicit-templates
19897 Compile your code with @option{-fno-implicit-templates} to disable the
19898 implicit generation of template instances, and explicitly instantiate
19899 all the ones you use. This approach requires more knowledge of exactly
19900 which instances you need than do the others, but it's less
19901 mysterious and allows greater control if you want to ensure that only
19902 the intended instances are used.
19904 If you are using Cfront-model code, you can probably get away with not
19905 using @option{-fno-implicit-templates} when compiling files that don't
19906 @samp{#include} the member template definitions.
19908 If you use one big file to do the instantiations, you may want to
19909 compile it without @option{-fno-implicit-templates} so you get all of the
19910 instances required by your explicit instantiations (but not by any
19911 other files) without having to specify them as well.
19913 In addition to forward declaration of explicit instantiations
19914 (with @code{extern}), G++ has extended the template instantiation
19915 syntax to support instantiation of the compiler support data for a
19916 template class (i.e.@: the vtable) without instantiating any of its
19917 members (with @code{inline}), and instantiation of only the static data
19918 members of a template class, without the support data or member
19919 functions (with @code{static}):
19922 inline template class Foo<int>;
19923 static template class Foo<int>;
19927 @node Bound member functions
19928 @section Extracting the Function Pointer from a Bound Pointer to Member Function
19930 @cindex pointer to member function
19931 @cindex bound pointer to member function
19933 In C++, pointer to member functions (PMFs) are implemented using a wide
19934 pointer of sorts to handle all the possible call mechanisms; the PMF
19935 needs to store information about how to adjust the @samp{this} pointer,
19936 and if the function pointed to is virtual, where to find the vtable, and
19937 where in the vtable to look for the member function. If you are using
19938 PMFs in an inner loop, you should really reconsider that decision. If
19939 that is not an option, you can extract the pointer to the function that
19940 would be called for a given object/PMF pair and call it directly inside
19941 the inner loop, to save a bit of time.
19943 Note that you still pay the penalty for the call through a
19944 function pointer; on most modern architectures, such a call defeats the
19945 branch prediction features of the CPU@. This is also true of normal
19946 virtual function calls.
19948 The syntax for this extension is
19952 extern int (A::*fp)();
19953 typedef int (*fptr)(A *);
19955 fptr p = (fptr)(a.*fp);
19958 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
19959 no object is needed to obtain the address of the function. They can be
19960 converted to function pointers directly:
19963 fptr p1 = (fptr)(&A::foo);
19966 @opindex Wno-pmf-conversions
19967 You must specify @option{-Wno-pmf-conversions} to use this extension.
19969 @node C++ Attributes
19970 @section C++-Specific Variable, Function, and Type Attributes
19972 Some attributes only make sense for C++ programs.
19975 @item abi_tag ("@var{tag}", ...)
19976 @cindex @code{abi_tag} function attribute
19977 @cindex @code{abi_tag} variable attribute
19978 @cindex @code{abi_tag} type attribute
19979 The @code{abi_tag} attribute can be applied to a function, variable, or class
19980 declaration. It modifies the mangled name of the entity to
19981 incorporate the tag name, in order to distinguish the function or
19982 class from an earlier version with a different ABI; perhaps the class
19983 has changed size, or the function has a different return type that is
19984 not encoded in the mangled name.
19986 The attribute can also be applied to an inline namespace, but does not
19987 affect the mangled name of the namespace; in this case it is only used
19988 for @option{-Wabi-tag} warnings and automatic tagging of functions and
19989 variables. Tagging inline namespaces is generally preferable to
19990 tagging individual declarations, but the latter is sometimes
19991 necessary, such as when only certain members of a class need to be
19994 The argument can be a list of strings of arbitrary length. The
19995 strings are sorted on output, so the order of the list is
19998 A redeclaration of an entity must not add new ABI tags,
19999 since doing so would change the mangled name.
20001 The ABI tags apply to a name, so all instantiations and
20002 specializations of a template have the same tags. The attribute will
20003 be ignored if applied to an explicit specialization or instantiation.
