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{w} floating point suffix
933 @cindex @code{q} floating point suffix
934 @cindex @code{W} floating point suffix
935 @cindex @code{Q} floating point suffix
937 As an extension, GNU C supports additional floating
938 types, @code{__float80} and @code{__float128} to support 80-bit
939 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
940 Support for additional types includes the arithmetic operators:
941 add, subtract, multiply, divide; unary arithmetic operators;
942 relational operators; equality operators; and conversions to and from
943 integer and other floating types. Use a suffix @samp{w} or @samp{W}
944 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
945 for @code{_float128}. You can declare complex types using the
946 corresponding internal complex type, @code{XCmode} for @code{__float80}
947 type and @code{TCmode} for @code{__float128} type:
950 typedef _Complex float __attribute__((mode(TC))) _Complex128;
951 typedef _Complex float __attribute__((mode(XC))) _Complex80;
954 Not all targets support additional floating-point types. @code{__float80}
955 and @code{__float128} types are supported on x86 and IA-64 targets.
956 The @code{__float128} type is supported on hppa HP-UX targets.
959 @section Half-Precision Floating Point
960 @cindex half-precision floating point
961 @cindex @code{__fp16} data type
963 On ARM targets, GCC supports half-precision (16-bit) floating point via
964 the @code{__fp16} type. You must enable this type explicitly
965 with the @option{-mfp16-format} command-line option in order to use it.
967 ARM supports two incompatible representations for half-precision
968 floating-point values. You must choose one of the representations and
969 use it consistently in your program.
971 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
972 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
973 There are 11 bits of significand precision, approximately 3
976 Specifying @option{-mfp16-format=alternative} selects the ARM
977 alternative format. This representation is similar to the IEEE
978 format, but does not support infinities or NaNs. Instead, the range
979 of exponents is extended, so that this format can represent normalized
980 values in the range of @math{2^{-14}} to 131008.
982 The @code{__fp16} type is a storage format only. For purposes
983 of arithmetic and other operations, @code{__fp16} values in C or C++
984 expressions are automatically promoted to @code{float}. In addition,
985 you cannot declare a function with a return value or parameters
986 of type @code{__fp16}.
988 Note that conversions from @code{double} to @code{__fp16}
989 involve an intermediate conversion to @code{float}. Because
990 of rounding, this can sometimes produce a different result than a
993 ARM provides hardware support for conversions between
994 @code{__fp16} and @code{float} values
995 as an extension to VFP and NEON (Advanced SIMD). GCC generates
996 code using these hardware instructions if you compile with
997 options to select an FPU that provides them;
998 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
999 in addition to the @option{-mfp16-format} option to select
1000 a half-precision format.
1002 Language-level support for the @code{__fp16} data type is
1003 independent of whether GCC generates code using hardware floating-point
1004 instructions. In cases where hardware support is not specified, GCC
1005 implements conversions between @code{__fp16} and @code{float} values
1009 @section Decimal Floating Types
1010 @cindex decimal floating types
1011 @cindex @code{_Decimal32} data type
1012 @cindex @code{_Decimal64} data type
1013 @cindex @code{_Decimal128} data type
1014 @cindex @code{df} integer suffix
1015 @cindex @code{dd} integer suffix
1016 @cindex @code{dl} integer suffix
1017 @cindex @code{DF} integer suffix
1018 @cindex @code{DD} integer suffix
1019 @cindex @code{DL} integer suffix
1021 As an extension, GNU C supports decimal floating types as
1022 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1023 floating types in GCC will evolve as the draft technical report changes.
1024 Calling conventions for any target might also change. Not all targets
1025 support decimal floating types.
1027 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1028 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1029 @code{float}, @code{double}, and @code{long double} whose radix is not
1030 specified by the C standard but is usually two.
1032 Support for decimal floating types includes the arithmetic operators
1033 add, subtract, multiply, divide; unary arithmetic operators;
1034 relational operators; equality operators; and conversions to and from
1035 integer and other floating types. Use a suffix @samp{df} or
1036 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1037 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1040 GCC support of decimal float as specified by the draft technical report
1045 When the value of a decimal floating type cannot be represented in the
1046 integer type to which it is being converted, the result is undefined
1047 rather than the result value specified by the draft technical report.
1050 GCC does not provide the C library functionality associated with
1051 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1052 @file{wchar.h}, which must come from a separate C library implementation.
1053 Because of this the GNU C compiler does not define macro
1054 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1055 the technical report.
1058 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1059 are supported by the DWARF 2 debug information format.
1065 ISO C99 supports floating-point numbers written not only in the usual
1066 decimal notation, such as @code{1.55e1}, but also numbers such as
1067 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1068 supports this in C90 mode (except in some cases when strictly
1069 conforming) and in C++. In that format the
1070 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1071 mandatory. The exponent is a decimal number that indicates the power of
1072 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1079 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1080 is the same as @code{1.55e1}.
1082 Unlike for floating-point numbers in the decimal notation the exponent
1083 is always required in the hexadecimal notation. Otherwise the compiler
1084 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1085 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1086 extension for floating-point constants of type @code{float}.
1089 @section Fixed-Point Types
1090 @cindex fixed-point types
1091 @cindex @code{_Fract} data type
1092 @cindex @code{_Accum} data type
1093 @cindex @code{_Sat} data type
1094 @cindex @code{hr} fixed-suffix
1095 @cindex @code{r} fixed-suffix
1096 @cindex @code{lr} fixed-suffix
1097 @cindex @code{llr} fixed-suffix
1098 @cindex @code{uhr} fixed-suffix
1099 @cindex @code{ur} fixed-suffix
1100 @cindex @code{ulr} fixed-suffix
1101 @cindex @code{ullr} fixed-suffix
1102 @cindex @code{hk} fixed-suffix
1103 @cindex @code{k} fixed-suffix
1104 @cindex @code{lk} fixed-suffix
1105 @cindex @code{llk} fixed-suffix
1106 @cindex @code{uhk} fixed-suffix
1107 @cindex @code{uk} fixed-suffix
1108 @cindex @code{ulk} fixed-suffix
1109 @cindex @code{ullk} fixed-suffix
1110 @cindex @code{HR} fixed-suffix
1111 @cindex @code{R} fixed-suffix
1112 @cindex @code{LR} fixed-suffix
1113 @cindex @code{LLR} fixed-suffix
1114 @cindex @code{UHR} fixed-suffix
1115 @cindex @code{UR} fixed-suffix
1116 @cindex @code{ULR} fixed-suffix
1117 @cindex @code{ULLR} fixed-suffix
1118 @cindex @code{HK} fixed-suffix
1119 @cindex @code{K} fixed-suffix
1120 @cindex @code{LK} fixed-suffix
1121 @cindex @code{LLK} fixed-suffix
1122 @cindex @code{UHK} fixed-suffix
1123 @cindex @code{UK} fixed-suffix
1124 @cindex @code{ULK} fixed-suffix
1125 @cindex @code{ULLK} fixed-suffix
1127 As an extension, GNU C supports fixed-point types as
1128 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1129 types in GCC will evolve as the draft technical report changes.
1130 Calling conventions for any target might also change. Not all targets
1131 support fixed-point types.
1133 The fixed-point types are
1134 @code{short _Fract},
1137 @code{long long _Fract},
1138 @code{unsigned short _Fract},
1139 @code{unsigned _Fract},
1140 @code{unsigned long _Fract},
1141 @code{unsigned long long _Fract},
1142 @code{_Sat short _Fract},
1144 @code{_Sat long _Fract},
1145 @code{_Sat long long _Fract},
1146 @code{_Sat unsigned short _Fract},
1147 @code{_Sat unsigned _Fract},
1148 @code{_Sat unsigned long _Fract},
1149 @code{_Sat unsigned long long _Fract},
1150 @code{short _Accum},
1153 @code{long long _Accum},
1154 @code{unsigned short _Accum},
1155 @code{unsigned _Accum},
1156 @code{unsigned long _Accum},
1157 @code{unsigned long long _Accum},
1158 @code{_Sat short _Accum},
1160 @code{_Sat long _Accum},
1161 @code{_Sat long long _Accum},
1162 @code{_Sat unsigned short _Accum},
1163 @code{_Sat unsigned _Accum},
1164 @code{_Sat unsigned long _Accum},
1165 @code{_Sat unsigned long long _Accum}.
1167 Fixed-point data values contain fractional and optional integral parts.
1168 The format of fixed-point data varies and depends on the target machine.
1170 Support for fixed-point types includes:
1173 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1175 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1177 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1179 binary shift operators (@code{<<}, @code{>>})
1181 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1183 equality operators (@code{==}, @code{!=})
1185 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1186 @code{<<=}, @code{>>=})
1188 conversions to and from integer, floating-point, or fixed-point types
1191 Use a suffix in a fixed-point literal constant:
1193 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1194 @code{_Sat short _Fract}
1195 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1196 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1197 @code{_Sat long _Fract}
1198 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1199 @code{_Sat long long _Fract}
1200 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1201 @code{_Sat unsigned short _Fract}
1202 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1203 @code{_Sat unsigned _Fract}
1204 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1205 @code{_Sat unsigned long _Fract}
1206 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1207 and @code{_Sat unsigned long long _Fract}
1208 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1209 @code{_Sat short _Accum}
1210 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1211 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1212 @code{_Sat long _Accum}
1213 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1214 @code{_Sat long long _Accum}
1215 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1216 @code{_Sat unsigned short _Accum}
1217 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1218 @code{_Sat unsigned _Accum}
1219 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1220 @code{_Sat unsigned long _Accum}
1221 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1222 and @code{_Sat unsigned long long _Accum}
1225 GCC support of fixed-point types as specified by the draft technical report
1230 Pragmas to control overflow and rounding behaviors are not implemented.
1233 Fixed-point types are supported by the DWARF 2 debug information format.
1235 @node Named Address Spaces
1236 @section Named Address Spaces
1237 @cindex Named Address Spaces
1239 As an extension, GNU C supports named address spaces as
1240 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1241 address spaces in GCC will evolve as the draft technical report
1242 changes. Calling conventions for any target might also change. At
1243 present, only the AVR, SPU, M32C, and RL78 targets support address
1244 spaces other than the generic address space.
1246 Address space identifiers may be used exactly like any other C type
1247 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1248 document for more details.
1250 @anchor{AVR Named Address Spaces}
1251 @subsection AVR Named Address Spaces
1253 On the AVR target, there are several address spaces that can be used
1254 in order to put read-only data into the flash memory and access that
1255 data by means of the special instructions @code{LPM} or @code{ELPM}
1256 needed to read from flash.
1258 Per default, any data including read-only data is located in RAM
1259 (the generic address space) so that non-generic address spaces are
1260 needed to locate read-only data in flash memory
1261 @emph{and} to generate the right instructions to access this data
1262 without using (inline) assembler code.
1266 @cindex @code{__flash} AVR Named Address Spaces
1267 The @code{__flash} qualifier locates data in the
1268 @code{.progmem.data} section. Data is read using the @code{LPM}
1269 instruction. Pointers to this address space are 16 bits wide.
1276 @cindex @code{__flash1} AVR Named Address Spaces
1277 @cindex @code{__flash2} AVR Named Address Spaces
1278 @cindex @code{__flash3} AVR Named Address Spaces
1279 @cindex @code{__flash4} AVR Named Address Spaces
1280 @cindex @code{__flash5} AVR Named Address Spaces
1281 These are 16-bit address spaces locating data in section
1282 @code{.progmem@var{N}.data} where @var{N} refers to
1283 address space @code{__flash@var{N}}.
1284 The compiler sets the @code{RAMPZ} segment register appropriately
1285 before reading data by means of the @code{ELPM} instruction.
1288 @cindex @code{__memx} AVR Named Address Spaces
1289 This is a 24-bit address space that linearizes flash and RAM:
1290 If the high bit of the address is set, data is read from
1291 RAM using the lower two bytes as RAM address.
1292 If the high bit of the address is clear, data is read from flash
1293 with @code{RAMPZ} set according to the high byte of the address.
1294 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1296 Objects in this address space are located in @code{.progmemx.data}.
1302 char my_read (const __flash char ** p)
1304 /* p is a pointer to RAM that points to a pointer to flash.
1305 The first indirection of p reads that flash pointer
1306 from RAM and the second indirection reads a char from this
1312 /* Locate array[] in flash memory */
1313 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1319 /* Return 17 by reading from flash memory */
1320 return array[array[i]];
1325 For each named address space supported by avr-gcc there is an equally
1326 named but uppercase built-in macro defined.
1327 The purpose is to facilitate testing if respective address space
1328 support is available or not:
1332 const __flash int var = 1;
1339 #include <avr/pgmspace.h> /* From AVR-LibC */
1341 const int var PROGMEM = 1;
1345 return (int) pgm_read_word (&var);
1347 #endif /* __FLASH */
1351 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1352 locates data in flash but
1353 accesses to these data read from generic address space, i.e.@:
1355 so that you need special accessors like @code{pgm_read_byte}
1356 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1357 together with attribute @code{progmem}.
1360 @b{Limitations and caveats}
1364 Reading across the 64@tie{}KiB section boundary of
1365 the @code{__flash} or @code{__flash@var{N}} address spaces
1366 shows undefined behavior. The only address space that
1367 supports reading across the 64@tie{}KiB flash segment boundaries is
1371 If you use one of the @code{__flash@var{N}} address spaces
1372 you must arrange your linker script to locate the
1373 @code{.progmem@var{N}.data} sections according to your needs.
1376 Any data or pointers to the non-generic address spaces must
1377 be qualified as @code{const}, i.e.@: as read-only data.
1378 This still applies if the data in one of these address
1379 spaces like software version number or calibration lookup table are intended to
1380 be changed after load time by, say, a boot loader. In this case
1381 the right qualification is @code{const} @code{volatile} so that the compiler
1382 must not optimize away known values or insert them
1383 as immediates into operands of instructions.
1386 The following code initializes a variable @code{pfoo}
1387 located in static storage with a 24-bit address:
1389 extern const __memx char foo;
1390 const __memx void *pfoo = &foo;
1394 Such code requires at least binutils 2.23, see
1395 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1399 @subsection M32C Named Address Spaces
1400 @cindex @code{__far} M32C Named Address Spaces
1402 On the M32C target, with the R8C and M16C CPU variants, variables
1403 qualified with @code{__far} are accessed using 32-bit addresses in
1404 order to access memory beyond the first 64@tie{}Ki bytes. If
1405 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1408 @subsection RL78 Named Address Spaces
1409 @cindex @code{__far} RL78 Named Address Spaces
1411 On the RL78 target, variables qualified with @code{__far} are accessed
1412 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1413 addresses. Non-far variables are assumed to appear in the topmost
1414 64@tie{}KiB of the address space.
1416 @subsection SPU Named Address Spaces
1417 @cindex @code{__ea} SPU Named Address Spaces
1419 On the SPU target variables may be declared as
1420 belonging to another address space by qualifying the type with the
1421 @code{__ea} address space identifier:
1428 The compiler generates special code to access the variable @code{i}.
1429 It may use runtime library
1430 support, or generate special machine instructions to access that address
1434 @section Arrays of Length Zero
1435 @cindex arrays of length zero
1436 @cindex zero-length arrays
1437 @cindex length-zero arrays
1438 @cindex flexible array members
1440 Zero-length arrays are allowed in GNU C@. They are very useful as the
1441 last element of a structure that is really a header for a variable-length
1450 struct line *thisline = (struct line *)
1451 malloc (sizeof (struct line) + this_length);
1452 thisline->length = this_length;
1455 In ISO C90, you would have to give @code{contents} a length of 1, which
1456 means either you waste space or complicate the argument to @code{malloc}.
1458 In ISO C99, you would use a @dfn{flexible array member}, which is
1459 slightly different in syntax and semantics:
1463 Flexible array members are written as @code{contents[]} without
1467 Flexible array members have incomplete type, and so the @code{sizeof}
1468 operator may not be applied. As a quirk of the original implementation
1469 of zero-length arrays, @code{sizeof} evaluates to zero.
1472 Flexible array members may only appear as the last member of a
1473 @code{struct} that is otherwise non-empty.
1476 A structure containing a flexible array member, or a union containing
1477 such a structure (possibly recursively), may not be a member of a
1478 structure or an element of an array. (However, these uses are
1479 permitted by GCC as extensions.)
1482 Non-empty initialization of zero-length
1483 arrays is treated like any case where there are more initializer
1484 elements than the array holds, in that a suitable warning about ``excess
1485 elements in array'' is given, and the excess elements (all of them, in
1486 this case) are ignored.
1488 GCC allows static initialization of flexible array members.
1489 This is equivalent to defining a new structure containing the original
1490 structure followed by an array of sufficient size to contain the data.
1491 E.g.@: in the following, @code{f1} is constructed as if it were declared
1497 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1500 struct f1 f1; int data[3];
1501 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1505 The convenience of this extension is that @code{f1} has the desired
1506 type, eliminating the need to consistently refer to @code{f2.f1}.
1508 This has symmetry with normal static arrays, in that an array of
1509 unknown size is also written with @code{[]}.
1511 Of course, this extension only makes sense if the extra data comes at
1512 the end of a top-level object, as otherwise we would be overwriting
1513 data at subsequent offsets. To avoid undue complication and confusion
1514 with initialization of deeply nested arrays, we simply disallow any
1515 non-empty initialization except when the structure is the top-level
1516 object. For example:
1519 struct foo @{ int x; int y[]; @};
1520 struct bar @{ struct foo z; @};
1522 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1523 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1524 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1525 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1528 @node Empty Structures
1529 @section Structures with No Members
1530 @cindex empty structures
1531 @cindex zero-size structures
1533 GCC permits a C structure to have no members:
1540 The structure has size zero. In C++, empty structures are part
1541 of the language. G++ treats empty structures as if they had a single
1542 member of type @code{char}.
1544 @node Variable Length
1545 @section Arrays of Variable Length
1546 @cindex variable-length arrays
1547 @cindex arrays of variable length
1550 Variable-length automatic arrays are allowed in ISO C99, and as an
1551 extension GCC accepts them in C90 mode and in C++. These arrays are
1552 declared like any other automatic arrays, but with a length that is not
1553 a constant expression. The storage is allocated at the point of
1554 declaration and deallocated when the block scope containing the declaration
1560 concat_fopen (char *s1, char *s2, char *mode)
1562 char str[strlen (s1) + strlen (s2) + 1];
1565 return fopen (str, mode);
1569 @cindex scope of a variable length array
1570 @cindex variable-length array scope
1571 @cindex deallocating variable length arrays
1572 Jumping or breaking out of the scope of the array name deallocates the
1573 storage. Jumping into the scope is not allowed; you get an error
1576 @cindex variable-length array in a structure
1577 As an extension, GCC accepts variable-length arrays as a member of
1578 a structure or a union. For example:
1584 struct S @{ int x[n]; @};
1588 @cindex @code{alloca} vs variable-length arrays
1589 You can use the function @code{alloca} to get an effect much like
1590 variable-length arrays. The function @code{alloca} is available in
1591 many other C implementations (but not in all). On the other hand,
1592 variable-length arrays are more elegant.
1594 There are other differences between these two methods. Space allocated
1595 with @code{alloca} exists until the containing @emph{function} returns.
1596 The space for a variable-length array is deallocated as soon as the array
1597 name's scope ends. (If you use both variable-length arrays and
1598 @code{alloca} in the same function, deallocation of a variable-length array
1599 also deallocates anything more recently allocated with @code{alloca}.)
1601 You can also use variable-length arrays as arguments to functions:
1605 tester (int len, char data[len][len])
1611 The length of an array is computed once when the storage is allocated
1612 and is remembered for the scope of the array in case you access it with
1615 If you want to pass the array first and the length afterward, you can
1616 use a forward declaration in the parameter list---another GNU extension.
1620 tester (int len; char data[len][len], int len)
1626 @cindex parameter forward declaration
1627 The @samp{int len} before the semicolon is a @dfn{parameter forward
1628 declaration}, and it serves the purpose of making the name @code{len}
1629 known when the declaration of @code{data} is parsed.
1631 You can write any number of such parameter forward declarations in the
1632 parameter list. They can be separated by commas or semicolons, but the
1633 last one must end with a semicolon, which is followed by the ``real''
1634 parameter declarations. Each forward declaration must match a ``real''
1635 declaration in parameter name and data type. ISO C99 does not support
1636 parameter forward declarations.
1638 @node Variadic Macros
1639 @section Macros with a Variable Number of Arguments.
1640 @cindex variable number of arguments
1641 @cindex macro with variable arguments
1642 @cindex rest argument (in macro)
1643 @cindex variadic macros
1645 In the ISO C standard of 1999, a macro can be declared to accept a
1646 variable number of arguments much as a function can. The syntax for
1647 defining the macro is similar to that of a function. Here is an
1651 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1655 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1656 such a macro, it represents the zero or more tokens until the closing
1657 parenthesis that ends the invocation, including any commas. This set of
1658 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1659 wherever it appears. See the CPP manual for more information.
1661 GCC has long supported variadic macros, and used a different syntax that
1662 allowed you to give a name to the variable arguments just like any other
1663 argument. Here is an example:
1666 #define debug(format, args...) fprintf (stderr, format, args)
1670 This is in all ways equivalent to the ISO C example above, but arguably
1671 more readable and descriptive.
1673 GNU CPP has two further variadic macro extensions, and permits them to
1674 be used with either of the above forms of macro definition.
1676 In standard C, you are not allowed to leave the variable argument out
1677 entirely; but you are allowed to pass an empty argument. For example,
1678 this invocation is invalid in ISO C, because there is no comma after
1685 GNU CPP permits you to completely omit the variable arguments in this
1686 way. In the above examples, the compiler would complain, though since
1687 the expansion of the macro still has the extra comma after the format
1690 To help solve this problem, CPP behaves specially for variable arguments
1691 used with the token paste operator, @samp{##}. If instead you write
1694 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1698 and if the variable arguments are omitted or empty, the @samp{##}
1699 operator causes the preprocessor to remove the comma before it. If you
1700 do provide some variable arguments in your macro invocation, GNU CPP
1701 does not complain about the paste operation and instead places the
1702 variable arguments after the comma. Just like any other pasted macro
1703 argument, these arguments are not macro expanded.
1705 @node Escaped Newlines
1706 @section Slightly Looser Rules for Escaped Newlines
1707 @cindex escaped newlines
1708 @cindex newlines (escaped)
1710 The preprocessor treatment of escaped newlines is more relaxed
1711 than that specified by the C90 standard, which requires the newline
1712 to immediately follow a backslash.
1713 GCC's implementation allows whitespace in the form
1714 of spaces, horizontal and vertical tabs, and form feeds between the
1715 backslash and the subsequent newline. The preprocessor issues a
1716 warning, but treats it as a valid escaped newline and combines the two
1717 lines to form a single logical line. This works within comments and
1718 tokens, as well as between tokens. Comments are @emph{not} treated as
1719 whitespace for the purposes of this relaxation, since they have not
1720 yet been replaced with spaces.
1723 @section Non-Lvalue Arrays May Have Subscripts
1724 @cindex subscripting
1725 @cindex arrays, non-lvalue
1727 @cindex subscripting and function values
1728 In ISO C99, arrays that are not lvalues still decay to pointers, and
1729 may be subscripted, although they may not be modified or used after
1730 the next sequence point and the unary @samp{&} operator may not be
1731 applied to them. As an extension, GNU C allows such arrays to be
1732 subscripted in C90 mode, though otherwise they do not decay to
1733 pointers outside C99 mode. For example,
1734 this is valid in GNU C though not valid in C90:
1738 struct foo @{int a[4];@};
1744 return f().a[index];
1750 @section Arithmetic on @code{void}- and Function-Pointers
1751 @cindex void pointers, arithmetic
1752 @cindex void, size of pointer to
1753 @cindex function pointers, arithmetic
1754 @cindex function, size of pointer to
1756 In GNU C, addition and subtraction operations are supported on pointers to
1757 @code{void} and on pointers to functions. This is done by treating the
1758 size of a @code{void} or of a function as 1.
1760 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1761 and on function types, and returns 1.
1763 @opindex Wpointer-arith
1764 The option @option{-Wpointer-arith} requests a warning if these extensions
1767 @node Pointers to Arrays
1768 @section Pointers to Arrays with Qualifiers Work as Expected
1769 @cindex pointers to arrays
1770 @cindex const qualifier
1772 In GNU C, pointers to arrays with qualifiers work similar to pointers
1773 to other qualified types. For example, a value of type @code{int (*)[5]}
1774 can be used to initialize a variable of type @code{const int (*)[5]}.
1775 These types are incompatible in ISO C because the @code{const} qualifier
1776 is formally attached to the element type of the array and not the
1781 transpose (int N, int M, double out[M][N], const double in[N][M]);
1785 transpose(3, 2, y, x);
1789 @section Non-Constant Initializers
1790 @cindex initializers, non-constant
1791 @cindex non-constant initializers
1793 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1794 automatic variable are not required to be constant expressions in GNU C@.
1795 Here is an example of an initializer with run-time varying elements:
1798 foo (float f, float g)
1800 float beat_freqs[2] = @{ f-g, f+g @};
1805 @node Compound Literals
1806 @section Compound Literals
1807 @cindex constructor expressions
1808 @cindex initializations in expressions
1809 @cindex structures, constructor expression
1810 @cindex expressions, constructor
1811 @cindex compound literals
1812 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1814 ISO C99 supports compound literals. A compound literal looks like
1815 a cast containing an initializer. Its value is an object of the
1816 type specified in the cast, containing the elements specified in
1817 the initializer; it is an lvalue. As an extension, GCC supports
1818 compound literals in C90 mode and in C++, though the semantics are
1819 somewhat different in C++.
1821 Usually, the specified type is a structure. Assume that
1822 @code{struct foo} and @code{structure} are declared as shown:
1825 struct foo @{int a; char b[2];@} structure;
1829 Here is an example of constructing a @code{struct foo} with a compound literal:
1832 structure = ((struct foo) @{x + y, 'a', 0@});
1836 This is equivalent to writing the following:
1840 struct foo temp = @{x + y, 'a', 0@};
1845 You can also construct an array, though this is dangerous in C++, as
1846 explained below. If all the elements of the compound literal are
1847 (made up of) simple constant expressions, suitable for use in
1848 initializers of objects of static storage duration, then the compound
1849 literal can be coerced to a pointer to its first element and used in
1850 such an initializer, as shown here:
1853 char **foo = (char *[]) @{ "x", "y", "z" @};
1856 Compound literals for scalar types and union types are
1857 also allowed, but then the compound literal is equivalent
1860 As a GNU extension, GCC allows initialization of objects with static storage
1861 duration by compound literals (which is not possible in ISO C99, because
1862 the initializer is not a constant).
1863 It is handled as if the object is initialized only with the bracket
1864 enclosed list if the types of the compound literal and the object match.
1865 The initializer list of the compound literal must be constant.
1866 If the object being initialized has array type of unknown size, the size is
1867 determined by compound literal size.
1870 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1871 static int y[] = (int []) @{1, 2, 3@};
1872 static int z[] = (int [3]) @{1@};
1876 The above lines are equivalent to the following:
1878 static struct foo x = @{1, 'a', 'b'@};
1879 static int y[] = @{1, 2, 3@};
1880 static int z[] = @{1, 0, 0@};
1883 In C, a compound literal designates an unnamed object with static or
1884 automatic storage duration. In C++, a compound literal designates a
1885 temporary object, which only lives until the end of its
1886 full-expression. As a result, well-defined C code that takes the
1887 address of a subobject of a compound literal can be undefined in C++,
1888 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1889 For instance, if the array compound literal example above appeared
1890 inside a function, any subsequent use of @samp{foo} in C++ has
1891 undefined behavior because the lifetime of the array ends after the
1892 declaration of @samp{foo}.
1894 As an optimization, the C++ compiler sometimes gives array compound
1895 literals longer lifetimes: when the array either appears outside a
1896 function or has const-qualified type. If @samp{foo} and its
1897 initializer had elements of @samp{char *const} type rather than
1898 @samp{char *}, or if @samp{foo} were a global variable, the array
1899 would have static storage duration. But it is probably safest just to
1900 avoid the use of array compound literals in code compiled as C++.
1902 @node Designated Inits
1903 @section Designated Initializers
1904 @cindex initializers with labeled elements
1905 @cindex labeled elements in initializers
1906 @cindex case labels in initializers
1907 @cindex designated initializers
1909 Standard C90 requires the elements of an initializer to appear in a fixed
1910 order, the same as the order of the elements in the array or structure
1913 In ISO C99 you can give the elements in any order, specifying the array
1914 indices or structure field names they apply to, and GNU C allows this as
1915 an extension in C90 mode as well. This extension is not
1916 implemented in GNU C++.
1918 To specify an array index, write
1919 @samp{[@var{index}] =} before the element value. For example,
1922 int a[6] = @{ [4] = 29, [2] = 15 @};
1929 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1933 The index values must be constant expressions, even if the array being
1934 initialized is automatic.
1936 An alternative syntax for this that has been obsolete since GCC 2.5 but
1937 GCC still accepts is to write @samp{[@var{index}]} before the element
1938 value, with no @samp{=}.
1940 To initialize a range of elements to the same value, write
1941 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1942 extension. For example,
1945 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1949 If the value in it has side-effects, the side-effects happen only once,
1950 not for each initialized field by the range initializer.
1953 Note that the length of the array is the highest value specified
1956 In a structure initializer, specify the name of a field to initialize
1957 with @samp{.@var{fieldname} =} before the element value. For example,
1958 given the following structure,
1961 struct point @{ int x, y; @};
1965 the following initialization
1968 struct point p = @{ .y = yvalue, .x = xvalue @};
1975 struct point p = @{ xvalue, yvalue @};
1978 Another syntax that has the same meaning, obsolete since GCC 2.5, is
1979 @samp{@var{fieldname}:}, as shown here:
1982 struct point p = @{ y: yvalue, x: xvalue @};
1985 Omitted field members are implicitly initialized the same as objects
1986 that have static storage duration.
1989 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1990 @dfn{designator}. You can also use a designator (or the obsolete colon
1991 syntax) when initializing a union, to specify which element of the union
1992 should be used. For example,
1995 union foo @{ int i; double d; @};
1997 union foo f = @{ .d = 4 @};
2001 converts 4 to a @code{double} to store it in the union using
2002 the second element. By contrast, casting 4 to type @code{union foo}
2003 stores it into the union as the integer @code{i}, since it is
2004 an integer. (@xref{Cast to Union}.)
2006 You can combine this technique of naming elements with ordinary C
2007 initialization of successive elements. Each initializer element that
2008 does not have a designator applies to the next consecutive element of the
2009 array or structure. For example,
2012 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2019 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2022 Labeling the elements of an array initializer is especially useful
2023 when the indices are characters or belong to an @code{enum} type.
2028 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2029 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2032 @cindex designator lists
2033 You can also write a series of @samp{.@var{fieldname}} and
2034 @samp{[@var{index}]} designators before an @samp{=} to specify a
2035 nested subobject to initialize; the list is taken relative to the
2036 subobject corresponding to the closest surrounding brace pair. For
2037 example, with the @samp{struct point} declaration above:
2040 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2044 If the same field is initialized multiple times, it has the value from
2045 the last initialization. If any such overridden initialization has
2046 side-effect, it is unspecified whether the side-effect happens or not.
2047 Currently, GCC discards them and issues a warning.
2050 @section Case Ranges
2052 @cindex ranges in case statements
2054 You can specify a range of consecutive values in a single @code{case} label,
2058 case @var{low} ... @var{high}:
2062 This has the same effect as the proper number of individual @code{case}
2063 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2065 This feature is especially useful for ranges of ASCII character codes:
2071 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2072 it may be parsed wrong when you use it with integer values. For example,
2087 @section Cast to a Union Type
2088 @cindex cast to a union
2089 @cindex union, casting to a
2091 A cast to union type is similar to other casts, except that the type
2092 specified is a union type. You can specify the type either with
2093 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2094 a constructor, not a cast, and hence does not yield an lvalue like
2095 normal casts. (@xref{Compound Literals}.)
2097 The types that may be cast to the union type are those of the members
2098 of the union. Thus, given the following union and variables:
2101 union foo @{ int i; double d; @};
2107 both @code{x} and @code{y} can be cast to type @code{union foo}.
2109 Using the cast as the right-hand side of an assignment to a variable of
2110 union type is equivalent to storing in a member of the union:
2115 u = (union foo) x @equiv{} u.i = x
2116 u = (union foo) y @equiv{} u.d = y
2119 You can also use the union cast as a function argument:
2122 void hack (union foo);
2124 hack ((union foo) x);
2127 @node Mixed Declarations
2128 @section Mixed Declarations and Code
2129 @cindex mixed declarations and code
2130 @cindex declarations, mixed with code
2131 @cindex code, mixed with declarations
2133 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2134 within compound statements. As an extension, GNU C also allows this in
2135 C90 mode. For example, you could do:
2144 Each identifier is visible from where it is declared until the end of
2145 the enclosing block.
2147 @node Function Attributes
2148 @section Declaring Attributes of Functions
2149 @cindex function attributes
2150 @cindex declaring attributes of functions
2151 @cindex @code{volatile} applied to function
2152 @cindex @code{const} applied to function
2154 In GNU C, you can use function attributes to declare certain things
2155 about functions called in your program which help the compiler
2156 optimize calls and check your code more carefully. For example, you
2157 can use attributes to declare that a function never returns
2158 (@code{noreturn}), returns a value depending only on its arguments
2159 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2161 You can also use attributes to control memory placement, code
2162 generation options or call/return conventions within the function
2163 being annotated. Many of these attributes are target-specific. For
2164 example, many targets support attributes for defining interrupt
2165 handler functions, which typically must follow special register usage
2166 and return conventions.
2168 Function attributes are introduced by the @code{__attribute__} keyword
2169 on a declaration, followed by an attribute specification inside double
2170 parentheses. You can specify multiple attributes in a declaration by
2171 separating them by commas within the double parentheses or by
2172 immediately following an attribute declaration with another attribute
2173 declaration. @xref{Attribute Syntax}, for the exact rules on
2174 attribute syntax and placement.
2176 GCC also supports attributes on
2177 variable declarations (@pxref{Variable Attributes}),
2178 labels (@pxref{Label Attributes}),
2179 enumerators (@pxref{Enumerator Attributes}),
2180 and types (@pxref{Type Attributes}).
2182 There is some overlap between the purposes of attributes and pragmas
2183 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2184 found convenient to use @code{__attribute__} to achieve a natural
2185 attachment of attributes to their corresponding declarations, whereas
2186 @code{#pragma} is of use for compatibility with other compilers
2187 or constructs that do not naturally form part of the grammar.
2189 In addition to the attributes documented here,
2190 GCC plugins may provide their own attributes.
2193 * Common Function Attributes::
2194 * AArch64 Function Attributes::
2195 * ARC Function Attributes::
2196 * ARM Function Attributes::
2197 * AVR Function Attributes::
2198 * Blackfin Function Attributes::
2199 * CR16 Function Attributes::
2200 * Epiphany Function Attributes::
2201 * H8/300 Function Attributes::
2202 * IA-64 Function Attributes::
2203 * M32C Function Attributes::
2204 * M32R/D Function Attributes::
2205 * m68k Function Attributes::
2206 * MCORE Function Attributes::
2207 * MeP Function Attributes::
2208 * MicroBlaze Function Attributes::
2209 * Microsoft Windows Function Attributes::
2210 * MIPS Function Attributes::
2211 * MSP430 Function Attributes::
2212 * NDS32 Function Attributes::
2213 * Nios II Function Attributes::
2214 * PowerPC Function Attributes::
2215 * RL78 Function Attributes::
2216 * RX Function Attributes::
2217 * S/390 Function Attributes::
2218 * SH Function Attributes::
2219 * SPU Function Attributes::
2220 * Symbian OS Function Attributes::
2221 * Visium Function Attributes::
2222 * x86 Function Attributes::
2223 * Xstormy16 Function Attributes::
2226 @node Common Function Attributes
2227 @subsection Common Function Attributes
2229 The following attributes are supported on most targets.
2232 @c Keep this table alphabetized by attribute name. Treat _ as space.
2234 @item alias ("@var{target}")
2235 @cindex @code{alias} function attribute
2236 The @code{alias} attribute causes the declaration to be emitted as an
2237 alias for another symbol, which must be specified. For instance,
2240 void __f () @{ /* @r{Do something.} */; @}
2241 void f () __attribute__ ((weak, alias ("__f")));
2245 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2246 mangled name for the target must be used. It is an error if @samp{__f}
2247 is not defined in the same translation unit.
2249 This attribute requires assembler and object file support,
2250 and may not be available on all targets.
2252 @item aligned (@var{alignment})
2253 @cindex @code{aligned} function attribute
2254 This attribute specifies a minimum alignment for the function,
2257 You cannot use this attribute to decrease the alignment of a function,
2258 only to increase it. However, when you explicitly specify a function
2259 alignment this overrides the effect of the
2260 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2263 Note that the effectiveness of @code{aligned} attributes may be
2264 limited by inherent limitations in your linker. On many systems, the
2265 linker is only able to arrange for functions to be aligned up to a
2266 certain maximum alignment. (For some linkers, the maximum supported
2267 alignment may be very very small.) See your linker documentation for
2268 further information.
2270 The @code{aligned} attribute can also be used for variables and fields
2271 (@pxref{Variable Attributes}.)
2274 @cindex @code{alloc_align} function attribute
2275 The @code{alloc_align} attribute is used to tell the compiler that the
2276 function return value points to memory, where the returned pointer minimum
2277 alignment is given by one of the functions parameters. GCC uses this
2278 information to improve pointer alignment analysis.
2280 The function parameter denoting the allocated alignment is specified by
2281 one integer argument, whose number is the argument of the attribute.
2282 Argument numbering starts at one.
2287 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2291 declares that @code{my_memalign} returns memory with minimum alignment
2292 given by parameter 1.
2295 @cindex @code{alloc_size} function attribute
2296 The @code{alloc_size} attribute is used to tell the compiler that the
2297 function return value points to memory, where the size is given by
2298 one or two of the functions parameters. GCC uses this
2299 information to improve the correctness of @code{__builtin_object_size}.
2301 The function parameter(s) denoting the allocated size are specified by
2302 one or two integer arguments supplied to the attribute. The allocated size
2303 is either the value of the single function argument specified or the product
2304 of the two function arguments specified. Argument numbering starts at
2310 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2311 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2315 declares that @code{my_calloc} returns memory of the size given by
2316 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2317 of the size given by parameter 2.
2320 @cindex @code{always_inline} function attribute
2321 Generally, functions are not inlined unless optimization is specified.
2322 For functions declared inline, this attribute inlines the function
2323 independent of any restrictions that otherwise apply to inlining.
2324 Failure to inline such a function is diagnosed as an error.
2325 Note that if such a function is called indirectly the compiler may
2326 or may not inline it depending on optimization level and a failure
2327 to inline an indirect call may or may not be diagnosed.
2330 @cindex @code{artificial} function attribute
2331 This attribute is useful for small inline wrappers that if possible
2332 should appear during debugging as a unit. Depending on the debug
2333 info format it either means marking the function as artificial
2334 or using the caller location for all instructions within the inlined
2337 @item assume_aligned
2338 @cindex @code{assume_aligned} function attribute
2339 The @code{assume_aligned} attribute is used to tell the compiler that the
2340 function return value points to memory, where the returned pointer minimum
2341 alignment is given by the first argument.
2342 If the attribute has two arguments, the second argument is misalignment offset.
2347 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2348 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2352 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2353 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2356 @item bnd_instrument
2357 @cindex @code{bnd_instrument} function attribute
2358 The @code{bnd_instrument} attribute on functions is used to inform the
2359 compiler that the function should be instrumented when compiled
2360 with the @option{-fchkp-instrument-marked-only} option.
2363 @cindex @code{bnd_legacy} function attribute
2364 @cindex Pointer Bounds Checker attributes
2365 The @code{bnd_legacy} attribute on functions is used to inform the
2366 compiler that the function should not be instrumented when compiled
2367 with the @option{-fcheck-pointer-bounds} option.
2370 @cindex @code{cold} function attribute
2371 The @code{cold} attribute on functions is used to inform the compiler that
2372 the function is unlikely to be executed. The function is optimized for
2373 size rather than speed and on many targets it is placed into a special
2374 subsection of the text section so all cold functions appear close together,
2375 improving code locality of non-cold parts of program. The paths leading
2376 to calls of cold functions within code are marked as unlikely by the branch
2377 prediction mechanism. It is thus useful to mark functions used to handle
2378 unlikely conditions, such as @code{perror}, as cold to improve optimization
2379 of hot functions that do call marked functions in rare occasions.
2381 When profile feedback is available, via @option{-fprofile-use}, cold functions
2382 are automatically detected and this attribute is ignored.
2385 @cindex @code{const} function attribute
2386 @cindex functions that have no side effects
2387 Many functions do not examine any values except their arguments, and
2388 have no effects except the return value. Basically this is just slightly
2389 more strict class than the @code{pure} attribute below, since function is not
2390 allowed to read global memory.
2392 @cindex pointer arguments
2393 Note that a function that has pointer arguments and examines the data
2394 pointed to must @emph{not} be declared @code{const}. Likewise, a
2395 function that calls a non-@code{const} function usually must not be
2396 @code{const}. It does not make sense for a @code{const} function to
2401 @itemx constructor (@var{priority})
2402 @itemx destructor (@var{priority})
2403 @cindex @code{constructor} function attribute
2404 @cindex @code{destructor} function attribute
2405 The @code{constructor} attribute causes the function to be called
2406 automatically before execution enters @code{main ()}. Similarly, the
2407 @code{destructor} attribute causes the function to be called
2408 automatically after @code{main ()} completes or @code{exit ()} is
2409 called. Functions with these attributes are useful for
2410 initializing data that is used implicitly during the execution of
2413 You may provide an optional integer priority to control the order in
2414 which constructor and destructor functions are run. A constructor
2415 with a smaller priority number runs before a constructor with a larger
2416 priority number; the opposite relationship holds for destructors. So,
2417 if you have a constructor that allocates a resource and a destructor
2418 that deallocates the same resource, both functions typically have the
2419 same priority. The priorities for constructor and destructor
2420 functions are the same as those specified for namespace-scope C++
2421 objects (@pxref{C++ Attributes}).
2423 These attributes are not currently implemented for Objective-C@.
2426 @itemx deprecated (@var{msg})
2427 @cindex @code{deprecated} function attribute
2428 The @code{deprecated} attribute results in a warning if the function
2429 is used anywhere in the source file. This is useful when identifying
2430 functions that are expected to be removed in a future version of a
2431 program. The warning also includes the location of the declaration
2432 of the deprecated function, to enable users to easily find further
2433 information about why the function is deprecated, or what they should
2434 do instead. Note that the warnings only occurs for uses:
2437 int old_fn () __attribute__ ((deprecated));
2439 int (*fn_ptr)() = old_fn;
2443 results in a warning on line 3 but not line 2. The optional @var{msg}
2444 argument, which must be a string, is printed in the warning if
2447 The @code{deprecated} attribute can also be used for variables and
2448 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2450 @item error ("@var{message}")
2451 @itemx warning ("@var{message}")
2452 @cindex @code{error} function attribute
2453 @cindex @code{warning} function attribute
2454 If the @code{error} or @code{warning} attribute
2455 is used on a function declaration and a call to such a function
2456 is not eliminated through dead code elimination or other optimizations,
2457 an error or warning (respectively) that includes @var{message} is diagnosed.
2459 for compile-time checking, especially together with @code{__builtin_constant_p}
2460 and inline functions where checking the inline function arguments is not
2461 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2463 While it is possible to leave the function undefined and thus invoke
2464 a link failure (to define the function with
2465 a message in @code{.gnu.warning*} section),
2466 when using these attributes the problem is diagnosed
2467 earlier and with exact location of the call even in presence of inline
2468 functions or when not emitting debugging information.
2470 @item externally_visible
2471 @cindex @code{externally_visible} function attribute
2472 This attribute, attached to a global variable or function, nullifies
2473 the effect of the @option{-fwhole-program} command-line option, so the
2474 object remains visible outside the current compilation unit.
2476 If @option{-fwhole-program} is used together with @option{-flto} and
2477 @command{gold} is used as the linker plugin,
2478 @code{externally_visible} attributes are automatically added to functions
2479 (not variable yet due to a current @command{gold} issue)
2480 that are accessed outside of LTO objects according to resolution file
2481 produced by @command{gold}.
2482 For other linkers that cannot generate resolution file,
2483 explicit @code{externally_visible} attributes are still necessary.
2486 @cindex @code{flatten} function attribute
2487 Generally, inlining into a function is limited. For a function marked with
2488 this attribute, every call inside this function is inlined, if possible.
2489 Whether the function itself is considered for inlining depends on its size and
2490 the current inlining parameters.
2492 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2493 @cindex @code{format} function attribute
2494 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2496 The @code{format} attribute specifies that a function takes @code{printf},
2497 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2498 should be type-checked against a format string. For example, the
2503 my_printf (void *my_object, const char *my_format, ...)
2504 __attribute__ ((format (printf, 2, 3)));
2508 causes the compiler to check the arguments in calls to @code{my_printf}
2509 for consistency with the @code{printf} style format string argument
2512 The parameter @var{archetype} determines how the format string is
2513 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2514 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2515 @code{strfmon}. (You can also use @code{__printf__},
2516 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2517 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2518 @code{ms_strftime} are also present.
2519 @var{archetype} values such as @code{printf} refer to the formats accepted
2520 by the system's C runtime library,
2521 while values prefixed with @samp{gnu_} always refer
2522 to the formats accepted by the GNU C Library. On Microsoft Windows
2523 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2524 @file{msvcrt.dll} library.
2525 The parameter @var{string-index}
2526 specifies which argument is the format string argument (starting
2527 from 1), while @var{first-to-check} is the number of the first
2528 argument to check against the format string. For functions
2529 where the arguments are not available to be checked (such as
2530 @code{vprintf}), specify the third parameter as zero. In this case the
2531 compiler only checks the format string for consistency. For
2532 @code{strftime} formats, the third parameter is required to be zero.
2533 Since non-static C++ methods have an implicit @code{this} argument, the
2534 arguments of such methods should be counted from two, not one, when
2535 giving values for @var{string-index} and @var{first-to-check}.
2537 In the example above, the format string (@code{my_format}) is the second
2538 argument of the function @code{my_print}, and the arguments to check
2539 start with the third argument, so the correct parameters for the format
2540 attribute are 2 and 3.
2542 @opindex ffreestanding
2543 @opindex fno-builtin
2544 The @code{format} attribute allows you to identify your own functions
2545 that take format strings as arguments, so that GCC can check the
2546 calls to these functions for errors. The compiler always (unless
2547 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2548 for the standard library functions @code{printf}, @code{fprintf},
2549 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2550 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2551 warnings are requested (using @option{-Wformat}), so there is no need to
2552 modify the header file @file{stdio.h}. In C99 mode, the functions
2553 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2554 @code{vsscanf} are also checked. Except in strictly conforming C
2555 standard modes, the X/Open function @code{strfmon} is also checked as
2556 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2557 @xref{C Dialect Options,,Options Controlling C Dialect}.
2559 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2560 recognized in the same context. Declarations including these format attributes
2561 are parsed for correct syntax, however the result of checking of such format
2562 strings is not yet defined, and is not carried out by this version of the
2565 The target may also provide additional types of format checks.
2566 @xref{Target Format Checks,,Format Checks Specific to Particular
2569 @item format_arg (@var{string-index})
2570 @cindex @code{format_arg} function attribute
2571 @opindex Wformat-nonliteral
2572 The @code{format_arg} attribute specifies that a function takes a format
2573 string for a @code{printf}, @code{scanf}, @code{strftime} or
2574 @code{strfmon} style function and modifies it (for example, to translate
2575 it into another language), so the result can be passed to a
2576 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2577 function (with the remaining arguments to the format function the same
2578 as they would have been for the unmodified string). For example, the
2583 my_dgettext (char *my_domain, const char *my_format)
2584 __attribute__ ((format_arg (2)));
2588 causes the compiler to check the arguments in calls to a @code{printf},
2589 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2590 format string argument is a call to the @code{my_dgettext} function, for
2591 consistency with the format string argument @code{my_format}. If the
2592 @code{format_arg} attribute had not been specified, all the compiler
2593 could tell in such calls to format functions would be that the format
2594 string argument is not constant; this would generate a warning when
2595 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2596 without the attribute.
2598 The parameter @var{string-index} specifies which argument is the format
2599 string argument (starting from one). Since non-static C++ methods have
2600 an implicit @code{this} argument, the arguments of such methods should
2601 be counted from two.
2603 The @code{format_arg} attribute allows you to identify your own
2604 functions that modify format strings, so that GCC can check the
2605 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2606 type function whose operands are a call to one of your own function.
2607 The compiler always treats @code{gettext}, @code{dgettext}, and
2608 @code{dcgettext} in this manner except when strict ISO C support is
2609 requested by @option{-ansi} or an appropriate @option{-std} option, or
2610 @option{-ffreestanding} or @option{-fno-builtin}
2611 is used. @xref{C Dialect Options,,Options
2612 Controlling C Dialect}.
2614 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2615 @code{NSString} reference for compatibility with the @code{format} attribute
2618 The target may also allow additional types in @code{format-arg} attributes.
2619 @xref{Target Format Checks,,Format Checks Specific to Particular
2623 @cindex @code{gnu_inline} function attribute
2624 This attribute should be used with a function that is also declared
2625 with the @code{inline} keyword. It directs GCC to treat the function
2626 as if it were defined in gnu90 mode even when compiling in C99 or
2629 If the function is declared @code{extern}, then this definition of the
2630 function is used only for inlining. In no case is the function
2631 compiled as a standalone function, not even if you take its address
2632 explicitly. Such an address becomes an external reference, as if you
2633 had only declared the function, and had not defined it. This has
2634 almost the effect of a macro. The way to use this is to put a
2635 function definition in a header file with this attribute, and put
2636 another copy of the function, without @code{extern}, in a library
2637 file. The definition in the header file causes most calls to the
2638 function to be inlined. If any uses of the function remain, they
2639 refer to the single copy in the library. Note that the two
2640 definitions of the functions need not be precisely the same, although
2641 if they do not have the same effect your program may behave oddly.
2643 In C, if the function is neither @code{extern} nor @code{static}, then
2644 the function is compiled as a standalone function, as well as being
2645 inlined where possible.
2647 This is how GCC traditionally handled functions declared
2648 @code{inline}. Since ISO C99 specifies a different semantics for
2649 @code{inline}, this function attribute is provided as a transition
2650 measure and as a useful feature in its own right. This attribute is
2651 available in GCC 4.1.3 and later. It is available if either of the
2652 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2653 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2654 Function is As Fast As a Macro}.
2656 In C++, this attribute does not depend on @code{extern} in any way,
2657 but it still requires the @code{inline} keyword to enable its special
2661 @cindex @code{hot} function attribute
2662 The @code{hot} attribute on a function is used to inform the compiler that
2663 the function is a hot spot of the compiled program. The function is
2664 optimized more aggressively and on many targets it is placed into a special
2665 subsection of the text section so all hot functions appear close together,
2668 When profile feedback is available, via @option{-fprofile-use}, hot functions
2669 are automatically detected and this attribute is ignored.
2671 @item ifunc ("@var{resolver}")
2672 @cindex @code{ifunc} function attribute
2673 @cindex indirect functions
2674 @cindex functions that are dynamically resolved
2675 The @code{ifunc} attribute is used to mark a function as an indirect
2676 function using the STT_GNU_IFUNC symbol type extension to the ELF
2677 standard. This allows the resolution of the symbol value to be
2678 determined dynamically at load time, and an optimized version of the
2679 routine can be selected for the particular processor or other system
2680 characteristics determined then. To use this attribute, first define
2681 the implementation functions available, and a resolver function that
2682 returns a pointer to the selected implementation function. The
2683 implementation functions' declarations must match the API of the
2684 function being implemented, the resolver's declaration is be a
2685 function returning pointer to void function returning void:
2688 void *my_memcpy (void *dst, const void *src, size_t len)
2693 static void (*resolve_memcpy (void)) (void)
2695 return my_memcpy; // we'll just always select this routine
2700 The exported header file declaring the function the user calls would
2704 extern void *memcpy (void *, const void *, size_t);
2708 allowing the user to call this as a regular function, unaware of the
2709 implementation. Finally, the indirect function needs to be defined in
2710 the same translation unit as the resolver function:
2713 void *memcpy (void *, const void *, size_t)
2714 __attribute__ ((ifunc ("resolve_memcpy")));
2717 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2718 and GNU C Library version 2.11.1 are required to use this feature.
2721 @itemx interrupt_handler
2722 Many GCC back ends support attributes to indicate that a function is
2723 an interrupt handler, which tells the compiler to generate function
2724 entry and exit sequences that differ from those from regular
2725 functions. The exact syntax and behavior are target-specific;
2726 refer to the following subsections for details.
2729 @cindex @code{leaf} function attribute
2730 Calls to external functions with this attribute must return to the current
2731 compilation unit only by return or by exception handling. In particular, leaf
2732 functions are not allowed to call callback function passed to it from the current
2733 compilation unit or directly call functions exported by the unit or longjmp
2734 into the unit. Leaf function might still call functions from other compilation
2735 units and thus they are not necessarily leaf in the sense that they contain no
2736 function calls at all.
2738 The attribute is intended for library functions to improve dataflow analysis.
2739 The compiler takes the hint that any data not escaping the current compilation unit can
2740 not be used or modified by the leaf function. For example, the @code{sin} function
2741 is a leaf function, but @code{qsort} is not.
2743 Note that leaf functions might invoke signals and signal handlers might be
2744 defined in the current compilation unit and use static variables. The only
2745 compliant way to write such a signal handler is to declare such variables
2748 The attribute has no effect on functions defined within the current compilation
2749 unit. This is to allow easy merging of multiple compilation units into one,
2750 for example, by using the link-time optimization. For this reason the
2751 attribute is not allowed on types to annotate indirect calls.
2755 @cindex @code{malloc} function attribute
2756 @cindex functions that behave like malloc
2757 This tells the compiler that a function is @code{malloc}-like, i.e.,
2758 that the pointer @var{P} returned by the function cannot alias any
2759 other pointer valid when the function returns, and moreover no
2760 pointers to valid objects occur in any storage addressed by @var{P}.
2762 Using this attribute can improve optimization. Functions like
2763 @code{malloc} and @code{calloc} have this property because they return
2764 a pointer to uninitialized or zeroed-out storage. However, functions
2765 like @code{realloc} do not have this property, as they can return a
2766 pointer to storage containing pointers.
2769 @cindex @code{no_icf} function attribute
2770 This function attribute prevents a functions from being merged with another
2771 semantically equivalent function.
2773 @item no_instrument_function
2774 @cindex @code{no_instrument_function} function attribute
2775 @opindex finstrument-functions
2776 If @option{-finstrument-functions} is given, profiling function calls are
2777 generated at entry and exit of most user-compiled functions.
2778 Functions with this attribute are not so instrumented.
2781 @cindex @code{no_reorder} function attribute
2782 Do not reorder functions or variables marked @code{no_reorder}
2783 against each other or top level assembler statements the executable.
2784 The actual order in the program will depend on the linker command
2785 line. Static variables marked like this are also not removed.
2786 This has a similar effect
2787 as the @option{-fno-toplevel-reorder} option, but only applies to the
2790 @item no_sanitize_address
2791 @itemx no_address_safety_analysis
2792 @cindex @code{no_sanitize_address} function attribute
2793 The @code{no_sanitize_address} attribute on functions is used
2794 to inform the compiler that it should not instrument memory accesses
2795 in the function when compiling with the @option{-fsanitize=address} option.
2796 The @code{no_address_safety_analysis} is a deprecated alias of the
2797 @code{no_sanitize_address} attribute, new code should use
2798 @code{no_sanitize_address}.
2800 @item no_sanitize_thread
2801 @cindex @code{no_sanitize_thread} function attribute
2802 The @code{no_sanitize_thread} attribute on functions is used
2803 to inform the compiler that it should not instrument memory accesses
2804 in the function when compiling with the @option{-fsanitize=thread} option.
2806 @item no_sanitize_undefined
2807 @cindex @code{no_sanitize_undefined} function attribute
2808 The @code{no_sanitize_undefined} attribute on functions is used
2809 to inform the compiler that it should not check for undefined behavior
2810 in the function when compiling with the @option{-fsanitize=undefined} option.
2812 @item no_split_stack
2813 @cindex @code{no_split_stack} function attribute
2814 @opindex fsplit-stack
2815 If @option{-fsplit-stack} is given, functions have a small
2816 prologue which decides whether to split the stack. Functions with the
2817 @code{no_split_stack} attribute do not have that prologue, and thus
2818 may run with only a small amount of stack space available.
2821 @cindex @code{noclone} function attribute
2822 This function attribute prevents a function from being considered for
2823 cloning---a mechanism that produces specialized copies of functions
2824 and which is (currently) performed by interprocedural constant
2828 @cindex @code{noinline} function attribute
2829 This function attribute prevents a function from being considered for
2831 @c Don't enumerate the optimizations by name here; we try to be
2832 @c future-compatible with this mechanism.