20005 The @option{-Wabi-tag} flag enables a warning about a class which does
20006 not have all the ABI tags used by its subobjects and virtual functions; for users with code
20007 that needs to coexist with an earlier ABI, using this option can help
20008 to find all affected types that need to be tagged.
20010 When a type involving an ABI tag is used as the type of a variable or
20011 return type of a function where that tag is not already present in the
20012 signature of the function, the tag is automatically applied to the
20013 variable or function. @option{-Wabi-tag} also warns about this
20014 situation; this warning can be avoided by explicitly tagging the
20015 variable or function or moving it into a tagged inline namespace.
20017 @item init_priority (@var{priority})
20018 @cindex @code{init_priority} variable attribute
20020 In Standard C++, objects defined at namespace scope are guaranteed to be
20021 initialized in an order in strict accordance with that of their definitions
20022 @emph{in a given translation unit}. No guarantee is made for initializations
20023 across translation units. However, GNU C++ allows users to control the
20024 order of initialization of objects defined at namespace scope with the
20025 @code{init_priority} attribute by specifying a relative @var{priority},
20026 a constant integral expression currently bounded between 101 and 65535
20027 inclusive. Lower numbers indicate a higher priority.
20029 In the following example, @code{A} would normally be created before
20030 @code{B}, but the @code{init_priority} attribute reverses that order:
20033 Some_Class A __attribute__ ((init_priority (2000)));
20034 Some_Class B __attribute__ ((init_priority (543)));
20038 Note that the particular values of @var{priority} do not matter; only their
20041 @item java_interface
20042 @cindex @code{java_interface} type attribute
20044 This type attribute informs C++ that the class is a Java interface. It may
20045 only be applied to classes declared within an @code{extern "Java"} block.
20046 Calls to methods declared in this interface are dispatched using GCJ's
20047 interface table mechanism, instead of regular virtual table dispatch.
20050 @cindex @code{warn_unused} type attribute
20052 For C++ types with non-trivial constructors and/or destructors it is
20053 impossible for the compiler to determine whether a variable of this
20054 type is truly unused if it is not referenced. This type attribute
20055 informs the compiler that variables of this type should be warned
20056 about if they appear to be unused, just like variables of fundamental
20059 This attribute is appropriate for types which just represent a value,
20060 such as @code{std::string}; it is not appropriate for types which
20061 control a resource, such as @code{std::mutex}.
20063 This attribute is also accepted in C, but it is unnecessary because C
20064 does not have constructors or destructors.
20068 See also @ref{Namespace Association}.
20070 @node Function Multiversioning
20071 @section Function Multiversioning
20072 @cindex function versions
20074 With the GNU C++ front end, for x86 targets, you may specify multiple
20075 versions of a function, where each function is specialized for a
20076 specific target feature. At runtime, the appropriate version of the
20077 function is automatically executed depending on the characteristics of
20078 the execution platform. Here is an example.
20081 __attribute__ ((target ("default")))
20084 // The default version of foo.
20088 __attribute__ ((target ("sse4.2")))
20091 // foo version for SSE4.2
20095 __attribute__ ((target ("arch=atom")))
20098 // foo version for the Intel ATOM processor
20102 __attribute__ ((target ("arch=amdfam10")))
20105 // foo version for the AMD Family 0x10 processors.
20112 assert ((*p) () == foo ());
20117 In the above example, four versions of function foo are created. The
20118 first version of foo with the target attribute "default" is the default
20119 version. This version gets executed when no other target specific
20120 version qualifies for execution on a particular platform. A new version
20121 of foo is created by using the same function signature but with a
20122 different target string. Function foo is called or a pointer to it is
20123 taken just like a regular function. GCC takes care of doing the
20124 dispatching to call the right version at runtime. Refer to the
20125 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
20126 Function Multiversioning} for more details.
20128 @node Namespace Association
20129 @section Namespace Association
20131 @strong{Caution:} The semantics of this extension are equivalent
20132 to C++ 2011 inline namespaces. Users should use inline namespaces
20133 instead as this extension will be removed in future versions of G++.