2833 If the function does not have side-effects, there are optimizations
2834 other than inlining that cause function calls to be optimized away,
2835 although the function call is live. To keep such calls from being
2842 (@pxref{Extended Asm}) in the called function, to serve as a special
2845 @item nonnull (@var{arg-index}, @dots{})
2846 @cindex @code{nonnull} function attribute
2847 @cindex functions with non-null pointer arguments
2848 The @code{nonnull} attribute specifies that some function parameters should
2849 be non-null pointers. For instance, the declaration:
2853 my_memcpy (void *dest, const void *src, size_t len)
2854 __attribute__((nonnull (1, 2)));
2858 causes the compiler to check that, in calls to @code{my_memcpy},
2859 arguments @var{dest} and @var{src} are non-null. If the compiler
2860 determines that a null pointer is passed in an argument slot marked
2861 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2862 is issued. The compiler may also choose to make optimizations based
2863 on the knowledge that certain function arguments will never be null.
2865 If no argument index list is given to the @code{nonnull} attribute,
2866 all pointer arguments are marked as non-null. To illustrate, the
2867 following declaration is equivalent to the previous example:
2871 my_memcpy (void *dest, const void *src, size_t len)
2872 __attribute__((nonnull));
2876 @cindex @code{noreturn} function attribute
2877 @cindex functions that never return
2878 A few standard library functions, such as @code{abort} and @code{exit},
2879 cannot return. GCC knows this automatically. Some programs define
2880 their own functions that never return. You can declare them
2881 @code{noreturn} to tell the compiler this fact. For example,
2885 void fatal () __attribute__ ((noreturn));
2888 fatal (/* @r{@dots{}} */)
2890 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2896 The @code{noreturn} keyword tells the compiler to assume that
2897 @code{fatal} cannot return. It can then optimize without regard to what
2898 would happen if @code{fatal} ever did return. This makes slightly
2899 better code. More importantly, it helps avoid spurious warnings of
2900 uninitialized variables.
2902 The @code{noreturn} keyword does not affect the exceptional path when that
2903 applies: a @code{noreturn}-marked function may still return to the caller
2904 by throwing an exception or calling @code{longjmp}.
2906 Do not assume that registers saved by the calling function are
2907 restored before calling the @code{noreturn} function.
2909 It does not make sense for a @code{noreturn} function to have a return
2910 type other than @code{void}.
2913 @cindex @code{nothrow} function attribute
2914 The @code{nothrow} attribute is used to inform the compiler that a
2915 function cannot throw an exception. For example, most functions in
2916 the standard C library can be guaranteed not to throw an exception
2917 with the notable exceptions of @code{qsort} and @code{bsearch} that
2918 take function pointer arguments.
2921 @cindex @code{noplt} function attribute
2922 The @code{noplt} attribute is the counterpart to option @option{-fno-plt} and
2923 does not use PLT for calls to functions marked with this attribute in position
2928 /* Externally defined function foo. */
2929 int foo () __attribute__ ((noplt));
2932 main (/* @r{@dots{}} */)
2941 The @code{noplt} attribute on function foo tells the compiler to assume that
2942 the function foo is externally defined and the call to foo must avoid the PLT
2943 in position independent code.
2945 Additionally, a few targets also convert calls to those functions that are
2946 marked to not use the PLT to use the GOT instead for non-position independent
2950 @cindex @code{optimize} function attribute
2951 The @code{optimize} attribute is used to specify that a function is to
2952 be compiled with different optimization options than specified on the
2953 command line. Arguments can either be numbers or strings. Numbers
2954 are assumed to be an optimization level. Strings that begin with
2955 @code{O} are assumed to be an optimization option, while other options
2956 are assumed to be used with a @code{-f} prefix. You can also use the
2957 @samp{#pragma GCC optimize} pragma to set the optimization options
2958 that affect more than one function.
2959 @xref{Function Specific Option Pragmas}, for details about the
2960 @samp{#pragma GCC optimize} pragma.
2962 This can be used for instance to have frequently-executed functions
2963 compiled with more aggressive optimization options that produce faster
2964 and larger code, while other functions can be compiled with less
2968 @cindex @code{pure} function attribute
2969 @cindex functions that have no side effects
2970 Many functions have no effects except the return value and their
2971 return value depends only on the parameters and/or global variables.
2972 Such a function can be subject
2973 to common subexpression elimination and loop optimization just as an
2974 arithmetic operator would be. These functions should be declared
2975 with the attribute @code{pure}. For example,
2978 int square (int) __attribute__ ((pure));
2982 says that the hypothetical function @code{square} is safe to call
2983 fewer times than the program says.
2985 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2986 Interesting non-pure functions are functions with infinite loops or those
2987 depending on volatile memory or other system resource, that may change between
2988 two consecutive calls (such as @code{feof} in a multithreading environment).
2990 @item returns_nonnull
2991 @cindex @code{returns_nonnull} function attribute
2992 The @code{returns_nonnull} attribute specifies that the function
2993 return value should be a non-null pointer. For instance, the declaration:
2997 mymalloc (size_t len) __attribute__((returns_nonnull));
3001 lets the compiler optimize callers based on the knowledge
3002 that the return value will never be null.
3005 @cindex @code{returns_twice} function attribute
3006 @cindex functions that return more than once
3007 The @code{returns_twice} attribute tells the compiler that a function may
3008 return more than one time. The compiler ensures that all registers
3009 are dead before calling such a function and emits a warning about
3010 the variables that may be clobbered after the second return from the
3011 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3012 The @code{longjmp}-like counterpart of such function, if any, might need
3013 to be marked with the @code{noreturn} attribute.
3015 @item section ("@var{section-name}")
3016 @cindex @code{section} function attribute
3017 @cindex functions in arbitrary sections
3018 Normally, the compiler places the code it generates in the @code{text} section.
3019 Sometimes, however, you need additional sections, or you need certain
3020 particular functions to appear in special sections. The @code{section}
3021 attribute specifies that a function lives in a particular section.
3022 For example, the declaration:
3025 extern void foobar (void) __attribute__ ((section ("bar")));
3029 puts the function @code{foobar} in the @code{bar} section.
3031 Some file formats do not support arbitrary sections so the @code{section}
3032 attribute is not available on all platforms.
3033 If you need to map the entire contents of a module to a particular
3034 section, consider using the facilities of the linker instead.
3037 @cindex @code{sentinel} function attribute
3038 This function attribute ensures that a parameter in a function call is
3039 an explicit @code{NULL}. The attribute is only valid on variadic
3040 functions. By default, the sentinel is located at position zero, the
3041 last parameter of the function call. If an optional integer position
3042 argument P is supplied to the attribute, the sentinel must be located at
3043 position P counting backwards from the end of the argument list.
3046 __attribute__ ((sentinel))
3048 __attribute__ ((sentinel(0)))
3051 The attribute is automatically set with a position of 0 for the built-in
3052 functions @code{execl} and @code{execlp}. The built-in function
3053 @code{execle} has the attribute set with a position of 1.
3055 A valid @code{NULL} in this context is defined as zero with any pointer
3056 type. If your system defines the @code{NULL} macro with an integer type
3057 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3058 with a copy that redefines NULL appropriately.
3060 The warnings for missing or incorrect sentinels are enabled with
3064 @cindex @code{stack_protect} function attribute
3065 This function attribute make a stack protection of the function if
3066 flags @option{fstack-protector} or @option{fstack-protector-strong}
3067 or @option{fstack-protector-explicit} are set.
3069 @item target (@var{options})
3070 @cindex @code{target} function attribute
3071 Multiple target back ends implement the @code{target} attribute
3072 to specify that a function is to
3073 be compiled with different target options than specified on the
3074 command line. This can be used for instance to have functions
3075 compiled with a different ISA (instruction set architecture) than the
3076 default. You can also use the @samp{#pragma GCC target} pragma to set
3077 more than one function to be compiled with specific target options.
3078 @xref{Function Specific Option Pragmas}, for details about the
3079 @samp{#pragma GCC target} pragma.
3081 For instance, on an x86, you could declare one function with the
3082 @code{target("sse4.1,arch=core2")} attribute and another with
3083 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3084 compiling the first function with @option{-msse4.1} and
3085 @option{-march=core2} options, and the second function with
3086 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3087 to make sure that a function is only invoked on a machine that
3088 supports the particular ISA it is compiled for (for example by using
3089 @code{cpuid} on x86 to determine what feature bits and architecture
3093 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3094 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3097 You can either use multiple
3098 strings separated by commas to specify multiple options,
3099 or separate the options with a comma (@samp{,}) within a single string.
3101 The options supported are specific to each target; refer to @ref{x86
3102 Function Attributes}, @ref{PowerPC Function Attributes},
3103 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3107 @cindex @code{unused} function attribute
3108 This attribute, attached to a function, means that the function is meant
3109 to be possibly unused. GCC does not produce a warning for this
3113 @cindex @code{used} function attribute
3114 This attribute, attached to a function, means that code must be emitted
3115 for the function even if it appears that the function is not referenced.
3116 This is useful, for example, when the function is referenced only in
3119 When applied to a member function of a C++ class template, the
3120 attribute also means that the function is instantiated if the
3121 class itself is instantiated.
3123 @item visibility ("@var{visibility_type}")
3124 @cindex @code{visibility} function attribute
3125 This attribute affects the linkage of the declaration to which it is attached.
3126 There are four supported @var{visibility_type} values: default,
3127 hidden, protected or internal visibility.
3130 void __attribute__ ((visibility ("protected")))
3131 f () @{ /* @r{Do something.} */; @}
3132 int i __attribute__ ((visibility ("hidden")));
3135 The possible values of @var{visibility_type} correspond to the
3136 visibility settings in the ELF gABI.
3139 @c keep this list of visibilities in alphabetical order.
3142 Default visibility is the normal case for the object file format.
3143 This value is available for the visibility attribute to override other
3144 options that may change the assumed visibility of entities.
3146 On ELF, default visibility means that the declaration is visible to other
3147 modules and, in shared libraries, means that the declared entity may be
3150 On Darwin, default visibility means that the declaration is visible to
3153 Default visibility corresponds to ``external linkage'' in the language.
3156 Hidden visibility indicates that the entity declared has a new
3157 form of linkage, which we call ``hidden linkage''. Two
3158 declarations of an object with hidden linkage refer to the same object
3159 if they are in the same shared object.
3162 Internal visibility is like hidden visibility, but with additional
3163 processor specific semantics. Unless otherwise specified by the
3164 psABI, GCC defines internal visibility to mean that a function is
3165 @emph{never} called from another module. Compare this with hidden
3166 functions which, while they cannot be referenced directly by other
3167 modules, can be referenced indirectly via function pointers. By
3168 indicating that a function cannot be called from outside the module,
3169 GCC may for instance omit the load of a PIC register since it is known
3170 that the calling function loaded the correct value.
3173 Protected visibility is like default visibility except that it
3174 indicates that references within the defining module bind to the
3175 definition in that module. That is, the declared entity cannot be
3176 overridden by another module.
3180 All visibilities are supported on many, but not all, ELF targets
3181 (supported when the assembler supports the @samp{.visibility}
3182 pseudo-op). Default visibility is supported everywhere. Hidden
3183 visibility is supported on Darwin targets.
3185 The visibility attribute should be applied only to declarations that
3186 would otherwise have external linkage. The attribute should be applied
3187 consistently, so that the same entity should not be declared with
3188 different settings of the attribute.
3190 In C++, the visibility attribute applies to types as well as functions
3191 and objects, because in C++ types have linkage. A class must not have
3192 greater visibility than its non-static data member types and bases,
3193 and class members default to the visibility of their class. Also, a
3194 declaration without explicit visibility is limited to the visibility
3197 In C++, you can mark member functions and static member variables of a
3198 class with the visibility attribute. This is useful if you know a
3199 particular method or static member variable should only be used from
3200 one shared object; then you can mark it hidden while the rest of the
3201 class has default visibility. Care must be taken to avoid breaking
3202 the One Definition Rule; for example, it is usually not useful to mark
3203 an inline method as hidden without marking the whole class as hidden.
3205 A C++ namespace declaration can also have the visibility attribute.
3208 namespace nspace1 __attribute__ ((visibility ("protected")))
3209 @{ /* @r{Do something.} */; @}
3212 This attribute applies only to the particular namespace body, not to
3213 other definitions of the same namespace; it is equivalent to using
3214 @samp{#pragma GCC visibility} before and after the namespace
3215 definition (@pxref{Visibility Pragmas}).
3217 In C++, if a template argument has limited visibility, this
3218 restriction is implicitly propagated to the template instantiation.
3219 Otherwise, template instantiations and specializations default to the
3220 visibility of their template.
3222 If both the template and enclosing class have explicit visibility, the
3223 visibility from the template is used.
3225 @item warn_unused_result
3226 @cindex @code{warn_unused_result} function attribute
3227 The @code{warn_unused_result} attribute causes a warning to be emitted
3228 if a caller of the function with this attribute does not use its
3229 return value. This is useful for functions where not checking
3230 the result is either a security problem or always a bug, such as
3234 int fn () __attribute__ ((warn_unused_result));
3237 if (fn () < 0) return -1;
3244 results in warning on line 5.
3247 @cindex @code{weak} function attribute
3248 The @code{weak} attribute causes the declaration to be emitted as a weak
3249 symbol rather than a global. This is primarily useful in defining
3250 library functions that can be overridden in user code, though it can
3251 also be used with non-function declarations. Weak symbols are supported
3252 for ELF targets, and also for a.out targets when using the GNU assembler
3256 @itemx weakref ("@var{target}")
3257 @cindex @code{weakref} function attribute
3258 The @code{weakref} attribute marks a declaration as a weak reference.
3259 Without arguments, it should be accompanied by an @code{alias} attribute
3260 naming the target symbol. Optionally, the @var{target} may be given as
3261 an argument to @code{weakref} itself. In either case, @code{weakref}
3262 implicitly marks the declaration as @code{weak}. Without a
3263 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3264 @code{weakref} is equivalent to @code{weak}.
3267 static int x() __attribute__ ((weakref ("y")));
3268 /* is equivalent to... */
3269 static int x() __attribute__ ((weak, weakref, alias ("y")));
3271 static int x() __attribute__ ((weakref));
3272 static int x() __attribute__ ((alias ("y")));
3275 A weak reference is an alias that does not by itself require a
3276 definition to be given for the target symbol. If the target symbol is
3277 only referenced through weak references, then it becomes a @code{weak}
3278 undefined symbol. If it is directly referenced, however, then such
3279 strong references prevail, and a definition is required for the
3280 symbol, not necessarily in the same translation unit.
3282 The effect is equivalent to moving all references to the alias to a
3283 separate translation unit, renaming the alias to the aliased symbol,
3284 declaring it as weak, compiling the two separate translation units and
3285 performing a reloadable link on them.
3287 At present, a declaration to which @code{weakref} is attached can
3288 only be @code{static}.
3293 @cindex lower memory region on the MSP430
3294 @cindex upper memory region on the MSP430
3295 @cindex either memory region on the MSP430
3296 On the MSP430 target these attributes can be used to specify whether
3297 the function or variable should be placed into low memory, high
3298 memory, or the placement should be left to the linker to decide. The
3299 attributes are only significant if compiling for the MSP430X
3302 The attributes work in conjunction with a linker script that has been
3303 augmented to specify where to place sections with a @code{.lower} and
3304 a @code{.upper} prefix. So for example as well as placing the
3305 @code{.data} section the script would also specify the placement of a
3306 @code{.lower.data} and a @code{.upper.data} section. The intention
3307 being that @code{lower} sections are placed into a small but easier to
3308 access memory region and the upper sections are placed into a larger, but
3309 slower to access region.
3311 The @code{either} attribute is special. It tells the linker to place
3312 the object into the corresponding @code{lower} section if there is
3313 room for it. If there is insufficient room then the object is placed
3314 into the corresponding @code{upper} section instead. Note - the
3315 placement algorithm is not very sophisticated. It will not attempt to
3316 find an optimal packing of the @code{lower} sections. It just makes
3317 one pass over the objects and does the best that it can. Using the
3318 @option{-ffunction-sections} and @option{-fdata-sections} command line
3319 options can help the packing however, since they produce smaller,
3320 easier to pack regions.
3324 @c This is the end of the target-independent attribute table
3326 @node AArch64 Function Attributes
3327 @subsection AArch64 Function Attributes
3329 The following target-specific function attributes are available for the
3330 AArch64 target. For the most part, these options mirror the behavior of
3331 similar command-line options (@pxref{AArch64 Options}), but on a
3335 @item general-regs-only
3336 @cindex @code{general-regs-only} function attribute, AArch64
3337 Indicates that no floating-point or Advanced SIMD registers should be
3338 used when generating code for this function. If the function explicitly
3339 uses floating-point code, then the compiler gives an error. This is
3340 the same behavior as that of the command-line option
3341 @option{-mgeneral-regs-only}.
3343 @item fix-cortex-a53-835769
3344 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3345 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3346 applied to this function. To explicitly disable the workaround for this
3347 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3348 This corresponds to the behavior of the command line options
3349 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3352 @cindex @code{cmodel=} function attribute, AArch64
3353 Indicates that code should be generated for a particular code model for
3354 this function. The behavior and permissible arguments are the same as
3355 for the command line option @option{-mcmodel=}.
3358 @cindex @code{strict-align} function attribute, AArch64
3359 Indicates that the compiler should not assume that unaligned memory references
3360 are handled by the system. The behavior is the same as for the command-line
3361 option @option{-mstrict-align}.
3363 @item omit-leaf-frame-pointer
3364 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3365 Indicates that the frame pointer should be omitted for a leaf function call.
3366 To keep the frame pointer, the inverse attribute
3367 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3368 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3369 and @option{-mno-omit-leaf-frame-pointer}.
3372 @cindex @code{tls-dialect=} function attribute, AArch64
3373 Specifies the TLS dialect to use for this function. The behavior and
3374 permissible arguments are the same as for the command-line option
3375 @option{-mtls-dialect=}.
3378 @cindex @code{arch=} function attribute, AArch64
3379 Specifies the architecture version and architectural extensions to use
3380 for this function. The behavior and permissible arguments are the same as
3381 for the @option{-march=} command-line option.
3384 @cindex @code{tune=} function attribute, AArch64
3385 Specifies the core for which to tune the performance of this function.
3386 The behavior and permissible arguments are the same as for the @option{-mtune=}
3387 command-line option.
3390 @cindex @code{cpu=} function attribute, AArch64
3391 Specifies the core for which to tune the performance of this function and also
3392 whose architectural features to use. The behavior and valid arguments are the
3393 same as for the @option{-mcpu=} command-line option.
3397 The above target attributes can be specified as follows:
3400 __attribute__((target("@var{attr-string}")))
3408 where @code{@var{attr-string}} is one of the attribute strings specified above.
3410 Additionally, the architectural extension string may be specified on its
3411 own. This can be used to turn on and off particular architectural extensions
3412 without having to specify a particular architecture version or core. Example:
3415 __attribute__((target("+crc+nocrypto")))
3423 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3424 extension and disables the @code{crypto} extension for the function @code{foo}
3425 without modifying an existing @option{-march=} or @option{-mcpu} option.
3427 Multiple target function attributes can be specified by separating them with
3428 a comma. For example:
3430 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3438 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3439 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3441 @subsubsection Inlining rules
3442 Specifying target attributes on individual functions or performing link-time
3443 optimization across translation units compiled with different target options
3444 can affect function inlining rules:
3446 In particular, a caller function can inline a callee function only if the
3447 architectural features available to the callee are a subset of the features
3448 available to the caller.
3449 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3450 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3451 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3452 because the all the architectural features that function @code{bar} requires
3453 are available to function @code{foo}. Conversely, function @code{bar} cannot
3454 inline function @code{foo}.
3456 Additionally inlining a function compiled with @option{-mstrict-align} into a
3457 function compiled without @code{-mstrict-align} is not allowed.
3458 However, inlining a function compiled without @option{-mstrict-align} into a
3459 function compiled with @option{-mstrict-align} is allowed.
3461 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3462 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3463 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3464 architectural feature rules specified above.
3466 @node ARC Function Attributes
3467 @subsection ARC Function Attributes
3469 These function attributes are supported by the ARC back end:
3473 @cindex @code{interrupt} function attribute, ARC
3474 Use this attribute to indicate
3475 that the specified function is an interrupt handler. The compiler generates
3476 function entry and exit sequences suitable for use in an interrupt handler
3477 when this attribute is present.
3479 On the ARC, you must specify the kind of interrupt to be handled
3480 in a parameter to the interrupt attribute like this:
3483 void f () __attribute__ ((interrupt ("ilink1")));
3486 Permissible values for this parameter are: @w{@code{ilink1}} and
3492 @cindex @code{long_call} function attribute, ARC
3493 @cindex @code{medium_call} function attribute, ARC
3494 @cindex @code{short_call} function attribute, ARC
3495 @cindex indirect calls, ARC
3496 These attributes specify how a particular function is called.
3497 These attributes override the
3498 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3499 command-line switches and @code{#pragma long_calls} settings.
3501 For ARC, a function marked with the @code{long_call} attribute is
3502 always called using register-indirect jump-and-link instructions,
3503 thereby enabling the called function to be placed anywhere within the
3504 32-bit address space. A function marked with the @code{medium_call}
3505 attribute will always be close enough to be called with an unconditional
3506 branch-and-link instruction, which has a 25-bit offset from
3507 the call site. A function marked with the @code{short_call}
3508 attribute will always be close enough to be called with a conditional
3509 branch-and-link instruction, which has a 21-bit offset from
3513 @node ARM Function Attributes
3514 @subsection ARM Function Attributes
3516 These function attributes are supported for ARM targets:
3520 @cindex @code{interrupt} function attribute, ARM
3521 Use this attribute to indicate
3522 that the specified function is an interrupt handler. The compiler generates
3523 function entry and exit sequences suitable for use in an interrupt handler
3524 when this attribute is present.
3526 You can specify the kind of interrupt to be handled by
3527 adding an optional parameter to the interrupt attribute like this:
3530 void f () __attribute__ ((interrupt ("IRQ")));
3534 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3535 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3537 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3538 may be called with a word-aligned stack pointer.
3541 @cindex @code{isr} function attribute, ARM
3542 Use this attribute on ARM to write Interrupt Service Routines. This is an
3543 alias to the @code{interrupt} attribute above.
3547 @cindex @code{long_call} function attribute, ARM
3548 @cindex @code{short_call} function attribute, ARM
3549 @cindex indirect calls, ARM
3550 These attributes specify how a particular function is called.
3551 These attributes override the
3552 @option{-mlong-calls} (@pxref{ARM Options})
3553 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3554 @code{long_call} attribute indicates that the function might be far
3555 away from the call site and require a different (more expensive)
3556 calling sequence. The @code{short_call} attribute always places
3557 the offset to the function from the call site into the @samp{BL}
3558 instruction directly.
3561 @cindex @code{naked} function attribute, ARM
3562 This attribute allows the compiler to construct the
3563 requisite function declaration, while allowing the body of the
3564 function to be assembly code. The specified function will not have
3565 prologue/epilogue sequences generated by the compiler. Only basic
3566 @code{asm} statements can safely be included in naked functions
3567 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3568 basic @code{asm} and C code may appear to work, they cannot be
3569 depended upon to work reliably and are not supported.
3572 @cindex @code{pcs} function attribute, ARM
3574 The @code{pcs} attribute can be used to control the calling convention
3575 used for a function on ARM. The attribute takes an argument that specifies
3576 the calling convention to use.
3578 When compiling using the AAPCS ABI (or a variant of it) then valid
3579 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3580 order to use a variant other than @code{"aapcs"} then the compiler must
3581 be permitted to use the appropriate co-processor registers (i.e., the
3582 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3586 /* Argument passed in r0, and result returned in r0+r1. */
3587 double f2d (float) __attribute__((pcs("aapcs")));
3590 Variadic functions always use the @code{"aapcs"} calling convention and
3591 the compiler rejects attempts to specify an alternative.
3593 @item target (@var{options})
3594 @cindex @code{target} function attribute
3595 As discussed in @ref{Common Function Attributes}, this attribute
3596 allows specification of target-specific compilation options.
3598 On ARM, the following options are allowed:
3602 @cindex @code{target("thumb")} function attribute, ARM
3603 Force code generation in the Thumb (T16/T32) ISA, depending on the
3607 @cindex @code{target("arm")} function attribute, ARM
3608 Force code generation in the ARM (A32) ISA.
3611 Functions from different modes can be inlined in the caller's mode.
3615 @node AVR Function Attributes
3616 @subsection AVR Function Attributes
3618 These function attributes are supported by the AVR back end:
3622 @cindex @code{interrupt} function attribute, AVR
3623 Use this attribute to indicate
3624 that the specified function is an interrupt handler. The compiler generates
3625 function entry and exit sequences suitable for use in an interrupt handler
3626 when this attribute is present.
3628 On the AVR, the hardware globally disables interrupts when an
3629 interrupt is executed. The first instruction of an interrupt handler
3630 declared with this attribute is a @code{SEI} instruction to
3631 re-enable interrupts. See also the @code{signal} function attribute
3632 that does not insert a @code{SEI} instruction. If both @code{signal} and
3633 @code{interrupt} are specified for the same function, @code{signal}
3634 is silently ignored.
3637 @cindex @code{naked} function attribute, AVR
3638 This attribute allows the compiler to construct the
3639 requisite function declaration, while allowing the body of the
3640 function to be assembly code. The specified function will not have
3641 prologue/epilogue sequences generated by the compiler. Only basic
3642 @code{asm} statements can safely be included in naked functions
3643 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3644 basic @code{asm} and C code may appear to work, they cannot be
3645 depended upon to work reliably and are not supported.
3649 @cindex @code{OS_main} function attribute, AVR
3650 @cindex @code{OS_task} function attribute, AVR
3651 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3652 do not save/restore any call-saved register in their prologue/epilogue.
3654 The @code{OS_main} attribute can be used when there @emph{is
3655 guarantee} that interrupts are disabled at the time when the function
3656 is entered. This saves resources when the stack pointer has to be
3657 changed to set up a frame for local variables.
3659 The @code{OS_task} attribute can be used when there is @emph{no
3660 guarantee} that interrupts are disabled at that time when the function
3661 is entered like for, e@.g@. task functions in a multi-threading operating
3662 system. In that case, changing the stack pointer register is
3663 guarded by save/clear/restore of the global interrupt enable flag.
3665 The differences to the @code{naked} function attribute are:
3667 @item @code{naked} functions do not have a return instruction whereas
3668 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3669 @code{RETI} return instruction.
3670 @item @code{naked} functions do not set up a frame for local variables
3671 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3676 @cindex @code{signal} function attribute, AVR
3677 Use this attribute on the AVR to indicate that the specified
3678 function is an interrupt handler. The compiler generates function
3679 entry and exit sequences suitable for use in an interrupt handler when this
3680 attribute is present.
3682 See also the @code{interrupt} function attribute.
3684 The AVR hardware globally disables interrupts when an interrupt is executed.
3685 Interrupt handler functions defined with the @code{signal} attribute
3686 do not re-enable interrupts. It is save to enable interrupts in a
3687 @code{signal} handler. This ``save'' only applies to the code
3688 generated by the compiler and not to the IRQ layout of the
3689 application which is responsibility of the application.
3691 If both @code{signal} and @code{interrupt} are specified for the same
3692 function, @code{signal} is silently ignored.
3695 @node Blackfin Function Attributes
3696 @subsection Blackfin Function Attributes
3698 These function attributes are supported by the Blackfin back end:
3702 @item exception_handler
3703 @cindex @code{exception_handler} function attribute
3704 @cindex exception handler functions, Blackfin
3705 Use this attribute on the Blackfin to indicate that the specified function
3706 is an exception handler. The compiler generates function entry and
3707 exit sequences suitable for use in an exception handler when this
3708 attribute is present.
3710 @item interrupt_handler
3711 @cindex @code{interrupt_handler} function attribute, Blackfin
3712 Use this attribute to
3713 indicate that the specified function is an interrupt handler. The compiler
3714 generates function entry and exit sequences suitable for use in an
3715 interrupt handler when this attribute is present.
3718 @cindex @code{kspisusp} function attribute, Blackfin
3719 @cindex User stack pointer in interrupts on the Blackfin
3720 When used together with @code{interrupt_handler}, @code{exception_handler}
3721 or @code{nmi_handler}, code is generated to load the stack pointer
3722 from the USP register in the function prologue.
3725 @cindex @code{l1_text} function attribute, Blackfin
3726 This attribute specifies a function to be placed into L1 Instruction
3727 SRAM@. The function is put into a specific section named @code{.l1.text}.
3728 With @option{-mfdpic}, function calls with a such function as the callee
3729 or caller uses inlined PLT.
3732 @cindex @code{l2} function attribute, Blackfin
3733 This attribute specifies a function to be placed into L2
3734 SRAM. The function is put into a specific section named
3735 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3740 @cindex indirect calls, Blackfin
3741 @cindex @code{longcall} function attribute, Blackfin
3742 @cindex @code{shortcall} function attribute, Blackfin
3743 The @code{longcall} attribute
3744 indicates that the function might be far away from the call site and
3745 require a different (more expensive) calling sequence. The
3746 @code{shortcall} attribute indicates that the function is always close
3747 enough for the shorter calling sequence to be used. These attributes
3748 override the @option{-mlongcall} switch.
3751 @cindex @code{nesting} function attribute, Blackfin
3752 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3753 Use this attribute together with @code{interrupt_handler},
3754 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3755 entry code should enable nested interrupts or exceptions.
3758 @cindex @code{nmi_handler} function attribute, Blackfin
3759 @cindex NMI handler functions on the Blackfin processor
3760 Use this attribute on the Blackfin to indicate that the specified function
3761 is an NMI handler. The compiler generates function entry and
3762 exit sequences suitable for use in an NMI handler when this
3763 attribute is present.
3766 @cindex @code{saveall} function attribute, Blackfin
3767 @cindex save all registers on the Blackfin
3768 Use this attribute to indicate that
3769 all registers except the stack pointer should be saved in the prologue
3770 regardless of whether they are used or not.
3773 @node CR16 Function Attributes
3774 @subsection CR16 Function Attributes
3776 These function attributes are supported by the CR16 back end:
3780 @cindex @code{interrupt} function attribute, CR16
3781 Use this attribute to indicate
3782 that the specified function is an interrupt handler. The compiler generates
3783 function entry and exit sequences suitable for use in an interrupt handler
3784 when this attribute is present.
3787 @node Epiphany Function Attributes
3788 @subsection Epiphany Function Attributes
3790 These function attributes are supported by the Epiphany back end:
3794 @cindex @code{disinterrupt} function attribute, Epiphany
3795 This attribute causes the compiler to emit
3796 instructions to disable interrupts for the duration of the given
3799 @item forwarder_section
3800 @cindex @code{forwarder_section} function attribute, Epiphany
3801 This attribute modifies the behavior of an interrupt handler.
3802 The interrupt handler may be in external memory which cannot be
3803 reached by a branch instruction, so generate a local memory trampoline
3804 to transfer control. The single parameter identifies the section where
3805 the trampoline is placed.
3808 @cindex @code{interrupt} function attribute, Epiphany
3809 Use this attribute to indicate
3810 that the specified function is an interrupt handler. The compiler generates
3811 function entry and exit sequences suitable for use in an interrupt handler
3812 when this attribute is present. It may also generate
3813 a special section with code to initialize the interrupt vector table.
3815 On Epiphany targets one or more optional parameters can be added like this:
3818 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3821 Permissible values for these parameters are: @w{@code{reset}},
3822 @w{@code{software_exception}}, @w{@code{page_miss}},
3823 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3824 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3825 Multiple parameters indicate that multiple entries in the interrupt
3826 vector table should be initialized for this function, i.e.@: for each
3827 parameter @w{@var{name}}, a jump to the function is emitted in
3828 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3829 entirely, in which case no interrupt vector table entry is provided.
3831 Note that interrupts are enabled inside the function
3832 unless the @code{disinterrupt} attribute is also specified.
3834 The following examples are all valid uses of these attributes on
3837 void __attribute__ ((interrupt)) universal_handler ();
3838 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3839 void __attribute__ ((interrupt ("dma0, dma1")))
3840 universal_dma_handler ();
3841 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3842 fast_timer_handler ();
3843 void __attribute__ ((interrupt ("dma0, dma1"),
3844 forwarder_section ("tramp")))
3845 external_dma_handler ();
3850 @cindex @code{long_call} function attribute, Epiphany
3851 @cindex @code{short_call} function attribute, Epiphany
3852 @cindex indirect calls, Epiphany
3853 These attributes specify how a particular function is called.
3854 These attributes override the
3855 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3856 command-line switch and @code{#pragma long_calls} settings.
3860 @node H8/300 Function Attributes
3861 @subsection H8/300 Function Attributes
3863 These function attributes are available for H8/300 targets:
3866 @item function_vector
3867 @cindex @code{function_vector} function attribute, H8/300
3868 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3869 that the specified function should be called through the function vector.
3870 Calling a function through the function vector reduces code size; however,
3871 the function vector has a limited size (maximum 128 entries on the H8/300
3872 and 64 entries on the H8/300H and H8S)
3873 and shares space with the interrupt vector.
3875 @item interrupt_handler
3876 @cindex @code{interrupt_handler} function attribute, H8/300
3877 Use this attribute on the H8/300, H8/300H, and H8S to
3878 indicate that the specified function is an interrupt handler. The compiler
3879 generates function entry and exit sequences suitable for use in an
3880 interrupt handler when this attribute is present.
3883 @cindex @code{saveall} function attribute, H8/300
3884 @cindex save all registers on the H8/300, H8/300H, and H8S
3885 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3886 all registers except the stack pointer should be saved in the prologue
3887 regardless of whether they are used or not.
3890 @node IA-64 Function Attributes
3891 @subsection IA-64 Function Attributes
3893 These function attributes are supported on IA-64 targets:
3896 @item syscall_linkage
3897 @cindex @code{syscall_linkage} function attribute, IA-64
3898 This attribute is used to modify the IA-64 calling convention by marking
3899 all input registers as live at all function exits. This makes it possible
3900 to restart a system call after an interrupt without having to save/restore
3901 the input registers. This also prevents kernel data from leaking into
3905 @cindex @code{version_id} function attribute, IA-64
3906 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
3907 symbol to contain a version string, thus allowing for function level
3908 versioning. HP-UX system header files may use function level versioning
3909 for some system calls.
3912 extern int foo () __attribute__((version_id ("20040821")));
3916 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
3919 @node M32C Function Attributes
3920 @subsection M32C Function Attributes
3922 These function attributes are supported by the M32C back end:
3926 @cindex @code{bank_switch} function attribute, M32C
3927 When added to an interrupt handler with the M32C port, causes the
3928 prologue and epilogue to use bank switching to preserve the registers
3929 rather than saving them on the stack.
3931 @item fast_interrupt
3932 @cindex @code{fast_interrupt} function attribute, M32C
3933 Use this attribute on the M32C port to indicate that the specified
3934 function is a fast interrupt handler. This is just like the
3935 @code{interrupt} attribute, except that @code{freit} is used to return
3936 instead of @code{reit}.
3938 @item function_vector
3939 @cindex @code{function_vector} function attribute, M16C/M32C
3940 On M16C/M32C targets, the @code{function_vector} attribute declares a
3941 special page subroutine call function. Use of this attribute reduces
3942 the code size by 2 bytes for each call generated to the
3943 subroutine. The argument to the attribute is the vector number entry
3944 from the special page vector table which contains the 16 low-order
3945 bits of the subroutine's entry address. Each vector table has special
3946 page number (18 to 255) that is used in @code{jsrs} instructions.
3947 Jump addresses of the routines are generated by adding 0x0F0000 (in
3948 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
3949 2-byte addresses set in the vector table. Therefore you need to ensure
3950 that all the special page vector routines should get mapped within the
3951 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
3954 In the following example 2 bytes are saved for each call to
3955 function @code{foo}.
3958 void foo (void) __attribute__((function_vector(0x18)));
3969 If functions are defined in one file and are called in another file,
3970 then be sure to write this declaration in both files.
3972 This attribute is ignored for R8C target.
3975 @cindex @code{interrupt} function attribute, M32C
3976 Use this attribute to indicate
3977 that the specified function is an interrupt handler. The compiler generates
3978 function entry and exit sequences suitable for use in an interrupt handler
3979 when this attribute is present.
3982 @node M32R/D Function Attributes
3983 @subsection M32R/D Function Attributes
3985 These function attributes are supported by the M32R/D back end:
3989 @cindex @code{interrupt} function attribute, M32R/D
3990 Use this attribute to indicate
3991 that the specified function is an interrupt handler. The compiler generates
3992 function entry and exit sequences suitable for use in an interrupt handler
3993 when this attribute is present.
3995 @item model (@var{model-name})
3996 @cindex @code{model} function attribute, M32R/D
3997 @cindex function addressability on the M32R/D
3999 On the M32R/D, use this attribute to set the addressability of an
4000 object, and of the code generated for a function. The identifier
4001 @var{model-name} is one of @code{small}, @code{medium}, or
4002 @code{large}, representing each of the code models.
4004 Small model objects live in the lower 16MB of memory (so that their
4005 addresses can be loaded with the @code{ld24} instruction), and are
4006 callable with the @code{bl} instruction.
4008 Medium model objects may live anywhere in the 32-bit address space (the
4009 compiler generates @code{seth/add3} instructions to load their addresses),
4010 and are callable with the @code{bl} instruction.
4012 Large model objects may live anywhere in the 32-bit address space (the
4013 compiler generates @code{seth/add3} instructions to load their addresses),
4014 and may not be reachable with the @code{bl} instruction (the compiler
4015 generates the much slower @code{seth/add3/jl} instruction sequence).
4018 @node m68k Function Attributes
4019 @subsection m68k Function Attributes
4021 These function attributes are supported by the m68k back end:
4025 @itemx interrupt_handler
4026 @cindex @code{interrupt} function attribute, m68k
4027 @cindex @code{interrupt_handler} function attribute, m68k
4028 Use this attribute to
4029 indicate that the specified function is an interrupt handler. The compiler
4030 generates function entry and exit sequences suitable for use in an
4031 interrupt handler when this attribute is present. Either name may be used.
4033 @item interrupt_thread
4034 @cindex @code{interrupt_thread} function attribute, fido
4035 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4036 that the specified function is an interrupt handler that is designed
4037 to run as a thread. The compiler omits generate prologue/epilogue
4038 sequences and replaces the return instruction with a @code{sleep}
4039 instruction. This attribute is available only on fido.
4042 @node MCORE Function Attributes
4043 @subsection MCORE Function Attributes
4045 These function attributes are supported by the MCORE back end:
4049 @cindex @code{naked} function attribute, MCORE
4050 This attribute allows the compiler to construct the
4051 requisite function declaration, while allowing the body of the
4052 function to be assembly code. The specified function will not have
4053 prologue/epilogue sequences generated by the compiler. Only basic
4054 @code{asm} statements can safely be included in naked functions
4055 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4056 basic @code{asm} and C code may appear to work, they cannot be
4057 depended upon to work reliably and are not supported.
4060 @node MeP Function Attributes
4061 @subsection MeP Function Attributes
4063 These function attributes are supported by the MeP back end:
4067 @cindex @code{disinterrupt} function attribute, MeP
4068 On MeP targets, this attribute causes the compiler to emit
4069 instructions to disable interrupts for the duration of the given
4073 @cindex @code{interrupt} function attribute, MeP
4074 Use this attribute to indicate
4075 that the specified function is an interrupt handler. The compiler generates
4076 function entry and exit sequences suitable for use in an interrupt handler
4077 when this attribute is present.
4080 @cindex @code{near} function attribute, MeP
4081 This attribute causes the compiler to assume the called
4082 function is close enough to use the normal calling convention,
4083 overriding the @option{-mtf} command-line option.
4086 @cindex @code{far} function attribute, MeP
4087 On MeP targets this causes the compiler to use a calling convention
4088 that assumes the called function is too far away for the built-in
4092 @cindex @code{vliw} function attribute, MeP
4093 The @code{vliw} attribute tells the compiler to emit
4094 instructions in VLIW mode instead of core mode. Note that this
4095 attribute is not allowed unless a VLIW coprocessor has been configured
4096 and enabled through command-line options.
4099 @node MicroBlaze Function Attributes
4100 @subsection MicroBlaze Function Attributes
4102 These function attributes are supported on MicroBlaze targets:
4105 @item save_volatiles
4106 @cindex @code{save_volatiles} function attribute, MicroBlaze
4107 Use this attribute to indicate that the function is
4108 an interrupt handler. All volatile registers (in addition to non-volatile
4109 registers) are saved in the function prologue. If the function is a leaf
4110 function, only volatiles used by the function are saved. A normal function
4111 return is generated instead of a return from interrupt.
4114 @cindex @code{break_handler} function attribute, MicroBlaze
4115 @cindex break handler functions
4116 Use this attribute to indicate that
4117 the specified function is a break handler. The compiler generates function
4118 entry and exit sequences suitable for use in an break handler when this
4119 attribute is present. The return from @code{break_handler} is done through
4120 the @code{rtbd} instead of @code{rtsd}.
4123 void f () __attribute__ ((break_handler));
4127 @node Microsoft Windows Function Attributes
4128 @subsection Microsoft Windows Function Attributes
4130 The following attributes are available on Microsoft Windows and Symbian OS
4135 @cindex @code{dllexport} function attribute
4136 @cindex @code{__declspec(dllexport)}
4137 On Microsoft Windows targets and Symbian OS targets the
4138 @code{dllexport} attribute causes the compiler to provide a global
4139 pointer to a pointer in a DLL, so that it can be referenced with the
4140 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4141 name is formed by combining @code{_imp__} and the function or variable
4144 You can use @code{__declspec(dllexport)} as a synonym for
4145 @code{__attribute__ ((dllexport))} for compatibility with other
4148 On systems that support the @code{visibility} attribute, this
4149 attribute also implies ``default'' visibility. It is an error to
4150 explicitly specify any other visibility.
4152 GCC's default behavior is to emit all inline functions with the
4153 @code{dllexport} attribute. Since this can cause object file-size bloat,
4154 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4155 ignore the attribute for inlined functions unless the
4156 @option{-fkeep-inline-functions} flag is used instead.
4158 The attribute is ignored for undefined symbols.
4160 When applied to C++ classes, the attribute marks defined non-inlined
4161 member functions and static data members as exports. Static consts
4162 initialized in-class are not marked unless they are also defined
4165 For Microsoft Windows targets there are alternative methods for
4166 including the symbol in the DLL's export table such as using a
4167 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4168 the @option{--export-all} linker flag.
4171 @cindex @code{dllimport} function attribute
4172 @cindex @code{__declspec(dllimport)}
4173 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4174 attribute causes the compiler to reference a function or variable via
4175 a global pointer to a pointer that is set up by the DLL exporting the
4176 symbol. The attribute implies @code{extern}. On Microsoft Windows
4177 targets, the pointer name is formed by combining @code{_imp__} and the
4178 function or variable name.
4180 You can use @code{__declspec(dllimport)} as a synonym for
4181 @code{__attribute__ ((dllimport))} for compatibility with other
4184 On systems that support the @code{visibility} attribute, this
4185 attribute also implies ``default'' visibility. It is an error to
4186 explicitly specify any other visibility.
4188 Currently, the attribute is ignored for inlined functions. If the
4189 attribute is applied to a symbol @emph{definition}, an error is reported.
4190 If a symbol previously declared @code{dllimport} is later defined, the
4191 attribute is ignored in subsequent references, and a warning is emitted.
4192 The attribute is also overridden by a subsequent declaration as
4195 When applied to C++ classes, the attribute marks non-inlined
4196 member functions and static data members as imports. However, the
4197 attribute is ignored for virtual methods to allow creation of vtables
4200 On the SH Symbian OS target the @code{dllimport} attribute also has
4201 another affect---it can cause the vtable and run-time type information
4202 for a class to be exported. This happens when the class has a
4203 dllimported constructor or a non-inline, non-pure virtual function
4204 and, for either of those two conditions, the class also has an inline
4205 constructor or destructor and has a key function that is defined in
4206 the current translation unit.
4208 For Microsoft Windows targets the use of the @code{dllimport}
4209 attribute on functions is not necessary, but provides a small
4210 performance benefit by eliminating a thunk in the DLL@. The use of the
4211 @code{dllimport} attribute on imported variables can be avoided by passing the
4212 @option{--enable-auto-import} switch to the GNU linker. As with
4213 functions, using the attribute for a variable eliminates a thunk in
4216 One drawback to using this attribute is that a pointer to a
4217 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4218 address. However, a pointer to a @emph{function} with the
4219 @code{dllimport} attribute can be used as a constant initializer; in
4220 this case, the address of a stub function in the import lib is
4221 referenced. On Microsoft Windows targets, the attribute can be disabled
4222 for functions by setting the @option{-mnop-fun-dllimport} flag.
4225 @node MIPS Function Attributes
4226 @subsection MIPS Function Attributes
4228 These function attributes are supported by the MIPS back end:
4232 @cindex @code{interrupt} function attribute, MIPS
4233 Use this attribute to indicate that the specified function is an interrupt
4234 handler. The compiler generates function entry and exit sequences suitable
4235 for use in an interrupt handler when this attribute is present.
4236 An optional argument is supported for the interrupt attribute which allows
4237 the interrupt mode to be described. By default GCC assumes the external
4238 interrupt controller (EIC) mode is in use, this can be explicitly set using
4239 @code{eic}. When interrupts are non-masked then the requested Interrupt
4240 Priority Level (IPL) is copied to the current IPL which has the effect of only
4241 enabling higher priority interrupts. To use vectored interrupt mode use
4242 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4243 the behaviour of the non-masked interrupt support and GCC will arrange to mask
4244 all interrupts from sw0 up to and including the specified interrupt vector.
4246 You can use the following attributes to modify the behavior
4247 of an interrupt handler:
4249 @item use_shadow_register_set
4250 @cindex @code{use_shadow_register_set} function attribute, MIPS
4251 Assume that the handler uses a shadow register set, instead of
4252 the main general-purpose registers. An optional argument @code{intstack} is
4253 supported to indicate that the shadow register set contains a valid stack
4256 @item keep_interrupts_masked
4257 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4258 Keep interrupts masked for the whole function. Without this attribute,
4259 GCC tries to reenable interrupts for as much of the function as it can.
4261 @item use_debug_exception_return
4262 @cindex @code{use_debug_exception_return} function attribute, MIPS
4263 Return using the @code{deret} instruction. Interrupt handlers that don't
4264 have this attribute return using @code{eret} instead.
4267 You can use any combination of these attributes, as shown below:
4269 void __attribute__ ((interrupt)) v0 ();
4270 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4271 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4272 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4273 void __attribute__ ((interrupt, use_shadow_register_set,
4274 keep_interrupts_masked)) v4 ();
4275 void __attribute__ ((interrupt, use_shadow_register_set,
4276 use_debug_exception_return)) v5 ();
4277 void __attribute__ ((interrupt, keep_interrupts_masked,
4278 use_debug_exception_return)) v6 ();
4279 void __attribute__ ((interrupt, use_shadow_register_set,
4280 keep_interrupts_masked,
4281 use_debug_exception_return)) v7 ();
4282 void __attribute__ ((interrupt("eic"))) v8 ();
4283 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4289 @cindex indirect calls, MIPS
4290 @cindex @code{long_call} function attribute, MIPS
4291 @cindex @code{near} function attribute, MIPS
4292 @cindex @code{far} function attribute, MIPS
4293 These attributes specify how a particular function is called on MIPS@.
4294 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4295 command-line switch. The @code{long_call} and @code{far} attributes are
4296 synonyms, and cause the compiler to always call
4297 the function by first loading its address into a register, and then using
4298 the contents of that register. The @code{near} attribute has the opposite
4299 effect; it specifies that non-PIC calls should be made using the more
4300 efficient @code{jal} instruction.
4304 @cindex @code{mips16} function attribute, MIPS
4305 @cindex @code{nomips16} function attribute, MIPS
4307 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4308 function attributes to locally select or turn off MIPS16 code generation.
4309 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4310 while MIPS16 code generation is disabled for functions with the
4311 @code{nomips16} attribute. These attributes override the
4312 @option{-mips16} and @option{-mno-mips16} options on the command line
4313 (@pxref{MIPS Options}).
4315 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4316 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4317 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4318 may interact badly with some GCC extensions such as @code{__builtin_apply}
4319 (@pxref{Constructing Calls}).
4321 @item micromips, MIPS
4322 @itemx nomicromips, MIPS
4323 @cindex @code{micromips} function attribute
4324 @cindex @code{nomicromips} function attribute
4326 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4327 function attributes to locally select or turn off microMIPS code generation.
4328 A function with the @code{micromips} attribute is emitted as microMIPS code,
4329 while microMIPS code generation is disabled for functions with the
4330 @code{nomicromips} attribute. These attributes override the
4331 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4332 (@pxref{MIPS Options}).
4334 When compiling files containing mixed microMIPS and non-microMIPS code, the
4335 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4337 not that within individual functions. Mixed microMIPS and non-microMIPS code
4338 may interact badly with some GCC extensions such as @code{__builtin_apply}
4339 (@pxref{Constructing Calls}).
4342 @cindex @code{nocompression} function attribute, MIPS
4343 On MIPS targets, you can use the @code{nocompression} function attribute
4344 to locally turn off MIPS16 and microMIPS code generation. This attribute
4345 overrides the @option{-mips16} and @option{-mmicromips} options on the
4346 command line (@pxref{MIPS Options}).
4349 @node MSP430 Function Attributes
4350 @subsection MSP430 Function Attributes
4352 These function attributes are supported by the MSP430 back end:
4356 @cindex @code{critical} function attribute, MSP430
4357 Critical functions disable interrupts upon entry and restore the
4358 previous interrupt state upon exit. Critical functions cannot also
4359 have the @code{naked} or @code{reentrant} attributes. They can have
4360 the @code{interrupt} attribute.
4363 @cindex @code{interrupt} function attribute, MSP430
4364 Use this attribute to indicate
4365 that the specified function is an interrupt handler. The compiler generates
4366 function entry and exit sequences suitable for use in an interrupt handler
4367 when this attribute is present.
4369 You can provide an argument to the interrupt
4370 attribute which specifies a name or number. If the argument is a
4371 number it indicates the slot in the interrupt vector table (0 - 31) to
4372 which this handler should be assigned. If the argument is a name it
4373 is treated as a symbolic name for the vector slot. These names should
4374 match up with appropriate entries in the linker script. By default
4375 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4376 @code{reset} for vector 31 are recognized.
4379 @cindex @code{naked} function attribute, MSP430
4380 This attribute allows the compiler to construct the
4381 requisite function declaration, while allowing the body of the
4382 function to be assembly code. The specified function will not have
4383 prologue/epilogue sequences generated by the compiler. Only basic
4384 @code{asm} statements can safely be included in naked functions
4385 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4386 basic @code{asm} and C code may appear to work, they cannot be
4387 depended upon to work reliably and are not supported.
4390 @cindex @code{reentrant} function attribute, MSP430
4391 Reentrant functions disable interrupts upon entry and enable them
4392 upon exit. Reentrant functions cannot also have the @code{naked}
4393 or @code{critical} attributes. They can have the @code{interrupt}
4397 @cindex @code{wakeup} function attribute, MSP430
4398 This attribute only applies to interrupt functions. It is silently
4399 ignored if applied to a non-interrupt function. A wakeup interrupt
4400 function will rouse the processor from any low-power state that it
4401 might be in when the function exits.
4404 @node NDS32 Function Attributes
4405 @subsection NDS32 Function Attributes
4407 These function attributes are supported by the NDS32 back end:
4411 @cindex @code{exception} function attribute
4412 @cindex exception handler functions, NDS32
4413 Use this attribute on the NDS32 target to indicate that the specified function
4414 is an exception handler. The compiler will generate corresponding sections
4415 for use in an exception handler.
4418 @cindex @code{interrupt} function attribute, NDS32
4419 On NDS32 target, this attribute indicates that the specified function
4420 is an interrupt handler. The compiler generates corresponding sections
4421 for use in an interrupt handler. You can use the following attributes
4422 to modify the behavior:
4425 @cindex @code{nested} function attribute, NDS32
4426 This interrupt service routine is interruptible.
4428 @cindex @code{not_nested} function attribute, NDS32
4429 This interrupt service routine is not interruptible.
4431 @cindex @code{nested_ready} function attribute, NDS32
4432 This interrupt service routine is interruptible after @code{PSW.GIE}
4433 (global interrupt enable) is set. This allows interrupt service routine to
4434 finish some short critical code before enabling interrupts.
4436 @cindex @code{save_all} function attribute, NDS32
4437 The system will help save all registers into stack before entering
4440 @cindex @code{partial_save} function attribute, NDS32
4441 The system will help save caller registers into stack before entering
4446 @cindex @code{naked} function attribute, NDS32
4447 This attribute allows the compiler to construct the
4448 requisite function declaration, while allowing the body of the
4449 function to be assembly code. The specified function will not have
4450 prologue/epilogue sequences generated by the compiler. Only basic
4451 @code{asm} statements can safely be included in naked functions
4452 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4453 basic @code{asm} and C code may appear to work, they cannot be
4454 depended upon to work reliably and are not supported.
4457 @cindex @code{reset} function attribute, NDS32
4458 @cindex reset handler functions
4459 Use this attribute on the NDS32 target to indicate that the specified function
4460 is a reset handler. The compiler will generate corresponding sections
4461 for use in a reset handler. You can use the following attributes
4462 to provide extra exception handling:
4465 @cindex @code{nmi} function attribute, NDS32
4466 Provide a user-defined function to handle NMI exception.
4468 @cindex @code{warm} function attribute, NDS32
4469 Provide a user-defined function to handle warm reset exception.
4473 @node Nios II Function Attributes
4474 @subsection Nios II Function Attributes
4476 These function attributes are supported by the Nios II back end:
4479 @item target (@var{options})
4480 @cindex @code{target} function attribute
4481 As discussed in @ref{Common Function Attributes}, this attribute
4482 allows specification of target-specific compilation options.
4484 When compiling for Nios II, the following options are allowed:
4487 @item custom-@var{insn}=@var{N}
4488 @itemx no-custom-@var{insn}
4489 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4490 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4491 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4492 custom instruction with encoding @var{N} when generating code that uses
4493 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4494 the custom instruction @var{insn}.
4495 These target attributes correspond to the
4496 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4497 command-line options, and support the same set of @var{insn} keywords.
4498 @xref{Nios II Options}, for more information.
4500 @item custom-fpu-cfg=@var{name}
4501 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4502 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4503 command-line option, to select a predefined set of custom instructions
4505 @xref{Nios II Options}, for more information.
4509 @node PowerPC Function Attributes
4510 @subsection PowerPC Function Attributes
4512 These function attributes are supported by the PowerPC back end:
4517 @cindex indirect calls, PowerPC
4518 @cindex @code{longcall} function attribute, PowerPC
4519 @cindex @code{shortcall} function attribute, PowerPC
4520 The @code{longcall} attribute
4521 indicates that the function might be far away from the call site and
4522 require a different (more expensive) calling sequence. The
4523 @code{shortcall} attribute indicates that the function is always close
4524 enough for the shorter calling sequence to be used. These attributes
4525 override both the @option{-mlongcall} switch and
4526 the @code{#pragma longcall} setting.
4528 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4529 calls are necessary.
4531 @item target (@var{options})
4532 @cindex @code{target} function attribute
4533 As discussed in @ref{Common Function Attributes}, this attribute
4534 allows specification of target-specific compilation options.
4536 On the PowerPC, the following options are allowed:
4541 @cindex @code{target("altivec")} function attribute, PowerPC
4542 Generate code that uses (does not use) AltiVec instructions. In
4543 32-bit code, you cannot enable AltiVec instructions unless
4544 @option{-mabi=altivec} is used on the command line.
4548 @cindex @code{target("cmpb")} function attribute, PowerPC
4549 Generate code that uses (does not use) the compare bytes instruction
4550 implemented on the POWER6 processor and other processors that support
4551 the PowerPC V2.05 architecture.
4555 @cindex @code{target("dlmzb")} function attribute, PowerPC
4556 Generate code that uses (does not use) the string-search @samp{dlmzb}
4557 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4558 generated by default when targeting those processors.
4562 @cindex @code{target("fprnd")} function attribute, PowerPC
4563 Generate code that uses (does not use) the FP round to integer
4564 instructions implemented on the POWER5+ processor and other processors
4565 that support the PowerPC V2.03 architecture.
4569 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4570 Generate code that uses (does not use) the decimal floating-point
4571 instructions implemented on some POWER processors.
4575 @cindex @code{target("isel")} function attribute, PowerPC
4576 Generate code that uses (does not use) ISEL instruction.
4580 @cindex @code{target("mfcrf")} function attribute, PowerPC
4581 Generate code that uses (does not use) the move from condition
4582 register field instruction implemented on the POWER4 processor and
4583 other processors that support the PowerPC V2.01 architecture.
4587 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4588 Generate code that uses (does not use) the FP move to/from general
4589 purpose register instructions implemented on the POWER6X processor and
4590 other processors that support the extended PowerPC V2.05 architecture.
4594 @cindex @code{target("mulhw")} function attribute, PowerPC
4595 Generate code that uses (does not use) the half-word multiply and
4596 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4597 These instructions are generated by default when targeting those
4602 @cindex @code{target("multiple")} function attribute, PowerPC
4603 Generate code that uses (does not use) the load multiple word
4604 instructions and the store multiple word instructions.