20135 A using-directive with @code{__attribute ((strong))} is stronger
20136 than a normal using-directive in two ways:
20140 Templates from the used namespace can be specialized and explicitly
20141 instantiated as though they were members of the using namespace.
20144 The using namespace is considered an associated namespace of all
20145 templates in the used namespace for purposes of argument-dependent
20149 The used namespace must be nested within the using namespace so that
20150 normal unqualified lookup works properly.
20152 This is useful for composing a namespace transparently from
20153 implementation namespaces. For example:
20158 template <class T> struct A @{ @};
20160 using namespace debug __attribute ((__strong__));
20161 template <> struct A<int> @{ @}; // @r{OK to specialize}
20163 template <class T> void f (A<T>);
20168 f (std::A<float>()); // @r{lookup finds} std::f
20174 @section Type Traits
20176 The C++ front end implements syntactic extensions that allow
20177 compile-time determination of
20178 various characteristics of a type (or of a
20182 @item __has_nothrow_assign (type)
20183 If @code{type} is const qualified or is a reference type then the trait is
20184 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
20185 is true, else if @code{type} is a cv class or union type with copy assignment
20186 operators that are known not to throw an exception then the trait is true,
20187 else it is false. Requires: @code{type} shall be a complete type,
20188 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20190 @item __has_nothrow_copy (type)
20191 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
20192 @code{type} is a cv class or union type with copy constructors that
20193 are known not to throw an exception then the trait is true, else it is false.
20194 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
20195 @code{void}, or an array of unknown bound.
20197 @item __has_nothrow_constructor (type)
20198 If @code{__has_trivial_constructor (type)} is true then the trait is
20199 true, else if @code{type} is a cv class or union type (or array
20200 thereof) with a default constructor that is known not to throw an
20201 exception then the trait is true, else it is false. Requires:
20202 @code{type} shall be a complete type, (possibly cv-qualified)
20203 @code{void}, or an array of unknown bound.
20205 @item __has_trivial_assign (type)
20206 If @code{type} is const qualified or is a reference type then the trait is
20207 false. Otherwise if @code{__is_pod (type)} is true then the trait is
20208 true, else if @code{type} is a cv class or union type with a trivial
20209 copy assignment ([class.copy]) then the trait is true, else it is
20210 false. Requires: @code{type} shall be a complete type, (possibly
20211 cv-qualified) @code{void}, or an array of unknown bound.
20213 @item __has_trivial_copy (type)
20214 If @code{__is_pod (type)} is true or @code{type} is a reference type
20215 then the trait is true, else if @code{type} is a cv class or union type
20216 with a trivial copy constructor ([class.copy]) then the trait
20217 is true, else it is false. Requires: @code{type} shall be a complete
20218 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20220 @item __has_trivial_constructor (type)
20221 If @code{__is_pod (type)} is true then the trait is true, else if
20222 @code{type} is a cv class or union type (or array thereof) with a
20223 trivial default constructor ([class.ctor]) then the trait is true,
20224 else it is false. Requires: @code{type} shall be a complete
20225 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20227 @item __has_trivial_destructor (type)
20228 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20229 the trait is true, else if @code{type} is a cv class or union type (or
20230 array thereof) with a trivial destructor ([class.dtor]) then the trait
20231 is true, else it is false. Requires: @code{type} shall be a complete
20232 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20234 @item __has_virtual_destructor (type)
20235 If @code{type} is a class type with a virtual destructor
20236 ([class.dtor]) then the trait is true, else it is false. Requires:
20237 @code{type} shall be a complete type, (possibly cv-qualified)
20238 @code{void}, or an array of unknown bound.
20240 @item __is_abstract (type)
20241 If @code{type} is an abstract class ([class.abstract]) then the trait
20242 is true, else it is false. Requires: @code{type} shall be a complete
20243 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20245 @item __is_base_of (base_type, derived_type)
20246 If @code{base_type} is a base class of @code{derived_type}
20247 ([class.derived]) then the trait is true, otherwise it is false.
20248 Top-level cv qualifications of @code{base_type} and
20249 @code{derived_type} are ignored. For the purposes of this trait, a
20250 class type is considered is own base. Requires: if @code{__is_class
20251 (base_type)} and @code{__is_class (derived_type)} are true and
20252 @code{base_type} and @code{derived_type} are not the same type
20253 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20254 type. Diagnostic is produced if this requirement is not met.
20256 @item __is_class (type)
20257 If @code{type} is a cv class type, and not a union type
20258 ([basic.compound]) the trait is true, else it is false.
20260 @item __is_empty (type)
20261 If @code{__is_class (type)} is false then the trait is false.
20262 Otherwise @code{type} is considered empty if and only if: @code{type}
20263 has no non-static data members, or all non-static data members, if
20264 any, are bit-fields of length 0, and @code{type} has no virtual
20265 members, and @code{type} has no virtual base classes, and @code{type}
20266 has no base classes @code{base_type} for which
20267 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20268 be a complete type, (possibly cv-qualified) @code{void}, or an array
20271 @item __is_enum (type)
20272 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20273 true, else it is false.
20275 @item __is_literal_type (type)
20276 If @code{type} is a literal type ([basic.types]) the trait is
20277 true, else it is false. Requires: @code{type} shall be a complete type,
20278 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20280 @item __is_pod (type)
20281 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20282 else it is false. Requires: @code{type} shall be a complete type,
20283 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20285 @item __is_polymorphic (type)
20286 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20287 is true, else it is false. Requires: @code{type} shall be a complete
20288 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20290 @item __is_standard_layout (type)
20291 If @code{type} is a standard-layout type ([basic.types]) the trait is
20292 true, else it is false. Requires: @code{type} shall be a complete
20293 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20295 @item __is_trivial (type)
20296 If @code{type} is a trivial type ([basic.types]) the trait is
20297 true, else it is false. Requires: @code{type} shall be a complete
20298 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20300 @item __is_union (type)
20301 If @code{type} is a cv union type ([basic.compound]) the trait is
20302 true, else it is false.
20304 @item __underlying_type (type)
20305 The underlying type of @code{type}. Requires: @code{type} shall be
20306 an enumeration type ([dcl.enum]).
20312 @section C++ Concepts
20314 C++ concepts provide much-improved support for generic programming. In
20315 particular, they allow the specification of constraints on template arguments.
20316 The constraints are used to extend the usual overloading and partial
20317 specialization capabilities of the language, allowing generic data structures
20318 and algorithms to be ``refined'' based on their properties rather than their
20321 The following keywords are reserved for concepts.
20325 States an expression as an assumption, and if possible, verifies that the
20326 assumption is valid. For example, @code{assume(n > 0)}.
20329 Introduces an axiom definition. Axioms introduce requirements on values.
20332 Introduces a universally quantified object in an axiom. For example,
20333 @code{forall (int n) n + 0 == n}).
20336 Introduces a concept definition. Concepts are sets of syntactic and semantic
20337 requirements on types and their values.
20340 Introduces constraints on template arguments or requirements for a member
20341 function of a class template.
20345 The front end also exposes a number of internal mechanism that can be used
20346 to simplify the writing of type traits. Note that some of these traits are
20347 likely to be removed in the future.
20350 @item __is_same (type1, type2)
20351 A binary type trait: true whenever the type arguments are the same.
20356 @node Java Exceptions
20357 @section Java Exceptions
20359 The Java language uses a slightly different exception handling model
20360 from C++. Normally, GNU C++ automatically detects when you are
20361 writing C++ code that uses Java exceptions, and handle them
20362 appropriately. However, if C++ code only needs to execute destructors
20363 when Java exceptions are thrown through it, GCC guesses incorrectly.