4608 @cindex @code{target("update")} function attribute, PowerPC
4609 Generate code that uses (does not use) the load or store instructions
4610 that update the base register to the address of the calculated memory
4615 @cindex @code{target("popcntb")} function attribute, PowerPC
4616 Generate code that uses (does not use) the popcount and double-precision
4617 FP reciprocal estimate instruction implemented on the POWER5
4618 processor and other processors that support the PowerPC V2.02
4623 @cindex @code{target("popcntd")} function attribute, PowerPC
4624 Generate code that uses (does not use) the popcount instruction
4625 implemented on the POWER7 processor and other processors that support
4626 the PowerPC V2.06 architecture.
4628 @item powerpc-gfxopt
4629 @itemx no-powerpc-gfxopt
4630 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4631 Generate code that uses (does not use) the optional PowerPC
4632 architecture instructions in the Graphics group, including
4633 floating-point select.
4636 @itemx no-powerpc-gpopt
4637 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4638 Generate code that uses (does not use) the optional PowerPC
4639 architecture instructions in the General Purpose group, including
4640 floating-point square root.
4642 @item recip-precision
4643 @itemx no-recip-precision
4644 @cindex @code{target("recip-precision")} function attribute, PowerPC
4645 Assume (do not assume) that the reciprocal estimate instructions
4646 provide higher-precision estimates than is mandated by the PowerPC
4651 @cindex @code{target("string")} function attribute, PowerPC
4652 Generate code that uses (does not use) the load string instructions
4653 and the store string word instructions to save multiple registers and
4654 do small block moves.
4658 @cindex @code{target("vsx")} function attribute, PowerPC
4659 Generate code that uses (does not use) vector/scalar (VSX)
4660 instructions, and also enable the use of built-in functions that allow
4661 more direct access to the VSX instruction set. In 32-bit code, you
4662 cannot enable VSX or AltiVec instructions unless
4663 @option{-mabi=altivec} is used on the command line.
4667 @cindex @code{target("friz")} function attribute, PowerPC
4668 Generate (do not generate) the @code{friz} instruction when the
4669 @option{-funsafe-math-optimizations} option is used to optimize
4670 rounding a floating-point value to 64-bit integer and back to floating
4671 point. The @code{friz} instruction does not return the same value if
4672 the floating-point number is too large to fit in an integer.
4674 @item avoid-indexed-addresses
4675 @itemx no-avoid-indexed-addresses
4676 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4677 Generate code that tries to avoid (not avoid) the use of indexed load
4678 or store instructions.
4682 @cindex @code{target("paired")} function attribute, PowerPC
4683 Generate code that uses (does not use) the generation of PAIRED simd
4688 @cindex @code{target("longcall")} function attribute, PowerPC
4689 Generate code that assumes (does not assume) that all calls are far
4690 away so that a longer more expensive calling sequence is required.
4693 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4694 Specify the architecture to generate code for when compiling the
4695 function. If you select the @code{target("cpu=power7")} attribute when
4696 generating 32-bit code, VSX and AltiVec instructions are not generated
4697 unless you use the @option{-mabi=altivec} option on the command line.
4699 @item tune=@var{TUNE}
4700 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4701 Specify the architecture to tune for when compiling the function. If
4702 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4703 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4704 compilation tunes for the @var{CPU} architecture, and not the
4705 default tuning specified on the command line.
4708 On the PowerPC, the inliner does not inline a
4709 function that has different target options than the caller, unless the
4710 callee has a subset of the target options of the caller.
4713 @node RL78 Function Attributes
4714 @subsection RL78 Function Attributes
4716 These function attributes are supported by the RL78 back end:
4720 @itemx brk_interrupt
4721 @cindex @code{interrupt} function attribute, RL78
4722 @cindex @code{brk_interrupt} function attribute, RL78
4723 These attributes indicate
4724 that the specified function is an interrupt handler. The compiler generates
4725 function entry and exit sequences suitable for use in an interrupt handler
4726 when this attribute is present.
4728 Use @code{brk_interrupt} instead of @code{interrupt} for
4729 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4730 that must end with @code{RETB} instead of @code{RETI}).
4733 @cindex @code{naked} function attribute, RL78
4734 This attribute allows the compiler to construct the
4735 requisite function declaration, while allowing the body of the
4736 function to be assembly code. The specified function will not have
4737 prologue/epilogue sequences generated by the compiler. Only basic
4738 @code{asm} statements can safely be included in naked functions
4739 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4740 basic @code{asm} and C code may appear to work, they cannot be
4741 depended upon to work reliably and are not supported.
4744 @node RX Function Attributes
4745 @subsection RX Function Attributes
4747 These function attributes are supported by the RX back end:
4750 @item fast_interrupt
4751 @cindex @code{fast_interrupt} function attribute, RX
4752 Use this attribute on the RX port to indicate that the specified
4753 function is a fast interrupt handler. This is just like the
4754 @code{interrupt} attribute, except that @code{freit} is used to return
4755 instead of @code{reit}.
4758 @cindex @code{interrupt} function attribute, RX
4759 Use this attribute to indicate
4760 that the specified function is an interrupt handler. The compiler generates
4761 function entry and exit sequences suitable for use in an interrupt handler
4762 when this attribute is present.
4764 On RX targets, you may specify one or more vector numbers as arguments
4765 to the attribute, as well as naming an alternate table name.
4766 Parameters are handled sequentially, so one handler can be assigned to
4767 multiple entries in multiple tables. One may also pass the magic
4768 string @code{"$default"} which causes the function to be used for any
4769 unfilled slots in the current table.
4771 This example shows a simple assignment of a function to one vector in
4772 the default table (note that preprocessor macros may be used for
4773 chip-specific symbolic vector names):
4775 void __attribute__ ((interrupt (5))) txd1_handler ();
4778 This example assigns a function to two slots in the default table
4779 (using preprocessor macros defined elsewhere) and makes it the default
4780 for the @code{dct} table:
4782 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4787 @cindex @code{naked} function attribute, RX
4788 This attribute allows the compiler to construct the
4789 requisite function declaration, while allowing the body of the
4790 function to be assembly code. The specified function will not have
4791 prologue/epilogue sequences generated by the compiler. Only basic
4792 @code{asm} statements can safely be included in naked functions
4793 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4794 basic @code{asm} and C code may appear to work, they cannot be
4795 depended upon to work reliably and are not supported.
4798 @cindex @code{vector} function attribute, RX
4799 This RX attribute is similar to the @code{interrupt} attribute, including its
4800 parameters, but does not make the function an interrupt-handler type
4801 function (i.e. it retains the normal C function calling ABI). See the
4802 @code{interrupt} attribute for a description of its arguments.
4805 @node S/390 Function Attributes
4806 @subsection S/390 Function Attributes
4808 These function attributes are supported on the S/390:
4811 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4812 @cindex @code{hotpatch} function attribute, S/390
4814 On S/390 System z targets, you can use this function attribute to
4815 make GCC generate a ``hot-patching'' function prologue. If the
4816 @option{-mhotpatch=} command-line option is used at the same time,
4817 the @code{hotpatch} attribute takes precedence. The first of the
4818 two arguments specifies the number of halfwords to be added before
4819 the function label. A second argument can be used to specify the
4820 number of halfwords to be added after the function label. For
4821 both arguments the maximum allowed value is 1000000.
4823 If both arguments are zero, hotpatching is disabled.
4826 @node SH Function Attributes
4827 @subsection SH Function Attributes
4829 These function attributes are supported on the SH family of processors:
4832 @item function_vector
4833 @cindex @code{function_vector} function attribute, SH
4834 @cindex calling functions through the function vector on SH2A
4835 On SH2A targets, this attribute declares a function to be called using the
4836 TBR relative addressing mode. The argument to this attribute is the entry
4837 number of the same function in a vector table containing all the TBR
4838 relative addressable functions. For correct operation the TBR must be setup
4839 accordingly to point to the start of the vector table before any functions with
4840 this attribute are invoked. Usually a good place to do the initialization is
4841 the startup routine. The TBR relative vector table can have at max 256 function
4842 entries. The jumps to these functions are generated using a SH2A specific,
4843 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
4844 from GNU binutils version 2.7 or later for this attribute to work correctly.
4846 In an application, for a function being called once, this attribute
4847 saves at least 8 bytes of code; and if other successive calls are being
4848 made to the same function, it saves 2 bytes of code per each of these
4851 @item interrupt_handler
4852 @cindex @code{interrupt_handler} function attribute, SH
4853 Use this attribute to
4854 indicate that the specified function is an interrupt handler. The compiler
4855 generates function entry and exit sequences suitable for use in an
4856 interrupt handler when this attribute is present.
4858 @item nosave_low_regs
4859 @cindex @code{nosave_low_regs} function attribute, SH
4860 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
4861 function should not save and restore registers R0..R7. This can be used on SH3*
4862 and SH4* targets that have a second R0..R7 register bank for non-reentrant
4866 @cindex @code{renesas} function attribute, SH
4867 On SH targets this attribute specifies that the function or struct follows the
4871 @cindex @code{resbank} function attribute, SH
4872 On the SH2A target, this attribute enables the high-speed register
4873 saving and restoration using a register bank for @code{interrupt_handler}
4874 routines. Saving to the bank is performed automatically after the CPU
4875 accepts an interrupt that uses a register bank.
4877 The nineteen 32-bit registers comprising general register R0 to R14,
4878 control register GBR, and system registers MACH, MACL, and PR and the
4879 vector table address offset are saved into a register bank. Register
4880 banks are stacked in first-in last-out (FILO) sequence. Restoration
4881 from the bank is executed by issuing a RESBANK instruction.
4884 @cindex @code{sp_switch} function attribute, SH
4885 Use this attribute on the SH to indicate an @code{interrupt_handler}
4886 function should switch to an alternate stack. It expects a string
4887 argument that names a global variable holding the address of the
4892 void f () __attribute__ ((interrupt_handler,
4893 sp_switch ("alt_stack")));
4897 @cindex @code{trap_exit} function attribute, SH
4898 Use this attribute on the SH for an @code{interrupt_handler} to return using
4899 @code{trapa} instead of @code{rte}. This attribute expects an integer
4900 argument specifying the trap number to be used.
4903 @cindex @code{trapa_handler} function attribute, SH
4904 On SH targets this function attribute is similar to @code{interrupt_handler}
4905 but it does not save and restore all registers.
4908 @node SPU Function Attributes
4909 @subsection SPU Function Attributes
4911 These function attributes are supported by the SPU back end:
4915 @cindex @code{naked} function attribute, SPU
4916 This attribute allows the compiler to construct the
4917 requisite function declaration, while allowing the body of the
4918 function to be assembly code. The specified function will not have
4919 prologue/epilogue sequences generated by the compiler. Only basic
4920 @code{asm} statements can safely be included in naked functions
4921 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4922 basic @code{asm} and C code may appear to work, they cannot be
4923 depended upon to work reliably and are not supported.
4926 @node Symbian OS Function Attributes
4927 @subsection Symbian OS Function Attributes
4929 @xref{Microsoft Windows Function Attributes}, for discussion of the
4930 @code{dllexport} and @code{dllimport} attributes.
4932 @node Visium Function Attributes
4933 @subsection Visium Function Attributes
4935 These function attributes are supported by the Visium back end:
4939 @cindex @code{interrupt} function attribute, Visium
4940 Use this attribute to indicate
4941 that the specified function is an interrupt handler. The compiler generates
4942 function entry and exit sequences suitable for use in an interrupt handler
4943 when this attribute is present.
4946 @node x86 Function Attributes
4947 @subsection x86 Function Attributes
4949 These function attributes are supported by the x86 back end:
4953 @cindex @code{cdecl} function attribute, x86-32
4954 @cindex functions that pop the argument stack on x86-32
4956 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
4957 assume that the calling function pops off the stack space used to
4958 pass arguments. This is
4959 useful to override the effects of the @option{-mrtd} switch.
4962 @cindex @code{fastcall} function attribute, x86-32
4963 @cindex functions that pop the argument stack on x86-32
4964 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
4965 pass the first argument (if of integral type) in the register ECX and
4966 the second argument (if of integral type) in the register EDX@. Subsequent
4967 and other typed arguments are passed on the stack. The called function
4968 pops the arguments off the stack. If the number of arguments is variable all
4969 arguments are pushed on the stack.
4972 @cindex @code{thiscall} function attribute, x86-32
4973 @cindex functions that pop the argument stack on x86-32
4974 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
4975 pass the first argument (if of integral type) in the register ECX.
4976 Subsequent and other typed arguments are passed on the stack. The called
4977 function pops the arguments off the stack.
4978 If the number of arguments is variable all arguments are pushed on the
4980 The @code{thiscall} attribute is intended for C++ non-static member functions.
4981 As a GCC extension, this calling convention can be used for C functions
4982 and for static member methods.
4986 @cindex @code{ms_abi} function attribute, x86
4987 @cindex @code{sysv_abi} function attribute, x86
4989 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
4990 to indicate which calling convention should be used for a function. The
4991 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
4992 while the @code{sysv_abi} attribute tells the compiler to use the ABI
4993 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
4994 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
4996 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
4997 requires the @option{-maccumulate-outgoing-args} option.
4999 @item callee_pop_aggregate_return (@var{number})
5000 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5002 On x86-32 targets, you can use this attribute to control how
5003 aggregates are returned in memory. If the caller is responsible for
5004 popping the hidden pointer together with the rest of the arguments, specify
5005 @var{number} equal to zero. If callee is responsible for popping the
5006 hidden pointer, specify @var{number} equal to one.
5008 The default x86-32 ABI assumes that the callee pops the
5009 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5010 the compiler assumes that the
5011 caller pops the stack for hidden pointer.
5013 @item ms_hook_prologue
5014 @cindex @code{ms_hook_prologue} function attribute, x86
5016 On 32-bit and 64-bit x86 targets, you can use
5017 this function attribute to make GCC generate the ``hot-patching'' function
5018 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5021 @item regparm (@var{number})
5022 @cindex @code{regparm} function attribute, x86
5023 @cindex functions that are passed arguments in registers on x86-32
5024 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5025 pass arguments number one to @var{number} if they are of integral type
5026 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5027 take a variable number of arguments continue to be passed all of their
5028 arguments on the stack.
5030 Beware that on some ELF systems this attribute is unsuitable for
5031 global functions in shared libraries with lazy binding (which is the
5032 default). Lazy binding sends the first call via resolving code in
5033 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5034 per the standard calling conventions. Solaris 8 is affected by this.
5035 Systems with the GNU C Library version 2.1 or higher
5036 and FreeBSD are believed to be
5037 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5038 disabled with the linker or the loader if desired, to avoid the
5042 @cindex @code{sseregparm} function attribute, x86
5043 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5044 causes the compiler to pass up to 3 floating-point arguments in
5045 SSE registers instead of on the stack. Functions that take a
5046 variable number of arguments continue to pass all of their
5047 floating-point arguments on the stack.
5049 @item force_align_arg_pointer
5050 @cindex @code{force_align_arg_pointer} function attribute, x86
5051 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5052 applied to individual function definitions, generating an alternate
5053 prologue and epilogue that realigns the run-time stack if necessary.
5054 This supports mixing legacy codes that run with a 4-byte aligned stack
5055 with modern codes that keep a 16-byte stack for SSE compatibility.
5058 @cindex @code{stdcall} function attribute, x86-32
5059 @cindex functions that pop the argument stack on x86-32
5060 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5061 assume that the called function pops off the stack space used to
5062 pass arguments, unless it takes a variable number of arguments.
5064 @item target (@var{options})
5065 @cindex @code{target} function attribute
5066 As discussed in @ref{Common Function Attributes}, this attribute
5067 allows specification of target-specific compilation options.
5069 On the x86, the following options are allowed:
5073 @cindex @code{target("abm")} function attribute, x86
5074 Enable/disable the generation of the advanced bit instructions.
5078 @cindex @code{target("aes")} function attribute, x86
5079 Enable/disable the generation of the AES instructions.
5082 @cindex @code{target("default")} function attribute, x86
5083 @xref{Function Multiversioning}, where it is used to specify the
5084 default function version.
5088 @cindex @code{target("mmx")} function attribute, x86
5089 Enable/disable the generation of the MMX instructions.
5093 @cindex @code{target("pclmul")} function attribute, x86
5094 Enable/disable the generation of the PCLMUL instructions.
5098 @cindex @code{target("popcnt")} function attribute, x86
5099 Enable/disable the generation of the POPCNT instruction.
5103 @cindex @code{target("sse")} function attribute, x86
5104 Enable/disable the generation of the SSE instructions.
5108 @cindex @code{target("sse2")} function attribute, x86
5109 Enable/disable the generation of the SSE2 instructions.
5113 @cindex @code{target("sse3")} function attribute, x86
5114 Enable/disable the generation of the SSE3 instructions.
5118 @cindex @code{target("sse4")} function attribute, x86
5119 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5124 @cindex @code{target("sse4.1")} function attribute, x86
5125 Enable/disable the generation of the sse4.1 instructions.
5129 @cindex @code{target("sse4.2")} function attribute, x86
5130 Enable/disable the generation of the sse4.2 instructions.
5134 @cindex @code{target("sse4a")} function attribute, x86
5135 Enable/disable the generation of the SSE4A instructions.
5139 @cindex @code{target("fma4")} function attribute, x86
5140 Enable/disable the generation of the FMA4 instructions.
5144 @cindex @code{target("xop")} function attribute, x86
5145 Enable/disable the generation of the XOP instructions.
5149 @cindex @code{target("lwp")} function attribute, x86
5150 Enable/disable the generation of the LWP instructions.
5154 @cindex @code{target("ssse3")} function attribute, x86
5155 Enable/disable the generation of the SSSE3 instructions.
5159 @cindex @code{target("cld")} function attribute, x86
5160 Enable/disable the generation of the CLD before string moves.
5162 @item fancy-math-387
5163 @itemx no-fancy-math-387
5164 @cindex @code{target("fancy-math-387")} function attribute, x86
5165 Enable/disable the generation of the @code{sin}, @code{cos}, and
5166 @code{sqrt} instructions on the 387 floating-point unit.
5169 @itemx no-fused-madd
5170 @cindex @code{target("fused-madd")} function attribute, x86
5171 Enable/disable the generation of the fused multiply/add instructions.
5175 @cindex @code{target("ieee-fp")} function attribute, x86
5176 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5178 @item inline-all-stringops
5179 @itemx no-inline-all-stringops
5180 @cindex @code{target("inline-all-stringops")} function attribute, x86
5181 Enable/disable inlining of string operations.
5183 @item inline-stringops-dynamically
5184 @itemx no-inline-stringops-dynamically
5185 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5186 Enable/disable the generation of the inline code to do small string
5187 operations and calling the library routines for large operations.
5189 @item align-stringops
5190 @itemx no-align-stringops
5191 @cindex @code{target("align-stringops")} function attribute, x86
5192 Do/do not align destination of inlined string operations.
5196 @cindex @code{target("recip")} function attribute, x86
5197 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5198 instructions followed an additional Newton-Raphson step instead of
5199 doing a floating-point division.
5201 @item arch=@var{ARCH}
5202 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5203 Specify the architecture to generate code for in compiling the function.
5205 @item tune=@var{TUNE}
5206 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5207 Specify the architecture to tune for in compiling the function.
5209 @item fpmath=@var{FPMATH}
5210 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5211 Specify which floating-point unit to use. You must specify the
5212 @code{target("fpmath=sse,387")} option as
5213 @code{target("fpmath=sse+387")} because the comma would separate
5217 On the x86, the inliner does not inline a
5218 function that has different target options than the caller, unless the
5219 callee has a subset of the target options of the caller. For example
5220 a function declared with @code{target("sse3")} can inline a function
5221 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5224 @node Xstormy16 Function Attributes
5225 @subsection Xstormy16 Function Attributes
5227 These function attributes are supported by the Xstormy16 back end:
5231 @cindex @code{interrupt} function attribute, Xstormy16
5232 Use this attribute to indicate
5233 that the specified function is an interrupt handler. The compiler generates
5234 function entry and exit sequences suitable for use in an interrupt handler
5235 when this attribute is present.
5238 @node Variable Attributes
5239 @section Specifying Attributes of Variables
5240 @cindex attribute of variables
5241 @cindex variable attributes
5243 The keyword @code{__attribute__} allows you to specify special
5244 attributes of variables or structure fields. This keyword is followed
5245 by an attribute specification inside double parentheses. Some
5246 attributes are currently defined generically for variables.
5247 Other attributes are defined for variables on particular target
5248 systems. Other attributes are available for functions
5249 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5250 enumerators (@pxref{Enumerator Attributes}), and for types
5251 (@pxref{Type Attributes}).
5252 Other front ends might define more attributes
5253 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5255 @xref{Attribute Syntax}, for details of the exact syntax for using
5259 * Common Variable Attributes::
5260 * AVR Variable Attributes::
5261 * Blackfin Variable Attributes::
5262 * H8/300 Variable Attributes::
5263 * IA-64 Variable Attributes::
5264 * M32R/D Variable Attributes::
5265 * MeP Variable Attributes::
5266 * Microsoft Windows Variable Attributes::
5267 * PowerPC Variable Attributes::
5268 * SPU Variable Attributes::
5269 * x86 Variable Attributes::
5270 * Xstormy16 Variable Attributes::
5273 @node Common Variable Attributes
5274 @subsection Common Variable Attributes
5276 The following attributes are supported on most targets.
5279 @cindex @code{aligned} variable attribute
5280 @item aligned (@var{alignment})
5281 This attribute specifies a minimum alignment for the variable or
5282 structure field, measured in bytes. For example, the declaration:
5285 int x __attribute__ ((aligned (16))) = 0;
5289 causes the compiler to allocate the global variable @code{x} on a
5290 16-byte boundary. On a 68040, this could be used in conjunction with
5291 an @code{asm} expression to access the @code{move16} instruction which
5292 requires 16-byte aligned operands.
5294 You can also specify the alignment of structure fields. For example, to
5295 create a double-word aligned @code{int} pair, you could write:
5298 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5302 This is an alternative to creating a union with a @code{double} member,
5303 which forces the union to be double-word aligned.
5305 As in the preceding examples, you can explicitly specify the alignment
5306 (in bytes) that you wish the compiler to use for a given variable or
5307 structure field. Alternatively, you can leave out the alignment factor
5308 and just ask the compiler to align a variable or field to the
5309 default alignment for the target architecture you are compiling for.
5310 The default alignment is sufficient for all scalar types, but may not be
5311 enough for all vector types on a target that supports vector operations.
5312 The default alignment is fixed for a particular target ABI.
5314 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5315 which is the largest alignment ever used for any data type on the
5316 target machine you are compiling for. For example, you could write:
5319 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5322 The compiler automatically sets the alignment for the declared
5323 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5324 often make copy operations more efficient, because the compiler can
5325 use whatever instructions copy the biggest chunks of memory when
5326 performing copies to or from the variables or fields that you have
5327 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5328 may change depending on command-line options.
5330 When used on a struct, or struct member, the @code{aligned} attribute can
5331 only increase the alignment; in order to decrease it, the @code{packed}
5332 attribute must be specified as well. When used as part of a typedef, the
5333 @code{aligned} attribute can both increase and decrease alignment, and
5334 specifying the @code{packed} attribute generates a warning.
5336 Note that the effectiveness of @code{aligned} attributes may be limited
5337 by inherent limitations in your linker. On many systems, the linker is
5338 only able to arrange for variables to be aligned up to a certain maximum
5339 alignment. (For some linkers, the maximum supported alignment may
5340 be very very small.) If your linker is only able to align variables
5341 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5342 in an @code{__attribute__} still only provides you with 8-byte
5343 alignment. See your linker documentation for further information.
5345 The @code{aligned} attribute can also be used for functions
5346 (@pxref{Common Function Attributes}.)
5348 @item cleanup (@var{cleanup_function})
5349 @cindex @code{cleanup} variable attribute
5350 The @code{cleanup} attribute runs a function when the variable goes
5351 out of scope. This attribute can only be applied to auto function
5352 scope variables; it may not be applied to parameters or variables
5353 with static storage duration. The function must take one parameter,
5354 a pointer to a type compatible with the variable. The return value
5355 of the function (if any) is ignored.
5357 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5358 is run during the stack unwinding that happens during the
5359 processing of the exception. Note that the @code{cleanup} attribute
5360 does not allow the exception to be caught, only to perform an action.
5361 It is undefined what happens if @var{cleanup_function} does not
5366 @cindex @code{common} variable attribute
5367 @cindex @code{nocommon} variable attribute
5370 The @code{common} attribute requests GCC to place a variable in
5371 ``common'' storage. The @code{nocommon} attribute requests the
5372 opposite---to allocate space for it directly.
5374 These attributes override the default chosen by the
5375 @option{-fno-common} and @option{-fcommon} flags respectively.
5378 @itemx deprecated (@var{msg})
5379 @cindex @code{deprecated} variable attribute
5380 The @code{deprecated} attribute results in a warning if the variable
5381 is used anywhere in the source file. This is useful when identifying
5382 variables that are expected to be removed in a future version of a
5383 program. The warning also includes the location of the declaration
5384 of the deprecated variable, to enable users to easily find further
5385 information about why the variable is deprecated, or what they should
5386 do instead. Note that the warning only occurs for uses:
5389 extern int old_var __attribute__ ((deprecated));
5391 int new_fn () @{ return old_var; @}
5395 results in a warning on line 3 but not line 2. The optional @var{msg}
5396 argument, which must be a string, is printed in the warning if
5399 The @code{deprecated} attribute can also be used for functions and
5400 types (@pxref{Common Function Attributes},
5401 @pxref{Common Type Attributes}).
5403 @item mode (@var{mode})
5404 @cindex @code{mode} variable attribute
5405 This attribute specifies the data type for the declaration---whichever
5406 type corresponds to the mode @var{mode}. This in effect lets you
5407 request an integer or floating-point type according to its width.
5409 You may also specify a mode of @code{byte} or @code{__byte__} to
5410 indicate the mode corresponding to a one-byte integer, @code{word} or
5411 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5412 or @code{__pointer__} for the mode used to represent pointers.
5415 @cindex @code{packed} variable attribute
5416 The @code{packed} attribute specifies that a variable or structure field
5417 should have the smallest possible alignment---one byte for a variable,
5418 and one bit for a field, unless you specify a larger value with the
5419 @code{aligned} attribute.
5421 Here is a structure in which the field @code{x} is packed, so that it
5422 immediately follows @code{a}:
5428 int x[2] __attribute__ ((packed));
5432 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5433 @code{packed} attribute on bit-fields of type @code{char}. This has
5434 been fixed in GCC 4.4 but the change can lead to differences in the
5435 structure layout. See the documentation of
5436 @option{-Wpacked-bitfield-compat} for more information.
5438 @item section ("@var{section-name}")
5439 @cindex @code{section} variable attribute
5440 Normally, the compiler places the objects it generates in sections like
5441 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5442 or you need certain particular variables to appear in special sections,
5443 for example to map to special hardware. The @code{section}
5444 attribute specifies that a variable (or function) lives in a particular
5445 section. For example, this small program uses several specific section names:
5448 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5449 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5450 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5451 int init_data __attribute__ ((section ("INITDATA")));
5455 /* @r{Initialize stack pointer} */
5456 init_sp (stack + sizeof (stack));
5458 /* @r{Initialize initialized data} */
5459 memcpy (&init_data, &data, &edata - &data);
5461 /* @r{Turn on the serial ports} */
5468 Use the @code{section} attribute with
5469 @emph{global} variables and not @emph{local} variables,
5470 as shown in the example.
5472 You may use the @code{section} attribute with initialized or
5473 uninitialized global variables but the linker requires
5474 each object be defined once, with the exception that uninitialized
5475 variables tentatively go in the @code{common} (or @code{bss}) section
5476 and can be multiply ``defined''. Using the @code{section} attribute
5477 changes what section the variable goes into and may cause the
5478 linker to issue an error if an uninitialized variable has multiple
5479 definitions. You can force a variable to be initialized with the
5480 @option{-fno-common} flag or the @code{nocommon} attribute.
5482 Some file formats do not support arbitrary sections so the @code{section}
5483 attribute is not available on all platforms.
5484 If you need to map the entire contents of a module to a particular
5485 section, consider using the facilities of the linker instead.
5487 @item tls_model ("@var{tls_model}")
5488 @cindex @code{tls_model} variable attribute
5489 The @code{tls_model} attribute sets thread-local storage model
5490 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5491 overriding @option{-ftls-model=} command-line switch on a per-variable
5493 The @var{tls_model} argument should be one of @code{global-dynamic},
5494 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5496 Not all targets support this attribute.
5499 @cindex @code{unused} variable attribute
5500 This attribute, attached to a variable, means that the variable is meant
5501 to be possibly unused. GCC does not produce a warning for this
5505 @cindex @code{used} variable attribute
5506 This attribute, attached to a variable with static storage, means that
5507 the variable must be emitted even if it appears that the variable is not
5510 When applied to a static data member of a C++ class template, the
5511 attribute also means that the member is instantiated if the
5512 class itself is instantiated.
5514 @item vector_size (@var{bytes})
5515 @cindex @code{vector_size} variable attribute
5516 This attribute specifies the vector size for the variable, measured in
5517 bytes. For example, the declaration:
5520 int foo __attribute__ ((vector_size (16)));
5524 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5525 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5526 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5528 This attribute is only applicable to integral and float scalars,
5529 although arrays, pointers, and function return values are allowed in
5530 conjunction with this construct.
5532 Aggregates with this attribute are invalid, even if they are of the same
5533 size as a corresponding scalar. For example, the declaration:
5536 struct S @{ int a; @};
5537 struct S __attribute__ ((vector_size (16))) foo;
5541 is invalid even if the size of the structure is the same as the size of
5545 @cindex @code{weak} variable attribute
5546 The @code{weak} attribute is described in
5547 @ref{Common Function Attributes}.
5551 @node AVR Variable Attributes
5552 @subsection AVR Variable Attributes
5556 @cindex @code{progmem} variable attribute, AVR
5557 The @code{progmem} attribute is used on the AVR to place read-only
5558 data in the non-volatile program memory (flash). The @code{progmem}
5559 attribute accomplishes this by putting respective variables into a
5560 section whose name starts with @code{.progmem}.
5562 This attribute works similar to the @code{section} attribute
5563 but adds additional checking. Notice that just like the
5564 @code{section} attribute, @code{progmem} affects the location
5565 of the data but not how this data is accessed.
5567 In order to read data located with the @code{progmem} attribute
5568 (inline) assembler must be used.
5570 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5571 #include <avr/pgmspace.h>
5573 /* Locate var in flash memory */
5574 const int var[2] PROGMEM = @{ 1, 2 @};
5576 int read_var (int i)
5578 /* Access var[] by accessor macro from avr/pgmspace.h */
5579 return (int) pgm_read_word (& var[i]);
5583 AVR is a Harvard architecture processor and data and read-only data
5584 normally resides in the data memory (RAM).
5586 See also the @ref{AVR Named Address Spaces} section for
5587 an alternate way to locate and access data in flash memory.
5590 @itemx io (@var{addr})
5591 @cindex @code{io} variable attribute, AVR
5592 Variables with the @code{io} attribute are used to address
5593 memory-mapped peripherals in the io address range.
5594 If an address is specified, the variable
5595 is assigned that address, and the value is interpreted as an
5596 address in the data address space.
5600 volatile int porta __attribute__((io (0x22)));
5603 The address specified in the address in the data address range.
5605 Otherwise, the variable it is not assigned an address, but the
5606 compiler will still use in/out instructions where applicable,
5607 assuming some other module assigns an address in the io address range.
5611 extern volatile int porta __attribute__((io));
5615 @itemx io_low (@var{addr})
5616 @cindex @code{io_low} variable attribute, AVR
5617 This is like the @code{io} attribute, but additionally it informs the
5618 compiler that the object lies in the lower half of the I/O area,
5619 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5623 @itemx address (@var{addr})
5624 @cindex @code{address} variable attribute, AVR
5625 Variables with the @code{address} attribute are used to address
5626 memory-mapped peripherals that may lie outside the io address range.
5629 volatile int porta __attribute__((address (0x600)));
5634 @node Blackfin Variable Attributes
5635 @subsection Blackfin Variable Attributes
5637 Three attributes are currently defined for the Blackfin.
5643 @cindex @code{l1_data} variable attribute, Blackfin
5644 @cindex @code{l1_data_A} variable attribute, Blackfin
5645 @cindex @code{l1_data_B} variable attribute, Blackfin
5646 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5647 Variables with @code{l1_data} attribute are put into the specific section
5648 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5649 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5650 attribute are put into the specific section named @code{.l1.data.B}.
5653 @cindex @code{l2} variable attribute, Blackfin
5654 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5655 Variables with @code{l2} attribute are put into the specific section
5656 named @code{.l2.data}.
5659 @node H8/300 Variable Attributes
5660 @subsection H8/300 Variable Attributes
5662 These variable attributes are available for H8/300 targets:
5666 @cindex @code{eightbit_data} variable attribute, H8/300
5667 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5668 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5669 variable should be placed into the eight-bit data section.
5670 The compiler generates more efficient code for certain operations
5671 on data in the eight-bit data area. Note the eight-bit data area is limited to
5674 You must use GAS and GLD from GNU binutils version 2.7 or later for
5675 this attribute to work correctly.
5678 @cindex @code{tiny_data} variable attribute, H8/300
5679 @cindex tiny data section on the H8/300H and H8S
5680 Use this attribute on the H8/300H and H8S to indicate that the specified
5681 variable should be placed into the tiny data section.
5682 The compiler generates more efficient code for loads and stores
5683 on data in the tiny data section. Note the tiny data area is limited to
5684 slightly under 32KB of data.
5688 @node IA-64 Variable Attributes
5689 @subsection IA-64 Variable Attributes
5691 The IA-64 back end supports the following variable attribute:
5694 @item model (@var{model-name})
5695 @cindex @code{model} variable attribute, IA-64
5697 On IA-64, use this attribute to set the addressability of an object.
5698 At present, the only supported identifier for @var{model-name} is
5699 @code{small}, indicating addressability via ``small'' (22-bit)
5700 addresses (so that their addresses can be loaded with the @code{addl}
5701 instruction). Caveat: such addressing is by definition not position
5702 independent and hence this attribute must not be used for objects
5703 defined by shared libraries.
5707 @node M32R/D Variable Attributes
5708 @subsection M32R/D Variable Attributes
5710 One attribute is currently defined for the M32R/D@.
5713 @item model (@var{model-name})
5714 @cindex @code{model-name} variable attribute, M32R/D
5715 @cindex variable addressability on the M32R/D
5716 Use this attribute on the M32R/D to set the addressability of an object.
5717 The identifier @var{model-name} is one of @code{small}, @code{medium},
5718 or @code{large}, representing each of the code models.
5720 Small model objects live in the lower 16MB of memory (so that their
5721 addresses can be loaded with the @code{ld24} instruction).
5723 Medium and large model objects may live anywhere in the 32-bit address space
5724 (the compiler generates @code{seth/add3} instructions to load their
5728 @node MeP Variable Attributes
5729 @subsection MeP Variable Attributes
5731 The MeP target has a number of addressing modes and busses. The
5732 @code{near} space spans the standard memory space's first 16 megabytes
5733 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5734 The @code{based} space is a 128-byte region in the memory space that
5735 is addressed relative to the @code{$tp} register. The @code{tiny}
5736 space is a 65536-byte region relative to the @code{$gp} register. In
5737 addition to these memory regions, the MeP target has a separate 16-bit
5738 control bus which is specified with @code{cb} attributes.
5743 @cindex @code{based} variable attribute, MeP
5744 Any variable with the @code{based} attribute is assigned to the
5745 @code{.based} section, and is accessed with relative to the
5746 @code{$tp} register.
5749 @cindex @code{tiny} variable attribute, MeP
5750 Likewise, the @code{tiny} attribute assigned variables to the
5751 @code{.tiny} section, relative to the @code{$gp} register.
5754 @cindex @code{near} variable attribute, MeP
5755 Variables with the @code{near} attribute are assumed to have addresses
5756 that fit in a 24-bit addressing mode. This is the default for large
5757 variables (@code{-mtiny=4} is the default) but this attribute can
5758 override @code{-mtiny=} for small variables, or override @code{-ml}.
5761 @cindex @code{far} variable attribute, MeP
5762 Variables with the @code{far} attribute are addressed using a full
5763 32-bit address. Since this covers the entire memory space, this
5764 allows modules to make no assumptions about where variables might be
5768 @cindex @code{io} variable attribute, MeP
5769 @itemx io (@var{addr})
5770 Variables with the @code{io} attribute are used to address
5771 memory-mapped peripherals. If an address is specified, the variable
5772 is assigned that address, else it is not assigned an address (it is
5773 assumed some other module assigns an address). Example:
5776 int timer_count __attribute__((io(0x123)));
5780 @itemx cb (@var{addr})
5781 @cindex @code{cb} variable attribute, MeP
5782 Variables with the @code{cb} attribute are used to access the control
5783 bus, using special instructions. @code{addr} indicates the control bus
5787 int cpu_clock __attribute__((cb(0x123)));
5792 @node Microsoft Windows Variable Attributes
5793 @subsection Microsoft Windows Variable Attributes
5795 You can use these attributes on Microsoft Windows targets.
5796 @ref{x86 Variable Attributes} for additional Windows compatibility
5797 attributes available on all x86 targets.
5802 @cindex @code{dllimport} variable attribute
5803 @cindex @code{dllexport} variable attribute
5804 The @code{dllimport} and @code{dllexport} attributes are described in
5805 @ref{Microsoft Windows Function Attributes}.
5808 @cindex @code{selectany} variable attribute
5809 The @code{selectany} attribute causes an initialized global variable to
5810 have link-once semantics. When multiple definitions of the variable are
5811 encountered by the linker, the first is selected and the remainder are
5812 discarded. Following usage by the Microsoft compiler, the linker is told
5813 @emph{not} to warn about size or content differences of the multiple
5816 Although the primary usage of this attribute is for POD types, the
5817 attribute can also be applied to global C++ objects that are initialized
5818 by a constructor. In this case, the static initialization and destruction
5819 code for the object is emitted in each translation defining the object,
5820 but the calls to the constructor and destructor are protected by a
5821 link-once guard variable.
5823 The @code{selectany} attribute is only available on Microsoft Windows
5824 targets. You can use @code{__declspec (selectany)} as a synonym for
5825 @code{__attribute__ ((selectany))} for compatibility with other
5829 @cindex @code{shared} variable attribute
5830 On Microsoft Windows, in addition to putting variable definitions in a named
5831 section, the section can also be shared among all running copies of an
5832 executable or DLL@. For example, this small program defines shared data
5833 by putting it in a named section @code{shared} and marking the section
5837 int foo __attribute__((section ("shared"), shared)) = 0;
5842 /* @r{Read and write foo. All running
5843 copies see the same value.} */
5849 You may only use the @code{shared} attribute along with @code{section}
5850 attribute with a fully-initialized global definition because of the way
5851 linkers work. See @code{section} attribute for more information.
5853 The @code{shared} attribute is only available on Microsoft Windows@.
5857 @node PowerPC Variable Attributes
5858 @subsection PowerPC Variable Attributes
5860 Three attributes currently are defined for PowerPC configurations:
5861 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5863 @cindex @code{ms_struct} variable attribute, PowerPC
5864 @cindex @code{gcc_struct} variable attribute, PowerPC
5865 For full documentation of the struct attributes please see the
5866 documentation in @ref{x86 Variable Attributes}.
5868 @cindex @code{altivec} variable attribute, PowerPC
5869 For documentation of @code{altivec} attribute please see the
5870 documentation in @ref{PowerPC Type Attributes}.
5872 @node SPU Variable Attributes
5873 @subsection SPU Variable Attributes
5875 @cindex @code{spu_vector} variable attribute, SPU
5876 The SPU supports the @code{spu_vector} attribute for variables. For
5877 documentation of this attribute please see the documentation in
5878 @ref{SPU Type Attributes}.
5880 @node x86 Variable Attributes
5881 @subsection x86 Variable Attributes
5883 Two attributes are currently defined for x86 configurations:
5884 @code{ms_struct} and @code{gcc_struct}.
5889 @cindex @code{ms_struct} variable attribute, x86
5890 @cindex @code{gcc_struct} variable attribute, x86
5892 If @code{packed} is used on a structure, or if bit-fields are used,
5893 it may be that the Microsoft ABI lays out the structure differently
5894 than the way GCC normally does. Particularly when moving packed
5895 data between functions compiled with GCC and the native Microsoft compiler
5896 (either via function call or as data in a file), it may be necessary to access
5899 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
5900 compilers to match the native Microsoft compiler.
5902 The Microsoft structure layout algorithm is fairly simple with the exception
5903 of the bit-field packing.
5904 The padding and alignment of members of structures and whether a bit-field
5905 can straddle a storage-unit boundary are determine by these rules:
5908 @item Structure members are stored sequentially in the order in which they are
5909 declared: the first member has the lowest memory address and the last member
5912 @item Every data object has an alignment requirement. The alignment requirement
5913 for all data except structures, unions, and arrays is either the size of the
5914 object or the current packing size (specified with either the
5915 @code{aligned} attribute or the @code{pack} pragma),
5916 whichever is less. For structures, unions, and arrays,
5917 the alignment requirement is the largest alignment requirement of its members.
5918 Every object is allocated an offset so that:
5921 offset % alignment_requirement == 0
5924 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
5925 unit if the integral types are the same size and if the next bit-field fits
5926 into the current allocation unit without crossing the boundary imposed by the
5927 common alignment requirements of the bit-fields.
5930 MSVC interprets zero-length bit-fields in the following ways:
5933 @item If a zero-length bit-field is inserted between two bit-fields that
5934 are normally coalesced, the bit-fields are not coalesced.
5941 unsigned long bf_1 : 12;
5943 unsigned long bf_2 : 12;
5948 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
5949 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
5951 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
5952 alignment of the zero-length bit-field is greater than the member that follows it,
5953 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
5974 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
5975 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
5976 bit-field does not affect the alignment of @code{bar} or, as a result, the size
5979 Taking this into account, it is important to note the following:
5982 @item If a zero-length bit-field follows a normal bit-field, the type of the
5983 zero-length bit-field may affect the alignment of the structure as whole. For
5984 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
5985 normal bit-field, and is of type short.
5987 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
5988 still affect the alignment of the structure:
5999 Here, @code{t4} takes up 4 bytes.
6002 @item Zero-length bit-fields following non-bit-field members are ignored:
6014 Here, @code{t5} takes up 2 bytes.
6018 @node Xstormy16 Variable Attributes
6019 @subsection Xstormy16 Variable Attributes
6021 One attribute is currently defined for xstormy16 configurations:
6026 @cindex @code{below100} variable attribute, Xstormy16
6028 If a variable has the @code{below100} attribute (@code{BELOW100} is
6029 allowed also), GCC places the variable in the first 0x100 bytes of
6030 memory and use special opcodes to access it. Such variables are
6031 placed in either the @code{.bss_below100} section or the
6032 @code{.data_below100} section.
6036 @node Type Attributes
6037 @section Specifying Attributes of Types
6038 @cindex attribute of types
6039 @cindex type attributes
6041 The keyword @code{__attribute__} allows you to specify special
6042 attributes of types. Some type attributes apply only to @code{struct}
6043 and @code{union} types, while others can apply to any type defined
6044 via a @code{typedef} declaration. Other attributes are defined for
6045 functions (@pxref{Function Attributes}), labels (@pxref{Label
6046 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6047 variables (@pxref{Variable Attributes}).
6049 The @code{__attribute__} keyword is followed by an attribute specification
6050 inside double parentheses.
6052 You may specify type attributes in an enum, struct or union type
6053 declaration or definition by placing them immediately after the
6054 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6055 syntax is to place them just past the closing curly brace of the
6058 You can also include type attributes in a @code{typedef} declaration.
6059 @xref{Attribute Syntax}, for details of the exact syntax for using
6063 * Common Type Attributes::
6064 * ARM Type Attributes::
6065 * MeP Type Attributes::
6066 * PowerPC Type Attributes::
6067 * SPU Type Attributes::
6068 * x86 Type Attributes::
6071 @node Common Type Attributes
6072 @subsection Common Type Attributes
6074 The following type attributes are supported on most targets.
6077 @cindex @code{aligned} type attribute
6078 @item aligned (@var{alignment})
6079 This attribute specifies a minimum alignment (in bytes) for variables
6080 of the specified type. For example, the declarations:
6083 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6084 typedef int more_aligned_int __attribute__ ((aligned (8)));
6088 force the compiler to ensure (as far as it can) that each variable whose
6089 type is @code{struct S} or @code{more_aligned_int} is allocated and
6090 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6091 variables of type @code{struct S} aligned to 8-byte boundaries allows
6092 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6093 store) instructions when copying one variable of type @code{struct S} to
6094 another, thus improving run-time efficiency.
6096 Note that the alignment of any given @code{struct} or @code{union} type
6097 is required by the ISO C standard to be at least a perfect multiple of
6098 the lowest common multiple of the alignments of all of the members of
6099 the @code{struct} or @code{union} in question. This means that you @emph{can}
6100 effectively adjust the alignment of a @code{struct} or @code{union}
6101 type by attaching an @code{aligned} attribute to any one of the members
6102 of such a type, but the notation illustrated in the example above is a
6103 more obvious, intuitive, and readable way to request the compiler to
6104 adjust the alignment of an entire @code{struct} or @code{union} type.
6106 As in the preceding example, you can explicitly specify the alignment
6107 (in bytes) that you wish the compiler to use for a given @code{struct}
6108 or @code{union} type. Alternatively, you can leave out the alignment factor
6109 and just ask the compiler to align a type to the maximum
6110 useful alignment for the target machine you are compiling for. For
6111 example, you could write:
6114 struct S @{ short f[3]; @} __attribute__ ((aligned));
6117 Whenever you leave out the alignment factor in an @code{aligned}
6118 attribute specification, the compiler automatically sets the alignment
6119 for the type to the largest alignment that is ever used for any data
6120 type on the target machine you are compiling for. Doing this can often
6121 make copy operations more efficient, because the compiler can use
6122 whatever instructions copy the biggest chunks of memory when performing
6123 copies to or from the variables that have types that you have aligned
6126 In the example above, if the size of each @code{short} is 2 bytes, then
6127 the size of the entire @code{struct S} type is 6 bytes. The smallest
6128 power of two that is greater than or equal to that is 8, so the
6129 compiler sets the alignment for the entire @code{struct S} type to 8
6132 Note that although you can ask the compiler to select a time-efficient
6133 alignment for a given type and then declare only individual stand-alone
6134 objects of that type, the compiler's ability to select a time-efficient
6135 alignment is primarily useful only when you plan to create arrays of
6136 variables having the relevant (efficiently aligned) type. If you
6137 declare or use arrays of variables of an efficiently-aligned type, then
6138 it is likely that your program also does pointer arithmetic (or
6139 subscripting, which amounts to the same thing) on pointers to the
6140 relevant type, and the code that the compiler generates for these
6141 pointer arithmetic operations is often more efficient for
6142 efficiently-aligned types than for other types.
6144 The @code{aligned} attribute can only increase the alignment; but you
6145 can decrease it by specifying @code{packed} as well. See below.
6147 Note that the effectiveness of @code{aligned} attributes may be limited
6148 by inherent limitations in your linker. On many systems, the linker is
6149 only able to arrange for variables to be aligned up to a certain maximum
6150 alignment. (For some linkers, the maximum supported alignment may
6151 be very very small.) If your linker is only able to align variables
6152 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6153 in an @code{__attribute__} still only provides you with 8-byte
6154 alignment. See your linker documentation for further information.
6156 @opindex fshort-enums
6157 Specifying this attribute for @code{struct} and @code{union} types is
6158 equivalent to specifying the @code{packed} attribute on each of the
6159 structure or union members. Specifying the @option{-fshort-enums}
6160 flag on the line is equivalent to specifying the @code{packed}
6161 attribute on all @code{enum} definitions.
6163 In the following example @code{struct my_packed_struct}'s members are
6164 packed closely together, but the internal layout of its @code{s} member
6165 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6169 struct my_unpacked_struct
6175 struct __attribute__ ((__packed__)) my_packed_struct
6179 struct my_unpacked_struct s;
6183 You may only specify this attribute on the definition of an @code{enum},
6184 @code{struct} or @code{union}, not on a @code{typedef} that does not
6185 also define the enumerated type, structure or union.
6187 @item bnd_variable_size
6188 @cindex @code{bnd_variable_size} type attribute
6189 @cindex Pointer Bounds Checker attributes
6190 When applied to a structure field, this attribute tells Pointer
6191 Bounds Checker that the size of this field should not be computed
6192 using static type information. It may be used to mark variably-sized
6193 static array fields placed at the end of a structure.
6201 S *p = (S *)malloc (sizeof(S) + 100);
6202 p->data[10] = 0; //Bounds violation
6206 By using an attribute for the field we may avoid unwanted bound
6213 char data[1] __attribute__((bnd_variable_size));
6215 S *p = (S *)malloc (sizeof(S) + 100);
6216 p->data[10] = 0; //OK
6220 @itemx deprecated (@var{msg})
6221 @cindex @code{deprecated} type attribute
6222 The @code{deprecated} attribute results in a warning if the type
6223 is used anywhere in the source file. This is useful when identifying
6224 types that are expected to be removed in a future version of a program.
6225 If possible, the warning also includes the location of the declaration
6226 of the deprecated type, to enable users to easily find further
6227 information about why the type is deprecated, or what they should do
6228 instead. Note that the warnings only occur for uses and then only
6229 if the type is being applied to an identifier that itself is not being
6230 declared as deprecated.
6233 typedef int T1 __attribute__ ((deprecated));
6237 typedef T1 T3 __attribute__ ((deprecated));
6238 T3 z __attribute__ ((deprecated));
6242 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6243 warning is issued for line 4 because T2 is not explicitly
6244 deprecated. Line 5 has no warning because T3 is explicitly
6245 deprecated. Similarly for line 6. The optional @var{msg}
6246 argument, which must be a string, is printed in the warning if
6249 The @code{deprecated} attribute can also be used for functions and
6250 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6252 @item designated_init
6253 @cindex @code{designated_init} type attribute
6254 This attribute may only be applied to structure types. It indicates
6255 that any initialization of an object of this type must use designated
6256 initializers rather than positional initializers. The intent of this
6257 attribute is to allow the programmer to indicate that a structure's
6258 layout may change, and that therefore relying on positional
6259 initialization will result in future breakage.
6261 GCC emits warnings based on this attribute by default; use
6262 @option{-Wno-designated-init} to suppress them.
6265 @cindex @code{may_alias} type attribute
6266 Accesses through pointers to types with this attribute are not subject
6267 to type-based alias analysis, but are instead assumed to be able to alias
6268 any other type of objects.
6269 In the context of section 6.5 paragraph 7 of the C99 standard,
6270 an lvalue expression
6271 dereferencing such a pointer is treated like having a character type.
6272 See @option{-fstrict-aliasing} for more information on aliasing issues.
6273 This extension exists to support some vector APIs, in which pointers to
6274 one vector type are permitted to alias pointers to a different vector type.
6276 Note that an object of a type with this attribute does not have any
6282 typedef short __attribute__((__may_alias__)) short_a;
6288 short_a *b = (short_a *) &a;
6292 if (a == 0x12345678)
6300 If you replaced @code{short_a} with @code{short} in the variable
6301 declaration, the above program would abort when compiled with
6302 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6306 @cindex @code{packed} type attribute
6307 This attribute, attached to @code{struct} or @code{union} type
6308 definition, specifies that each member (other than zero-width bit-fields)
6309 of the structure or union is placed to minimize the memory required. When
6310 attached to an @code{enum} definition, it indicates that the smallest
6311 integral type should be used.
6313 @item transparent_union
6314 @cindex @code{transparent_union} type attribute
6316 This attribute, attached to a @code{union} type definition, indicates
6317 that any function parameter having that union type causes calls to that
6318 function to be treated in a special way.
6320 First, the argument corresponding to a transparent union type can be of
6321 any type in the union; no cast is required. Also, if the union contains
6322 a pointer type, the corresponding argument can be a null pointer
6323 constant or a void pointer expression; and if the union contains a void
6324 pointer type, the corresponding argument can be any pointer expression.
6325 If the union member type is a pointer, qualifiers like @code{const} on
6326 the referenced type must be respected, just as with normal pointer
6329 Second, the argument is passed to the function using the calling
6330 conventions of the first member of the transparent union, not the calling
6331 conventions of the union itself. All members of the union must have the
6332 same machine representation; this is necessary for this argument passing
6335 Transparent unions are designed for library functions that have multiple
6336 interfaces for compatibility reasons. For example, suppose the
6337 @code{wait} function must accept either a value of type @code{int *} to
6338 comply with POSIX, or a value of type @code{union wait *} to comply with
6339 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6340 @code{wait} would accept both kinds of arguments, but it would also
6341 accept any other pointer type and this would make argument type checking
6342 less useful. Instead, @code{<sys/wait.h>} might define the interface
6346 typedef union __attribute__ ((__transparent_union__))
6350 @} wait_status_ptr_t;
6352 pid_t wait (wait_status_ptr_t);
6356 This interface allows either @code{int *} or @code{union wait *}
6357 arguments to be passed, using the @code{int *} calling convention.
6358 The program can call @code{wait} with arguments of either type:
6361 int w1 () @{ int w; return wait (&w); @}
6362 int w2 () @{ union wait w; return wait (&w); @}
6366 With this interface, @code{wait}'s implementation might look like this:
6369 pid_t wait (wait_status_ptr_t p)
6371 return waitpid (-1, p.__ip, 0);
6376 @cindex @code{unused} type attribute
6377 When attached to a type (including a @code{union} or a @code{struct}),
6378 this attribute means that variables of that type are meant to appear
6379 possibly unused. GCC does not produce a warning for any variables of
6380 that type, even if the variable appears to do nothing. This is often
6381 the case with lock or thread classes, which are usually defined and then
6382 not referenced, but contain constructors and destructors that have
6383 nontrivial bookkeeping functions.
6386 @cindex @code{visibility} type attribute
6387 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6388 applied to class, struct, union and enum types. Unlike other type
6389 attributes, the attribute must appear between the initial keyword and
6390 the name of the type; it cannot appear after the body of the type.
6392 Note that the type visibility is applied to vague linkage entities
6393 associated with the class (vtable, typeinfo node, etc.). In
6394 particular, if a class is thrown as an exception in one shared object
6395 and caught in another, the class must have default visibility.
6396 Otherwise the two shared objects are unable to use the same
6397 typeinfo node and exception handling will break.
6401 To specify multiple attributes, separate them by commas within the
6402 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6405 @node ARM Type Attributes
6406 @subsection ARM Type Attributes
6408 @cindex @code{notshared} type attribute, ARM
6409 On those ARM targets that support @code{dllimport} (such as Symbian
6410 OS), you can use the @code{notshared} attribute to indicate that the
6411 virtual table and other similar data for a class should not be
6412 exported from a DLL@. For example:
6415 class __declspec(notshared) C @{
6417 __declspec(dllimport) C();
6421 __declspec(dllexport)
6426 In this code, @code{C::C} is exported from the current DLL, but the
6427 virtual table for @code{C} is not exported. (You can use
6428 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6429 most Symbian OS code uses @code{__declspec}.)
6431 @node MeP Type Attributes
6432 @subsection MeP Type Attributes
6434 @cindex @code{based} type attribute, MeP
6435 @cindex @code{tiny} type attribute, MeP
6436 @cindex @code{near} type attribute, MeP
6437 @cindex @code{far} type attribute, MeP
6438 Many of the MeP variable attributes may be applied to types as well.
6439 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6440 @code{far} attributes may be applied to either. The @code{io} and
6441 @code{cb} attributes may not be applied to types.
6443 @node PowerPC Type Attributes
6444 @subsection PowerPC Type Attributes
6446 Three attributes currently are defined for PowerPC configurations:
6447 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6449 @cindex @code{ms_struct} type attribute, PowerPC
6450 @cindex @code{gcc_struct} type attribute, PowerPC
6451 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6452 attributes please see the documentation in @ref{x86 Type Attributes}.
6454 @cindex @code{altivec} type attribute, PowerPC
6455 The @code{altivec} attribute allows one to declare AltiVec vector data
6456 types supported by the AltiVec Programming Interface Manual. The
6457 attribute requires an argument to specify one of three vector types:
6458 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6459 and @code{bool__} (always followed by unsigned).
6462 __attribute__((altivec(vector__)))
6463 __attribute__((altivec(pixel__))) unsigned short
6464 __attribute__((altivec(bool__))) unsigned
6467 These attributes mainly are intended to support the @code{__vector},
6468 @code{__pixel}, and @code{__bool} AltiVec keywords.
6470 @node SPU Type Attributes
6471 @subsection SPU Type Attributes
6473 @cindex @code{spu_vector} type attribute, SPU
6474 The SPU supports the @code{spu_vector} attribute for types. This attribute
6475 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6476 Language Extensions Specification. It is intended to support the
6477 @code{__vector} keyword.
6479 @node x86 Type Attributes
6480 @subsection x86 Type Attributes
6482 Two attributes are currently defined for x86 configurations:
6483 @code{ms_struct} and @code{gcc_struct}.
6489 @cindex @code{ms_struct} type attribute, x86
6490 @cindex @code{gcc_struct} type attribute, x86
6492 If @code{packed} is used on a structure, or if bit-fields are used
6493 it may be that the Microsoft ABI packs them differently
6494 than GCC normally packs them. Particularly when moving packed
6495 data between functions compiled with GCC and the native Microsoft compiler
6496 (either via function call or as data in a file), it may be necessary to access
6499 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
6500 compilers to match the native Microsoft compiler.
6503 @node Label Attributes
6504 @section Label Attributes
6505 @cindex Label Attributes
6507 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6508 details of the exact syntax for using attributes. Other attributes are
6509 available for functions (@pxref{Function Attributes}), variables
6510 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6511 and for types (@pxref{Type Attributes}).
6513 This example uses the @code{cold} label attribute to indicate the
6514 @code{ErrorHandling} branch is unlikely to be taken and that the
6515 @code{ErrorHandling} label is unused:
6519 asm goto ("some asm" : : : : NoError);
6521 /* This branch (the fall-through from the asm) is less commonly used */
6523 __attribute__((cold, unused)); /* Semi-colon is required here */
6528 printf("no error\n");
6534 @cindex @code{unused} label attribute
6535 This feature is intended for program-generated code that may contain
6536 unused labels, but which is compiled with @option{-Wall}. It is
6537 not normally appropriate to use in it human-written code, though it
6538 could be useful in cases where the code that jumps to the label is
6539 contained within an @code{#ifdef} conditional.
6542 @cindex @code{hot} label attribute
6543 The @code{hot} attribute on a label is used to inform the compiler that
6544 the path following the label is more likely than paths that are not so
6545 annotated. This attribute is used in cases where @code{__builtin_expect}
6546 cannot be used, for instance with computed goto or @code{asm goto}.
6549 @cindex @code{cold} label attribute
6550 The @code{cold} attribute on labels is used to inform the compiler that
6551 the path following the label is unlikely to be executed. This attribute
6552 is used in cases where @code{__builtin_expect} cannot be used, for instance
6553 with computed goto or @code{asm goto}.
6557 @node Enumerator Attributes
6558 @section Enumerator Attributes
6559 @cindex Enumerator Attributes
6561 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6562 details of the exact syntax for using attributes. Other attributes are
6563 available for functions (@pxref{Function Attributes}), variables
6564 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6565 and for types (@pxref{Type Attributes}).
6567 This example uses the @code{deprecated} enumerator attribute to indicate the
6568 @code{oldval} enumerator is deprecated:
6572 oldval __attribute__((deprecated)),
6585 @cindex @code{deprecated} enumerator attribute
6586 The @code{deprecated} attribute results in a warning if the enumerator
6587 is used anywhere in the source file. This is useful when identifying
6588 enumerators that are expected to be removed in a future version of a
6589 program. The warning also includes the location of the declaration
6590 of the deprecated enumerator, to enable users to easily find further
6591 information about why the enumerator is deprecated, or what they should
6592 do instead. Note that the warnings only occurs for uses.
6596 @node Attribute Syntax
6597 @section Attribute Syntax
6598 @cindex attribute syntax
6600 This section describes the syntax with which @code{__attribute__} may be
6601 used, and the constructs to which attribute specifiers bind, for the C
6602 language. Some details may vary for C++ and Objective-C@. Because of
6603 infelicities in the grammar for attributes, some forms described here
6604 may not be successfully parsed in all cases.
6606 There are some problems with the semantics of attributes in C++. For
6607 example, there are no manglings for attributes, although they may affect
6608 code generation, so problems may arise when attributed types are used in
6609 conjunction with templates or overloading. Similarly, @code{typeid}
6610 does not distinguish between types with different attributes. Support
6611 for attributes in C++ may be restricted in future to attributes on
6612 declarations only, but not on nested declarators.
6614 @xref{Function Attributes}, for details of the semantics of attributes
6615 applying to functions. @xref{Variable Attributes}, for details of the
6616 semantics of attributes applying to variables. @xref{Type Attributes},
6617 for details of the semantics of attributes applying to structure, union
6618 and enumerated types.
6619 @xref{Label Attributes}, for details of the semantics of attributes
6621 @xref{Enumerator Attributes}, for details of the semantics of attributes
6622 applying to enumerators.
6624 An @dfn{attribute specifier} is of the form
6625 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6626 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6627 each attribute is one of the following:
6631 Empty. Empty attributes are ignored.
6635 (which may be an identifier such as @code{unused}, or a reserved
6636 word such as @code{const}).
6639 An attribute name followed by a parenthesized list of
6640 parameters for the attribute.
6641 These parameters take one of the following forms:
6645 An identifier. For example, @code{mode} attributes use this form.
6648 An identifier followed by a comma and a non-empty comma-separated list
6649 of expressions. For example, @code{format} attributes use this form.
6652 A possibly empty comma-separated list of expressions. For example,
6653 @code{format_arg} attributes use this form with the list being a single
6654 integer constant expression, and @code{alias} attributes use this form
6655 with the list being a single string constant.
6659 An @dfn{attribute specifier list} is a sequence of one or more attribute
6660 specifiers, not separated by any other tokens.
6662 You may optionally specify attribute names with @samp{__}
6663 preceding and following the name.
6664 This allows you to use them in header files without
6665 being concerned about a possible macro of the same name. For example,
6666 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6669 @subsubheading Label Attributes
6671 In GNU C, an attribute specifier list may appear after the colon following a
6672 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6673 attributes on labels if the attribute specifier is immediately
6674 followed by a semicolon (i.e., the label applies to an empty
6675 statement). If the semicolon is missing, C++ label attributes are
6676 ambiguous, as it is permissible for a declaration, which could begin
6677 with an attribute list, to be labelled in C++. Declarations cannot be
6678 labelled in C90 or C99, so the ambiguity does not arise there.
6680 @subsubheading Enumerator Attributes
6682 In GNU C, an attribute specifier list may appear as part of an enumerator.
6683 The attribute goes after the enumeration constant, before @code{=}, if
6684 present. The optional attribute in the enumerator appertains to the
6685 enumeration constant. It is not possible to place the attribute after
6686 the constant expression, if present.
6688 @subsubheading Type Attributes
6690 An attribute specifier list may appear as part of a @code{struct},
6691 @code{union} or @code{enum} specifier. It may go either immediately
6692 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6693 the closing brace. The former syntax is preferred.
6694 Where attribute specifiers follow the closing brace, they are considered
6695 to relate to the structure, union or enumerated type defined, not to any
6696 enclosing declaration the type specifier appears in, and the type
6697 defined is not complete until after the attribute specifiers.
6698 @c Otherwise, there would be the following problems: a shift/reduce
6699 @c conflict between attributes binding the struct/union/enum and
6700 @c binding to the list of specifiers/qualifiers; and "aligned"
6701 @c attributes could use sizeof for the structure, but the size could be
6702 @c changed later by "packed" attributes.
6705 @subsubheading All other attributes
6707 Otherwise, an attribute specifier appears as part of a declaration,
6708 counting declarations of unnamed parameters and type names, and relates
6709 to that declaration (which may be nested in another declaration, for
6710 example in the case of a parameter declaration), or to a particular declarator
6711 within a declaration. Where an
6712 attribute specifier is applied to a parameter declared as a function or
6713 an array, it should apply to the function or array rather than the
6714 pointer to which the parameter is implicitly converted, but this is not
6715 yet correctly implemented.
6717 Any list of specifiers and qualifiers at the start of a declaration may
6718 contain attribute specifiers, whether or not such a list may in that
6719 context contain storage class specifiers. (Some attributes, however,
6720 are essentially in the nature of storage class specifiers, and only make
6721 sense where storage class specifiers may be used; for example,
6722 @code{section}.) There is one necessary limitation to this syntax: the
6723 first old-style parameter declaration in a function definition cannot
6724 begin with an attribute specifier, because such an attribute applies to
6725 the function instead by syntax described below (which, however, is not
6726 yet implemented in this case). In some other cases, attribute
6727 specifiers are permitted by this grammar but not yet supported by the
6728 compiler. All attribute specifiers in this place relate to the
6729 declaration as a whole. In the obsolescent usage where a type of
6730 @code{int} is implied by the absence of type specifiers, such a list of
6731 specifiers and qualifiers may be an attribute specifier list with no
6732 other specifiers or qualifiers.
6734 At present, the first parameter in a function prototype must have some
6735 type specifier that is not an attribute specifier; this resolves an
6736 ambiguity in the interpretation of @code{void f(int
6737 (__attribute__((foo)) x))}, but is subject to change. At present, if
6738 the parentheses of a function declarator contain only attributes then
6739 those attributes are ignored, rather than yielding an error or warning
6740 or implying a single parameter of type int, but this is subject to
6743 An attribute specifier list may appear immediately before a declarator
6744 (other than the first) in a comma-separated list of declarators in a
6745 declaration of more than one identifier using a single list of
6746 specifiers and qualifiers. Such attribute specifiers apply
6747 only to the identifier before whose declarator they appear. For
6751 __attribute__((noreturn)) void d0 (void),
6752 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6757 the @code{noreturn} attribute applies to all the functions
6758 declared; the @code{format} attribute only applies to @code{d1}.
6760 An attribute specifier list may appear immediately before the comma,
6761 @code{=} or semicolon terminating the declaration of an identifier other
6762 than a function definition. Such attribute specifiers apply
6763 to the declared object or function. Where an
6764 assembler name for an object or function is specified (@pxref{Asm
6765 Labels}), the attribute must follow the @code{asm}
6768 An attribute specifier list may, in future, be permitted to appear after
6769 the declarator in a function definition (before any old-style parameter
6770 declarations or the function body).
6772 Attribute specifiers may be mixed with type qualifiers appearing inside
6773 the @code{[]} of a parameter array declarator, in the C99 construct by
6774 which such qualifiers are applied to the pointer to which the array is
6775 implicitly converted. Such attribute specifiers apply to the pointer,
6776 not to the array, but at present this is not implemented and they are
6779 An attribute specifier list may appear at the start of a nested
6780 declarator. At present, there are some limitations in this usage: the
6781 attributes correctly apply to the declarator, but for most individual
6782 attributes the semantics this implies are not implemented.
6783 When attribute specifiers follow the @code{*} of a pointer
6784 declarator, they may be mixed with any type qualifiers present.
6785 The following describes the formal semantics of this syntax. It makes the
6786 most sense if you are familiar with the formal specification of
6787 declarators in the ISO C standard.
6789 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
6790 D1}, where @code{T} contains declaration specifiers that specify a type
6791 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
6792 contains an identifier @var{ident}. The type specified for @var{ident}
6793 for derived declarators whose type does not include an attribute
6794 specifier is as in the ISO C standard.
6796 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
6797 and the declaration @code{T D} specifies the type
6798 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6799 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6800 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
6802 If @code{D1} has the form @code{*
6803 @var{type-qualifier-and-attribute-specifier-list} D}, and the
6804 declaration @code{T D} specifies the type
6805 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6806 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6807 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
6813 void (__attribute__((noreturn)) ****f) (void);
6817 specifies the type ``pointer to pointer to pointer to pointer to
6818 non-returning function returning @code{void}''. As another example,
6821 char *__attribute__((aligned(8))) *f;
6825 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
6826 Note again that this does not work with most attributes; for example,
6827 the usage of @samp{aligned} and @samp{noreturn} attributes given above
6828 is not yet supported.
6830 For compatibility with existing code written for compiler versions that
6831 did not implement attributes on nested declarators, some laxity is
6832 allowed in the placing of attributes. If an attribute that only applies
6833 to types is applied to a declaration, it is treated as applying to
6834 the type of that declaration. If an attribute that only applies to
6835 declarations is applied to the type of a declaration, it is treated
6836 as applying to that declaration; and, for compatibility with code
6837 placing the attributes immediately before the identifier declared, such
6838 an attribute applied to a function return type is treated as
6839 applying to the function type, and such an attribute applied to an array
6840 element type is treated as applying to the array type. If an
6841 attribute that only applies to function types is applied to a
6842 pointer-to-function type, it is treated as applying to the pointer
6843 target type; if such an attribute is applied to a function return type
6844 that is not a pointer-to-function type, it is treated as applying
6845 to the function type.
6847 @node Function Prototypes
6848 @section Prototypes and Old-Style Function Definitions
6849 @cindex function prototype declarations
6850 @cindex old-style function definitions
6851 @cindex promotion of formal parameters
6853 GNU C extends ISO C to allow a function prototype to override a later
6854 old-style non-prototype definition. Consider the following example:
6857 /* @r{Use prototypes unless the compiler is old-fashioned.} */
6864 /* @r{Prototype function declaration.} */
6865 int isroot P((uid_t));
6867 /* @r{Old-style function definition.} */
6869 isroot (x) /* @r{??? lossage here ???} */
6876 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
6877 not allow this example, because subword arguments in old-style
6878 non-prototype definitions are promoted. Therefore in this example the
6879 function definition's argument is really an @code{int}, which does not
6880 match the prototype argument type of @code{short}.
6882 This restriction of ISO C makes it hard to write code that is portable
6883 to traditional C compilers, because the programmer does not know
6884 whether the @code{uid_t} type is @code{short}, @code{int}, or
6885 @code{long}. Therefore, in cases like these GNU C allows a prototype
6886 to override a later old-style definition. More precisely, in GNU C, a
6887 function prototype argument type overrides the argument type specified
6888 by a later old-style definition if the former type is the same as the
6889 latter type before promotion. Thus in GNU C the above example is
6890 equivalent to the following:
6903 GNU C++ does not support old-style function definitions, so this
6904 extension is irrelevant.
6907 @section C++ Style Comments
6909 @cindex C++ comments
6910 @cindex comments, C++ style
6912 In GNU C, you may use C++ style comments, which start with @samp{//} and
6913 continue until the end of the line. Many other C implementations allow
6914 such comments, and they are included in the 1999 C standard. However,
6915 C++ style comments are not recognized if you specify an @option{-std}
6916 option specifying a version of ISO C before C99, or @option{-ansi}
6917 (equivalent to @option{-std=c90}).
6920 @section Dollar Signs in Identifier Names
6922 @cindex dollar signs in identifier names
6923 @cindex identifier names, dollar signs in
6925 In GNU C, you may normally use dollar signs in identifier names.
6926 This is because many traditional C implementations allow such identifiers.
6927 However, dollar signs in identifiers are not supported on a few target
6928 machines, typically because the target assembler does not allow them.
6930 @node Character Escapes
6931 @section The Character @key{ESC} in Constants
6933 You can use the sequence @samp{\e} in a string or character constant to
6934 stand for the ASCII character @key{ESC}.
6937 @section Inquiring on Alignment of Types or Variables
6939 @cindex type alignment
6940 @cindex variable alignment
6942 The keyword @code{__alignof__} allows you to inquire about how an object
6943 is aligned, or the minimum alignment usually required by a type. Its
6944 syntax is just like @code{sizeof}.
6946 For example, if the target machine requires a @code{double} value to be
6947 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
6948 This is true on many RISC machines. On more traditional machine
6949 designs, @code{__alignof__ (double)} is 4 or even 2.
6951 Some machines never actually require alignment; they allow reference to any
6952 data type even at an odd address. For these machines, @code{__alignof__}
6953 reports the smallest alignment that GCC gives the data type, usually as
6954 mandated by the target ABI.
6956 If the operand of @code{__alignof__} is an lvalue rather than a type,
6957 its value is the required alignment for its type, taking into account
6958 any minimum alignment specified with GCC's @code{__attribute__}
6959 extension (@pxref{Variable Attributes}). For example, after this
6963 struct foo @{ int x; char y; @} foo1;
6967 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
6968 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
6970 It is an error to ask for the alignment of an incomplete type.
6974 @section An Inline Function is As Fast As a Macro
6975 @cindex inline functions
6976 @cindex integrating function code
6978 @cindex macros, inline alternative
6980 By declaring a function inline, you can direct GCC to make
6981 calls to that function faster. One way GCC can achieve this is to
6982 integrate that function's code into the code for its callers. This
6983 makes execution faster by eliminating the function-call overhead; in
6984 addition, if any of the actual argument values are constant, their
6985 known values may permit simplifications at compile time so that not
6986 all of the inline function's code needs to be included. The effect on
6987 code size is less predictable; object code may be larger or smaller
6988 with function inlining, depending on the particular case. You can
6989 also direct GCC to try to integrate all ``simple enough'' functions
6990 into their callers with the option @option{-finline-functions}.
6992 GCC implements three different semantics of declaring a function
6993 inline. One is available with @option{-std=gnu89} or
6994 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
6995 on all inline declarations, another when
6996 @option{-std=c99}, @option{-std=c11},
6997 @option{-std=gnu99} or @option{-std=gnu11}
6998 (without @option{-fgnu89-inline}), and the third
6999 is used when compiling C++.
7001 To declare a function inline, use the @code{inline} keyword in its
7002 declaration, like this:
7012 If you are writing a header file to be included in ISO C90 programs, write
7013 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7015 The three types of inlining behave similarly in two important cases:
7016 when the @code{inline} keyword is used on a @code{static} function,
7017 like the example above, and when a function is first declared without
7018 using the @code{inline} keyword and then is defined with
7019 @code{inline}, like this:
7022 extern int inc (int *a);
7030 In both of these common cases, the program behaves the same as if you
7031 had not used the @code{inline} keyword, except for its speed.
7033 @cindex inline functions, omission of
7034 @opindex fkeep-inline-functions
7035 When a function is both inline and @code{static}, if all calls to the
7036 function are integrated into the caller, and the function's address is
7037 never used, then the function's own assembler code is never referenced.
7038 In this case, GCC does not actually output assembler code for the
7039 function, unless you specify the option @option{-fkeep-inline-functions}.
7040 Some calls cannot be integrated for various reasons (in particular,
7041 calls that precede the function's definition cannot be integrated, and
7042 neither can recursive calls within the definition). If there is a
7043 nonintegrated call, then the function is compiled to assembler code as
7044 usual. The function must also be compiled as usual if the program
7045 refers to its address, because that can't be inlined.
7048 Note that certain usages in a function definition can make it unsuitable
7049 for inline substitution. Among these usages are: variadic functions, use of
7050 @code{alloca}, use of variable-length data types (@pxref{Variable Length}),
7051 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
7052 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
7053 warns when a function marked @code{inline} could not be substituted,
7054 and gives the reason for the failure.
7056 @cindex automatic @code{inline} for C++ member fns
7057 @cindex @code{inline} automatic for C++ member fns
7058 @cindex member fns, automatically @code{inline}
7059 @cindex C++ member fns, automatically @code{inline}
7060 @opindex fno-default-inline
7061 As required by ISO C++, GCC considers member functions defined within
7062 the body of a class to be marked inline even if they are
7063 not explicitly declared with the @code{inline} keyword. You can
7064 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7065 Options,,Options Controlling C++ Dialect}.
7067 GCC does not inline any functions when not optimizing unless you specify
7068 the @samp{always_inline} attribute for the function, like this:
7071 /* @r{Prototype.} */
7072 inline void foo (const char) __attribute__((always_inline));
7075 The remainder of this section is specific to GNU C90 inlining.
7077 @cindex non-static inline function
7078 When an inline function is not @code{static}, then the compiler must assume
7079 that there may be calls from other source files; since a global symbol can
7080 be defined only once in any program, the function must not be defined in
7081 the other source files, so the calls therein cannot be integrated.
7082 Therefore, a non-@code{static} inline function is always compiled on its
7083 own in the usual fashion.
7085 If you specify both @code{inline} and @code{extern} in the function
7086 definition, then the definition is used only for inlining. In no case
7087 is the function compiled on its own, not even if you refer to its
7088 address explicitly. Such an address becomes an external reference, as
7089 if you had only declared the function, and had not defined it.
7091 This combination of @code{inline} and @code{extern} has almost the
7092 effect of a macro. The way to use it is to put a function definition in
7093 a header file with these keywords, and put another copy of the
7094 definition (lacking @code{inline} and @code{extern}) in a library file.
7095 The definition in the header file causes most calls to the function
7096 to be inlined. If any uses of the function remain, they refer to
7097 the single copy in the library.
7100 @section When is a Volatile Object Accessed?
7101 @cindex accessing volatiles
7102 @cindex volatile read
7103 @cindex volatile write
7104 @cindex volatile access
7106 C has the concept of volatile objects. These are normally accessed by
7107 pointers and used for accessing hardware or inter-thread
7108 communication. The standard encourages compilers to refrain from
7109 optimizations concerning accesses to volatile objects, but leaves it
7110 implementation defined as to what constitutes a volatile access. The
7111 minimum requirement is that at a sequence point all previous accesses
7112 to volatile objects have stabilized and no subsequent accesses have
7113 occurred. Thus an implementation is free to reorder and combine
7114 volatile accesses that occur between sequence points, but cannot do
7115 so for accesses across a sequence point. The use of volatile does
7116 not allow you to violate the restriction on updating objects multiple
7117 times between two sequence points.
7119 Accesses to non-volatile objects are not ordered with respect to
7120 volatile accesses. You cannot use a volatile object as a memory
7121 barrier to order a sequence of writes to non-volatile memory. For
7125 int *ptr = @var{something};
7127 *ptr = @var{something};
7132 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7133 that the write to @var{*ptr} occurs by the time the update
7134 of @var{vobj} happens. If you need this guarantee, you must use
7135 a stronger memory barrier such as:
7138 int *ptr = @var{something};
7140 *ptr = @var{something};
7141 asm volatile ("" : : : "memory");
7145 A scalar volatile object is read when it is accessed in a void context:
7148 volatile int *src = @var{somevalue};
7152 Such expressions are rvalues, and GCC implements this as a
7153 read of the volatile object being pointed to.
7155 Assignments are also expressions and have an rvalue. However when
7156 assigning to a scalar volatile, the volatile object is not reread,
7157 regardless of whether the assignment expression's rvalue is used or
7158 not. If the assignment's rvalue is used, the value is that assigned
7159 to the volatile object. For instance, there is no read of @var{vobj}
7160 in all the following cases:
7165 vobj = @var{something};
7166 obj = vobj = @var{something};
7167 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7168 obj = (@var{something}, vobj = @var{anotherthing});
7171 If you need to read the volatile object after an assignment has
7172 occurred, you must use a separate expression with an intervening
7175 As bit-fields are not individually addressable, volatile bit-fields may
7176 be implicitly read when written to, or when adjacent bit-fields are
7177 accessed. Bit-field operations may be optimized such that adjacent
7178 bit-fields are only partially accessed, if they straddle a storage unit
7179 boundary. For these reasons it is unwise to use volatile bit-fields to
7182 @node Using Assembly Language with C
7183 @section How to Use Inline Assembly Language in C Code
7184 @cindex @code{asm} keyword
7185 @cindex assembly language in C
7186 @cindex inline assembly language
7187 @cindex mixing assembly language and C
7189 The @code{asm} keyword allows you to embed assembler instructions
7190 within C code. GCC provides two forms of inline @code{asm}
7191 statements. A @dfn{basic @code{asm}} statement is one with no
7192 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7193 statement (@pxref{Extended Asm}) includes one or more operands.
7194 The extended form is preferred for mixing C and assembly language
7195 within a function, but to include assembly language at
7196 top level you must use basic @code{asm}.
7198 You can also use the @code{asm} keyword to override the assembler name
7199 for a C symbol, or to place a C variable in a specific register.
7202 * Basic Asm:: Inline assembler without operands.
7203 * Extended Asm:: Inline assembler with operands.
7204 * Constraints:: Constraints for @code{asm} operands
7205 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7206 * Explicit Reg Vars:: Defining variables residing in specified registers.
7207 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7211 @subsection Basic Asm --- Assembler Instructions Without Operands
7212 @cindex basic @code{asm}
7213 @cindex assembly language in C, basic
7215 A basic @code{asm} statement has the following syntax:
7218 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7221 The @code{asm} keyword is a GNU extension.
7222 When writing code that can be compiled with @option{-ansi} and the
7223 various @option{-std} options, use @code{__asm__} instead of
7224 @code{asm} (@pxref{Alternate Keywords}).
7226 @subsubheading Qualifiers
7229 The optional @code{volatile} qualifier has no effect.
7230 All basic @code{asm} blocks are implicitly volatile.
7233 @subsubheading Parameters
7236 @item AssemblerInstructions
7237 This is a literal string that specifies the assembler code. The string can
7238 contain any instructions recognized by the assembler, including directives.
7239 GCC does not parse the assembler instructions themselves and
7240 does not know what they mean or even whether they are valid assembler input.
7242 You may place multiple assembler instructions together in a single @code{asm}
7243 string, separated by the characters normally used in assembly code for the
7244 system. A combination that works in most places is a newline to break the
7245 line, plus a tab character (written as @samp{\n\t}).
7246 Some assemblers allow semicolons as a line separator. However,
7247 note that some assembler dialects use semicolons to start a comment.
7250 @subsubheading Remarks
7251 Using extended @code{asm} typically produces smaller, safer, and more
7252 efficient code, and in most cases it is a better solution than basic
7253 @code{asm}. However, there are two situations where only basic @code{asm}
7258 Extended @code{asm} statements have to be inside a C
7259 function, so to write inline assembly language at file scope (``top-level''),
7260 outside of C functions, you must use basic @code{asm}.
7261 You can use this technique to emit assembler directives,
7262 define assembly language macros that can be invoked elsewhere in the file,
7263 or write entire functions in assembly language.
7267 with the @code{naked} attribute also require basic @code{asm}
7268 (@pxref{Function Attributes}).
7271 Safely accessing C data and calling functions from basic @code{asm} is more
7272 complex than it may appear. To access C data, it is better to use extended
7275 Do not expect a sequence of @code{asm} statements to remain perfectly
7276 consecutive after compilation. If certain instructions need to remain
7277 consecutive in the output, put them in a single multi-instruction @code{asm}
7278 statement. Note that GCC's optimizers can move @code{asm} statements
7279 relative to other code, including across jumps.
7281 @code{asm} statements may not perform jumps into other @code{asm} statements.
7282 GCC does not know about these jumps, and therefore cannot take
7283 account of them when deciding how to optimize. Jumps from @code{asm} to C
7284 labels are only supported in extended @code{asm}.
7286 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7287 assembly code when optimizing. This can lead to unexpected duplicate
7288 symbol errors during compilation if your assembly code defines symbols or
7291 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7292 visibility of any symbols it references. This may result in GCC discarding
7293 those symbols as unreferenced.
7295 The compiler copies the assembler instructions in a basic @code{asm}
7296 verbatim to the assembly language output file, without
7297 processing dialects or any of the @samp{%} operators that are available with
7298 extended @code{asm}. This results in minor differences between basic
7299 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7300 registers you might use @samp{%eax} in basic @code{asm} and
7301 @samp{%%eax} in extended @code{asm}.
7303 On targets such as x86 that support multiple assembler dialects,
7304 all basic @code{asm} blocks use the assembler dialect specified by the
7305 @option{-masm} command-line option (@pxref{x86 Options}).
7306 Basic @code{asm} provides no
7307 mechanism to provide different assembler strings for different dialects.
7309 Here is an example of basic @code{asm} for i386:
7312 /* Note that this code will not compile with -masm=intel */
7313 #define DebugBreak() asm("int $3")
7317 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7318 @cindex extended @code{asm}
7319 @cindex assembly language in C, extended
7321 With extended @code{asm} you can read and write C variables from
7322 assembler and perform jumps from assembler code to C labels.
7323 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7324 the operand parameters after the assembler template:
7327 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7328 : @var{OutputOperands}
7329 @r{[} : @var{InputOperands}
7330 @r{[} : @var{Clobbers} @r{]} @r{]})
7332 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7334 : @var{InputOperands}
7339 The @code{asm} keyword is a GNU extension.
7340 When writing code that can be compiled with @option{-ansi} and the
7341 various @option{-std} options, use @code{__asm__} instead of
7342 @code{asm} (@pxref{Alternate Keywords}).
7344 @subsubheading Qualifiers
7348 The typical use of extended @code{asm} statements is to manipulate input
7349 values to produce output values. However, your @code{asm} statements may
7350 also produce side effects. If so, you may need to use the @code{volatile}
7351 qualifier to disable certain optimizations. @xref{Volatile}.
7354 This qualifier informs the compiler that the @code{asm} statement may
7355 perform a jump to one of the labels listed in the @var{GotoLabels}.
7359 @subsubheading Parameters
7361 @item AssemblerTemplate
7362 This is a literal string that is the template for the assembler code. It is a
7363 combination of fixed text and tokens that refer to the input, output,
7364 and goto parameters. @xref{AssemblerTemplate}.
7366 @item OutputOperands
7367 A comma-separated list of the C variables modified by the instructions in the
7368 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7371 A comma-separated list of C expressions read by the instructions in the
7372 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7375 A comma-separated list of registers or other values changed by the
7376 @var{AssemblerTemplate}, beyond those listed as outputs.
7377 An empty list is permitted. @xref{Clobbers}.
7380 When you are using the @code{goto} form of @code{asm}, this section contains
7381 the list of all C labels to which the code in the
7382 @var{AssemblerTemplate} may jump.
7385 @code{asm} statements may not perform jumps into other @code{asm} statements,
7386 only to the listed @var{GotoLabels}.
7387 GCC's optimizers do not know about other jumps; therefore they cannot take
7388 account of them when deciding how to optimize.
7391 The total number of input + output + goto operands is limited to 30.
7393 @subsubheading Remarks
7394 The @code{asm} statement allows you to include assembly instructions directly
7395 within C code. This may help you to maximize performance in time-sensitive
7396 code or to access assembly instructions that are not readily available to C
7399 Note that extended @code{asm} statements must be inside a function. Only
7400 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7401 Functions declared with the @code{naked} attribute also require basic
7402 @code{asm} (@pxref{Function Attributes}).
7404 While the uses of @code{asm} are many and varied, it may help to think of an
7405 @code{asm} statement as a series of low-level instructions that convert input
7406 parameters to output parameters. So a simple (if not particularly useful)
7407 example for i386 using @code{asm} might look like this:
7413 asm ("mov %1, %0\n\t"
7418 printf("%d\n", dst);
7421 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7424 @subsubsection Volatile
7425 @cindex volatile @code{asm}
7426 @cindex @code{asm} volatile
7428 GCC's optimizers sometimes discard @code{asm} statements if they determine
7429 there is no need for the output variables. Also, the optimizers may move
7430 code out of loops if they believe that the code will always return the same
7431 result (i.e. none of its input values change between calls). Using the
7432 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7433 that have no output operands, including @code{asm goto} statements,
7434 are implicitly volatile.
7436 This i386 code demonstrates a case that does not use (or require) the
7437 @code{volatile} qualifier. If it is performing assertion checking, this code
7438 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7439 unreferenced by any code. As a result, the optimizers can discard the
7440 @code{asm} statement, which in turn removes the need for the entire
7441 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7442 isn't needed you allow the optimizers to produce the most efficient code
7446 void DoCheck(uint32_t dwSomeValue)
7450 // Assumes dwSomeValue is not zero.
7460 The next example shows a case where the optimizers can recognize that the input
7461 (@code{dwSomeValue}) never changes during the execution of the function and can
7462 therefore move the @code{asm} outside the loop to produce more efficient code.
7463 Again, using @code{volatile} disables this type of optimization.
7466 void do_print(uint32_t dwSomeValue)
7470 for (uint32_t x=0; x < 5; x++)
7472 // Assumes dwSomeValue is not zero.
7478 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7483 The following example demonstrates a case where you need to use the
7484 @code{volatile} qualifier.
7485 It uses the x86 @code{rdtsc} instruction, which reads
7486 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7487 the optimizers might assume that the @code{asm} block will always return the
7488 same value and therefore optimize away the second call.
7493 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7494 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7495 "or %%rdx, %0" // 'Or' in the lower bits.
7500 printf("msr: %llx\n", msr);
7504 // Reprint the timestamp
7505 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7506 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7507 "or %%rdx, %0" // 'Or' in the lower bits.
7512 printf("msr: %llx\n", msr);
7515 GCC's optimizers do not treat this code like the non-volatile code in the
7516 earlier examples. They do not move it out of loops or omit it on the
7517 assumption that the result from a previous call is still valid.
7519 Note that the compiler can move even volatile @code{asm} instructions relative
7520 to other code, including across jump instructions. For example, on many
7521 targets there is a system register that controls the rounding mode of
7522 floating-point operations. Setting it with a volatile @code{asm}, as in the
7523 following PowerPC example, does not work reliably.
7526 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7530 The compiler may move the addition back before the volatile @code{asm}. To
7531 make it work as expected, add an artificial dependency to the @code{asm} by
7532 referencing a variable in the subsequent code, for example:
7535 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7539 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7540 assembly code when optimizing. This can lead to unexpected duplicate symbol
7541 errors during compilation if your asm code defines symbols or labels.
7543 (@pxref{AssemblerTemplate}) may help resolve this problem.
7545 @anchor{AssemblerTemplate}
7546 @subsubsection Assembler Template
7547 @cindex @code{asm} assembler template
7549 An assembler template is a literal string containing assembler instructions.
7550 The compiler replaces tokens in the template that refer
7551 to inputs, outputs, and goto labels,
7552 and then outputs the resulting string to the assembler. The
7553 string can contain any instructions recognized by the assembler, including
7554 directives. GCC does not parse the assembler instructions
7555 themselves and does not know what they mean or even whether they are valid
7556 assembler input. However, it does count the statements
7557 (@pxref{Size of an asm}).
7559 You may place multiple assembler instructions together in a single @code{asm}
7560 string, separated by the characters normally used in assembly code for the
7561 system. A combination that works in most places is a newline to break the
7562 line, plus a tab character to move to the instruction field (written as
7564 Some assemblers allow semicolons as a line separator. However, note
7565 that some assembler dialects use semicolons to start a comment.
7567 Do not expect a sequence of @code{asm} statements to remain perfectly
7568 consecutive after compilation, even when you are using the @code{volatile}
7569 qualifier. If certain instructions need to remain consecutive in the output,
7570 put them in a single multi-instruction asm statement.
7572 Accessing data from C programs without using input/output operands (such as
7573 by using global symbols directly from the assembler template) may not work as
7574 expected. Similarly, calling functions directly from an assembler template
7575 requires a detailed understanding of the target assembler and ABI.
7577 Since GCC does not parse the assembler template,
7578 it has no visibility of any
7579 symbols it references. This may result in GCC discarding those symbols as
7580 unreferenced unless they are also listed as input, output, or goto operands.
7582 @subsubheading Special format strings
7584 In addition to the tokens described by the input, output, and goto operands,
7585 these tokens have special meanings in the assembler template:
7589 Outputs a single @samp{%} into the assembler code.
7592 Outputs a number that is unique to each instance of the @code{asm}
7593 statement in the entire compilation. This option is useful when creating local
7594 labels and referring to them multiple times in a single template that
7595 generates multiple assembler instructions.
7600 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7601 into the assembler code. When unescaped, these characters have special
7602 meaning to indicate multiple assembler dialects, as described below.
7605 @subsubheading Multiple assembler dialects in @code{asm} templates
7607 On targets such as x86, GCC supports multiple assembler dialects.
7608 The @option{-masm} option controls which dialect GCC uses as its
7609 default for inline assembler. The target-specific documentation for the
7610 @option{-masm} option contains the list of supported dialects, as well as the
7611 default dialect if the option is not specified. This information may be
7612 important to understand, since assembler code that works correctly when
7613 compiled using one dialect will likely fail if compiled using another.
7616 If your code needs to support multiple assembler dialects (for example, if
7617 you are writing public headers that need to support a variety of compilation
7618 options), use constructs of this form:
7621 @{ dialect0 | dialect1 | dialect2... @}
7624 This construct outputs @code{dialect0}
7625 when using dialect #0 to compile the code,
7626 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7627 braces than the number of dialects the compiler supports, the construct
7630 For example, if an x86 compiler supports two dialects
7631 (@samp{att}, @samp{intel}), an
7632 assembler template such as this:
7635 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7639 is equivalent to one of
7642 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7643 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7646 Using that same compiler, this code:
7649 "xchg@{l@}\t@{%%@}ebx, %1"
7653 corresponds to either
7656 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7657 "xchg\tebx, %1" @r{/* intel dialect */}
7660 There is no support for nesting dialect alternatives.
7662 @anchor{OutputOperands}
7663 @subsubsection Output Operands
7664 @cindex @code{asm} output operands
7666 An @code{asm} statement has zero or more output operands indicating the names
7667 of C variables modified by the assembler code.
7669 In this i386 example, @code{old} (referred to in the template string as
7670 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7671 (@code{%2}) is an input:
7676 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7677 "sbb %0,%0" // Use the CF to calculate old.
7678 : "=r" (old), "+rm" (*Base)
7685 Operands are separated by commas. Each operand has this format:
7688 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7692 @item asmSymbolicName
7693 Specifies a symbolic name for the operand.
7694 Reference the name in the assembler template
7695 by enclosing it in square brackets
7696 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7697 that contains the definition. Any valid C variable name is acceptable,
7698 including names already defined in the surrounding code. No two operands
7699 within the same @code{asm} statement can use the same symbolic name.
7701 When not using an @var{asmSymbolicName}, use the (zero-based) position
7703 in the list of operands in the assembler template. For example if there are
7704 three output operands, use @samp{%0} in the template to refer to the first,
7705 @samp{%1} for the second, and @samp{%2} for the third.
7708 A string constant specifying constraints on the placement of the operand;
7709 @xref{Constraints}, for details.
7711 Output constraints must begin with either @samp{=} (a variable overwriting an
7712 existing value) or @samp{+} (when reading and writing). When using
7713 @samp{=}, do not assume the location contains the existing value
7714 on entry to the @code{asm}, except
7715 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7717 After the prefix, there must be one or more additional constraints
7718 (@pxref{Constraints}) that describe where the value resides. Common
7719 constraints include @samp{r} for register and @samp{m} for memory.
7720 When you list more than one possible location (for example, @code{"=rm"}),
7721 the compiler chooses the most efficient one based on the current context.
7722 If you list as many alternates as the @code{asm} statement allows, you permit
7723 the optimizers to produce the best possible code.
7724 If you must use a specific register, but your Machine Constraints do not
7725 provide sufficient control to select the specific register you want,
7726 local register variables may provide a solution (@pxref{Local Reg Vars}).
7729 Specifies a C lvalue expression to hold the output, typically a variable name.
7730 The enclosing parentheses are a required part of the syntax.
7734 When the compiler selects the registers to use to
7735 represent the output operands, it does not use any of the clobbered registers
7738 Output operand expressions must be lvalues. The compiler cannot check whether
7739 the operands have data types that are reasonable for the instruction being
7740 executed. For output expressions that are not directly addressable (for
7741 example a bit-field), the constraint must allow a register. In that case, GCC
7742 uses the register as the output of the @code{asm}, and then stores that
7743 register into the output.
7745 Operands using the @samp{+} constraint modifier count as two operands
7746 (that is, both as input and output) towards the total maximum of 30 operands
7747 per @code{asm} statement.
7749 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7750 operands that must not overlap an input. Otherwise,
7751 GCC may allocate the output operand in the same register as an unrelated
7752 input operand, on the assumption that the assembler code consumes its
7753 inputs before producing outputs. This assumption may be false if the assembler
7754 code actually consists of more than one instruction.
7756 The same problem can occur if one output parameter (@var{a}) allows a register
7757 constraint and another output parameter (@var{b}) allows a memory constraint.
7758 The code generated by GCC to access the memory address in @var{b} can contain
7759 registers which @emph{might} be shared by @var{a}, and GCC considers those
7760 registers to be inputs to the asm. As above, GCC assumes that such input
7761 registers are consumed before any outputs are written. This assumption may
7762 result in incorrect behavior if the asm writes to @var{a} before using
7763 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7764 ensures that modifying @var{a} does not affect the address referenced by
7765 @var{b}. Otherwise, the location of @var{b}
7766 is undefined if @var{a} is modified before using @var{b}.
7768 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7769 instead of simply @samp{%2}). Typically these qualifiers are hardware
7770 dependent. The list of supported modifiers for x86 is found at
7771 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7773 If the C code that follows the @code{asm} makes no use of any of the output
7774 operands, use @code{volatile} for the @code{asm} statement to prevent the
7775 optimizers from discarding the @code{asm} statement as unneeded
7776 (see @ref{Volatile}).
7778 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7779 references the first output operand as @code{%0} (were there a second, it
7780 would be @code{%1}, etc). The number of the first input operand is one greater
7781 than that of the last output operand. In this i386 example, that makes
7782 @code{Mask} referenced as @code{%1}:
7785 uint32_t Mask = 1234;
7794 That code overwrites the variable @code{Index} (@samp{=}),
7795 placing the value in a register (@samp{r}).
7796 Using the generic @samp{r} constraint instead of a constraint for a specific
7797 register allows the compiler to pick the register to use, which can result
7798 in more efficient code. This may not be possible if an assembler instruction
7799 requires a specific register.
7801 The following i386 example uses the @var{asmSymbolicName} syntax.
7803 same result as the code above, but some may consider it more readable or more
7804 maintainable since reordering index numbers is not necessary when adding or
7805 removing operands. The names @code{aIndex} and @code{aMask}
7806 are only used in this example to emphasize which
7807 names get used where.
7808 It is acceptable to reuse the names @code{Index} and @code{Mask}.
7811 uint32_t Mask = 1234;
7814 asm ("bsfl %[aMask], %[aIndex]"
7815 : [aIndex] "=r" (Index)
7816 : [aMask] "r" (Mask)
7820 Here are some more examples of output operands.
7827 asm ("mov %[e], %[d]"
7832 Here, @code{d} may either be in a register or in memory. Since the compiler
7833 might already have the current value of the @code{uint32_t} location
7834 pointed to by @code{e}
7835 in a register, you can enable it to choose the best location
7836 for @code{d} by specifying both constraints.
7838 @anchor{FlagOutputOperands}
7839 @subsection Flag Output Operands
7840 @cindex @code{asm} flag output operands
7842 Some targets have a special register that holds the ``flags'' for the
7843 result of an operation or comparison. Normally, the contents of that
7844 register are either unmodifed by the asm, or the asm is considered to
7845 clobber the contents.
7847 On some targets, a special form of output operand exists by which
7848 conditions in the flags register may be outputs of the asm. The set of
7849 conditions supported are target specific, but the general rule is that
7850 the output variable must be a scalar integer, and the value will be boolean.
7851 When supported, the target will define the preprocessor symbol
7852 @code{__GCC_ASM_FLAG_OUTPUTS__}.
7854 Because of the special nature of the flag output operands, the constraint
7855 may not include alternatives.
7857 Most often, the target has only one flags register, and thus is an implied
7858 operand of many instructions. In this case, the operand should not be
7859 referenced within the assembler template via @code{%0} etc, as there's
7860 no corresponding text in the assembly language.
7864 The flag output constraints for the x86 family are of the form
7865 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
7866 conditions defined in the ISA manual for @code{j@var{cc}} or
7871 ``above'' or unsigned greater than
7873 ``above or equal'' or unsigned greater than or equal
7875 ``below'' or unsigned less than
7877 ``below or equal'' or unsigned less than or equal
7882 ``equal'' or zero flag set
7886 signed greater than or equal
7890 signed less than or equal
7911 ``not'' @var{flag}, or inverted versions of those above
7916 @anchor{InputOperands}
7917 @subsubsection Input Operands
7918 @cindex @code{asm} input operands
7919 @cindex @code{asm} expressions
7921 Input operands make values from C variables and expressions available to the
7924 Operands are separated by commas. Each operand has this format:
7927 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
7931 @item asmSymbolicName
7932 Specifies a symbolic name for the operand.
7933 Reference the name in the assembler template
7934 by enclosing it in square brackets
7935 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7936 that contains the definition. Any valid C variable name is acceptable,
7937 including names already defined in the surrounding code. No two operands
7938 within the same @code{asm} statement can use the same symbolic name.
7940 When not using an @var{asmSymbolicName}, use the (zero-based) position
7942 in the list of operands in the assembler template. For example if there are
7943 two output operands and three inputs,
7944 use @samp{%2} in the template to refer to the first input operand,
7945 @samp{%3} for the second, and @samp{%4} for the third.
7948 A string constant specifying constraints on the placement of the operand;
7949 @xref{Constraints}, for details.
7951 Input constraint strings may not begin with either @samp{=} or @samp{+}.
7952 When you list more than one possible location (for example, @samp{"irm"}),
7953 the compiler chooses the most efficient one based on the current context.
7954 If you must use a specific register, but your Machine Constraints do not
7955 provide sufficient control to select the specific register you want,
7956 local register variables may provide a solution (@pxref{Local Reg Vars}).
7958 Input constraints can also be digits (for example, @code{"0"}). This indicates
7959 that the specified input must be in the same place as the output constraint
7960 at the (zero-based) index in the output constraint list.
7961 When using @var{asmSymbolicName} syntax for the output operands,
7962 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
7965 This is the C variable or expression being passed to the @code{asm} statement
7966 as input. The enclosing parentheses are a required part of the syntax.
7970 When the compiler selects the registers to use to represent the input
7971 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
7973 If there are no output operands but there are input operands, place two
7974 consecutive colons where the output operands would go:
7977 __asm__ ("some instructions"
7979 : "r" (Offset / 8));
7982 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
7983 (except for inputs tied to outputs). The compiler assumes that on exit from
7984 the @code{asm} statement these operands contain the same values as they
7985 had before executing the statement.
7986 It is @emph{not} possible to use clobbers
7987 to inform the compiler that the values in these inputs are changing. One
7988 common work-around is to tie the changing input variable to an output variable
7989 that never gets used. Note, however, that if the code that follows the
7990 @code{asm} statement makes no use of any of the output operands, the GCC
7991 optimizers may discard the @code{asm} statement as unneeded
7992 (see @ref{Volatile}).
7994 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7995 instead of simply @samp{%2}). Typically these qualifiers are hardware
7996 dependent. The list of supported modifiers for x86 is found at
7997 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7999 In this example using the fictitious @code{combine} instruction, the
8000 constraint @code{"0"} for input operand 1 says that it must occupy the same
8001 location as output operand 0. Only input operands may use numbers in
8002 constraints, and they must each refer to an output operand. Only a number (or
8003 the symbolic assembler name) in the constraint can guarantee that one operand
8004 is in the same place as another. The mere fact that @code{foo} is the value of
8005 both operands is not enough to guarantee that they are in the same place in
8006 the generated assembler code.
8009 asm ("combine %2, %0"
8011 : "0" (foo), "g" (bar));
8014 Here is an example using symbolic names.
8017 asm ("cmoveq %1, %2, %[result]"
8018 : [result] "=r"(result)
8019 : "r" (test), "r" (new), "[result]" (old));
8023 @subsubsection Clobbers
8024 @cindex @code{asm} clobbers
8026 While the compiler is aware of changes to entries listed in the output
8027 operands, the inline @code{asm} code may modify more than just the outputs. For
8028 example, calculations may require additional registers, or the processor may
8029 overwrite a register as a side effect of a particular assembler instruction.
8030 In order to inform the compiler of these changes, list them in the clobber
8031 list. Clobber list items are either register names or the special clobbers
8032 (listed below). Each clobber list item is a string constant
8033 enclosed in double quotes and separated by commas.
8035 Clobber descriptions may not in any way overlap with an input or output
8036 operand. For example, you may not have an operand describing a register class
8037 with one member when listing that register in the clobber list. Variables
8038 declared to live in specific registers (@pxref{Explicit Reg Vars}) and used
8039 as @code{asm} input or output operands must have no part mentioned in the
8040 clobber description. In particular, there is no way to specify that input
8041 operands get modified without also specifying them as output operands.
8043 When the compiler selects which registers to use to represent input and output
8044 operands, it does not use any of the clobbered registers. As a result,
8045 clobbered registers are available for any use in the assembler code.
8047 Here is a realistic example for the VAX showing the use of clobbered
8051 asm volatile ("movc3 %0, %1, %2"
8053 : "g" (from), "g" (to), "g" (count)
8054 : "r0", "r1", "r2", "r3", "r4", "r5");
8057 Also, there are two special clobber arguments:
8061 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8062 register. On some machines, GCC represents the condition codes as a specific
8063 hardware register; @code{"cc"} serves to name this register.
8064 On other machines, condition code handling is different,
8065 and specifying @code{"cc"} has no effect. But
8066 it is valid no matter what the target.
8069 The @code{"memory"} clobber tells the compiler that the assembly code
8071 reads or writes to items other than those listed in the input and output
8072 operands (for example, accessing the memory pointed to by one of the input
8073 parameters). To ensure memory contains correct values, GCC may need to flush
8074 specific register values to memory before executing the @code{asm}. Further,
8075 the compiler does not assume that any values read from memory before an
8076 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8078 Using the @code{"memory"} clobber effectively forms a read/write
8079 memory barrier for the compiler.
8081 Note that this clobber does not prevent the @emph{processor} from doing
8082 speculative reads past the @code{asm} statement. To prevent that, you need
8083 processor-specific fence instructions.
8085 Flushing registers to memory has performance implications and may be an issue
8086 for time-sensitive code. You can use a trick to avoid this if the size of
8087 the memory being accessed is known at compile time. For example, if accessing
8088 ten bytes of a string, use a memory input like:
8090 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8095 @subsubsection Goto Labels
8096 @cindex @code{asm} goto labels
8098 @code{asm goto} allows assembly code to jump to one or more C labels. The
8099 @var{GotoLabels} section in an @code{asm goto} statement contains
8101 list of all C labels to which the assembler code may jump. GCC assumes that
8102 @code{asm} execution falls through to the next statement (if this is not the
8103 case, consider using the @code{__builtin_unreachable} intrinsic after the
8104 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8105 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8108 An @code{asm goto} statement cannot have outputs.
8109 This is due to an internal restriction of
8110 the compiler: control transfer instructions cannot have outputs.
8111 If the assembler code does modify anything, use the @code{"memory"} clobber
8113 optimizers to flush all register values to memory and reload them if
8114 necessary after the @code{asm} statement.
8116 Also note that an @code{asm goto} statement is always implicitly
8117 considered volatile.
8119 To reference a label in the assembler template,
8120 prefix it with @samp{%l} (lowercase @samp{L}) followed
8121 by its (zero-based) position in @var{GotoLabels} plus the number of input
8122 operands. For example, if the @code{asm} has three inputs and references two
8123 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8125 Alternately, you can reference labels using the actual C label name enclosed
8126 in brackets. For example, to reference a label named @code{carry}, you can
8127 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8128 section when using this approach.
8130 Here is an example of @code{asm goto} for i386:
8137 : "r" (p1), "r" (p2)
8147 The following example shows an @code{asm goto} that uses a memory clobber.
8153 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8164 @anchor{x86Operandmodifiers}
8165 @subsubsection x86 Operand Modifiers
8167 References to input, output, and goto operands in the assembler template
8168 of extended @code{asm} statements can use
8169 modifiers to affect the way the operands are formatted in
8170 the code output to the assembler. For example, the
8171 following code uses the @samp{h} and @samp{b} modifiers for x86:
8175 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8179 These modifiers generate this assembler code:
8185 The rest of this discussion uses the following code for illustrative purposes.
8194 asm volatile goto ("some assembler instructions here"
8196 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8197 : /* No clobbers. */
8202 With no modifiers, this is what the output from the operands would be for the
8203 @samp{att} and @samp{intel} dialects of assembler:
8205 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8206 @headitem Operand @tab masm=att @tab masm=intel
8215 @tab @code{OFFSET FLAT:.L2}
8218 The table below shows the list of supported modifiers and their effects.
8220 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8221 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8223 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8228 @tab Print the QImode name of the register.
8233 @tab Print the QImode name for a ``high'' register.
8238 @tab Print the HImode name of the register.
8243 @tab Print the SImode name of the register.
8248 @tab Print the DImode name of the register.
8253 @tab Print the label name with no punctuation.
8258 @tab Require a constant operand and print the constant expression with no punctuation.
8264 @anchor{x86floatingpointasmoperands}
8265 @subsubsection x86 Floating-Point @code{asm} Operands
8267 On x86 targets, there are several rules on the usage of stack-like registers
8268 in the operands of an @code{asm}. These rules apply only to the operands
8269 that are stack-like registers:
8273 Given a set of input registers that die in an @code{asm}, it is
8274 necessary to know which are implicitly popped by the @code{asm}, and
8275 which must be explicitly popped by GCC@.
8277 An input register that is implicitly popped by the @code{asm} must be
8278 explicitly clobbered, unless it is constrained to match an
8282 For any input register that is implicitly popped by an @code{asm}, it is
8283 necessary to know how to adjust the stack to compensate for the pop.
8284 If any non-popped input is closer to the top of the reg-stack than
8285 the implicitly popped register, it would not be possible to know what the
8286 stack looked like---it's not clear how the rest of the stack ``slides
8289 All implicitly popped input registers must be closer to the top of
8290 the reg-stack than any input that is not implicitly popped.
8292 It is possible that if an input dies in an @code{asm}, the compiler might
8293 use the input register for an output reload. Consider this example:
8296 asm ("foo" : "=t" (a) : "f" (b));
8300 This code says that input @code{b} is not popped by the @code{asm}, and that
8301 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8302 deeper after the @code{asm} than it was before. But, it is possible that
8303 reload may think that it can use the same register for both the input and
8306 To prevent this from happening,
8307 if any input operand uses the @samp{f} constraint, all output register
8308 constraints must use the @samp{&} early-clobber modifier.
8310 The example above is correctly written as:
8313 asm ("foo" : "=&t" (a) : "f" (b));
8317 Some operands need to be in particular places on the stack. All
8318 output operands fall in this category---GCC has no other way to
8319 know which registers the outputs appear in unless you indicate
8320 this in the constraints.
8322 Output operands must specifically indicate which register an output
8323 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8324 constraints must select a class with a single register.
8327 Output operands may not be ``inserted'' between existing stack registers.
8328 Since no 387 opcode uses a read/write operand, all output operands
8329 are dead before the @code{asm}, and are pushed by the @code{asm}.
8330 It makes no sense to push anywhere but the top of the reg-stack.
8332 Output operands must start at the top of the reg-stack: output
8333 operands may not ``skip'' a register.
8336 Some @code{asm} statements may need extra stack space for internal
8337 calculations. This can be guaranteed by clobbering stack registers
8338 unrelated to the inputs and outputs.
8343 takes one input, which is internally popped, and produces two outputs.
8346 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8350 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8351 and replaces them with one output. The @code{st(1)} clobber is necessary
8352 for the compiler to know that @code{fyl2xp1} pops both inputs.
8355 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8363 @subsection Controlling Names Used in Assembler Code
8364 @cindex assembler names for identifiers
8365 @cindex names used in assembler code
8366 @cindex identifiers, names in assembler code
8368 You can specify the name to be used in the assembler code for a C
8369 function or variable by writing the @code{asm} (or @code{__asm__})
8370 keyword after the declarator as follows:
8373 int foo asm ("myfoo") = 2;
8377 This specifies that the name to be used for the variable @code{foo} in
8378 the assembler code should be @samp{myfoo} rather than the usual
8381 On systems where an underscore is normally prepended to the name of a C
8382 function or variable, this feature allows you to define names for the
8383 linker that do not start with an underscore.
8385 It does not make sense to use this feature with a non-static local
8386 variable since such variables do not have assembler names. If you are
8387 trying to put the variable in a particular register, see @ref{Explicit
8388 Reg Vars}. GCC presently accepts such code with a warning, but will
8389 probably be changed to issue an error, rather than a warning, in the
8392 You cannot use @code{asm} in this way in a function @emph{definition}; but
8393 you can get the same effect by writing a declaration for the function
8394 before its definition and putting @code{asm} there, like this:
8397 extern func () asm ("FUNC");
8404 It is up to you to make sure that the assembler names you choose do not
8405 conflict with any other assembler symbols. Also, you must not use a
8406 register name; that would produce completely invalid assembler code. GCC
8407 does not as yet have the ability to store static variables in registers.
8408 Perhaps that will be added.
8410 @node Explicit Reg Vars
8411 @subsection Variables in Specified Registers
8412 @cindex explicit register variables
8413 @cindex variables in specified registers
8414 @cindex specified registers
8415 @cindex registers, global allocation
8417 GNU C allows you to put a few global variables into specified hardware
8418 registers. You can also specify the register in which an ordinary
8419 register variable should be allocated.
8423 Global register variables reserve registers throughout the program.
8424 This may be useful in programs such as programming language
8425 interpreters that have a couple of global variables that are accessed
8429 Local register variables in specific registers do not reserve the
8430 registers, except at the point where they are used as input or output
8431 operands in an @code{asm} statement and the @code{asm} statement itself is
8432 not deleted. The compiler's data flow analysis is capable of determining
8433 where the specified registers contain live values, and where they are
8434 available for other uses. Stores into local register variables may be deleted
8435 when they appear to be dead according to dataflow analysis. References
8436 to local register variables may be deleted or moved or simplified.
8438 These local variables are sometimes convenient for use with the extended
8439 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
8440 output of the assembler instruction directly into a particular register.
8441 (This works provided the register you specify fits the constraints
8442 specified for that operand in the @code{asm}.)
8450 @node Global Reg Vars
8451 @subsubsection Defining Global Register Variables
8452 @cindex global register variables
8453 @cindex registers, global variables in
8455 You can define a global register variable in GNU C like this:
8458 register int *foo asm ("a5");
8462 Here @code{a5} is the name of the register that should be used. Choose a
8463 register that is normally saved and restored by function calls on your
8464 machine, so that library routines will not clobber it.
8466 Naturally the register name is CPU-dependent, so you need to
8467 conditionalize your program according to CPU type. The register
8468 @code{a5} is a good choice on a 68000 for a variable of pointer
8469 type. On machines with register windows, be sure to choose a ``global''
8470 register that is not affected magically by the function call mechanism.
8472 In addition, different operating systems on the same CPU may differ in how they
8473 name the registers; then you need additional conditionals. For
8474 example, some 68000 operating systems call this register @code{%a5}.
8476 Eventually there may be a way of asking the compiler to choose a register
8477 automatically, but first we need to figure out how it should choose and
8478 how to enable you to guide the choice. No solution is evident.