20364 Sample problematic code is:
20367 struct S @{ ~S(); @};
20368 extern void bar(); // @r{is written in Java, and may throw exceptions}
20377 The usual effect of an incorrect guess is a link failure, complaining of
20378 a missing routine called @samp{__gxx_personality_v0}.
20380 You can inform the compiler that Java exceptions are to be used in a
20381 translation unit, irrespective of what it might think, by writing
20382 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20383 @samp{#pragma} must appear before any functions that throw or catch
20384 exceptions, or run destructors when exceptions are thrown through them.
20386 You cannot mix Java and C++ exceptions in the same translation unit. It
20387 is believed to be safe to throw a C++ exception from one file through
20388 another file compiled for the Java exception model, or vice versa, but
20389 there may be bugs in this area.
20391 @node Deprecated Features
20392 @section Deprecated Features
20394 In the past, the GNU C++ compiler was extended to experiment with new
20395 features, at a time when the C++ language was still evolving. Now that
20396 the C++ standard is complete, some of those features are superseded by
20397 superior alternatives. Using the old features might cause a warning in
20398 some cases that the feature will be dropped in the future. In other
20399 cases, the feature might be gone already.
20401 While the list below is not exhaustive, it documents some of the options
20402 that are now deprecated:
20405 @item -fexternal-templates
20406 @itemx -falt-external-templates
20407 These are two of the many ways for G++ to implement template
20408 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20409 defines how template definitions have to be organized across
20410 implementation units. G++ has an implicit instantiation mechanism that
20411 should work just fine for standard-conforming code.
20413 @item -fstrict-prototype
20414 @itemx -fno-strict-prototype
20415 Previously it was possible to use an empty prototype parameter list to
20416 indicate an unspecified number of parameters (like C), rather than no
20417 parameters, as C++ demands. This feature has been removed, except where
20418 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20421 G++ allows a virtual function returning @samp{void *} to be overridden
20422 by one returning a different pointer type. This extension to the
20423 covariant return type rules is now deprecated and will be removed from a
20426 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20427 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20428 and are now removed from G++. Code using these operators should be
20429 modified to use @code{std::min} and @code{std::max} instead.
20431 The named return value extension has been deprecated, and is now
20434 The use of initializer lists with new expressions has been deprecated,
20435 and is now removed from G++.
20437 Floating and complex non-type template parameters have been deprecated,
20438 and are now removed from G++.
20440 The implicit typename extension has been deprecated and is now
20443 The use of default arguments in function pointers, function typedefs
20444 and other places where they are not permitted by the standard is
20445 deprecated and will be removed from a future version of G++.
20447 G++ allows floating-point literals to appear in integral constant expressions,
20448 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20449 This extension is deprecated and will be removed from a future version.
20451 G++ allows static data members of const floating-point type to be declared
20452 with an initializer in a class definition. The standard only allows
20453 initializers for static members of const integral types and const
20454 enumeration types so this extension has been deprecated and will be removed
20455 from a future version.
20457 @node Backwards Compatibility
20458 @section Backwards Compatibility
20459 @cindex Backwards Compatibility
20460 @cindex ARM [Annotated C++ Reference Manual]
20462 Now that there is a definitive ISO standard C++, G++ has a specification
20463 to adhere to. The C++ language evolved over time, and features that
20464 used to be acceptable in previous drafts of the standard, such as the ARM
20465 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20466 compilation of C++ written to such drafts, G++ contains some backwards
20467 compatibilities. @emph{All such backwards compatibility features are
20468 liable to disappear in future versions of G++.} They should be considered
20469 deprecated. @xref{Deprecated Features}.
20473 If a variable is declared at for scope, it used to remain in scope until
20474 the end of the scope that contained the for statement (rather than just
20475 within the for scope). G++ retains this, but issues a warning, if such a
20476 variable is accessed outside the for scope.
20478 @item Implicit C language
20479 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20480 scope to set the language. On such systems, all header files are
20481 implicitly scoped inside a C language scope. Also, an empty prototype
20482 @code{()} is treated as an unspecified number of arguments, rather
20483 than no arguments, as C++ demands.
20486 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20487 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr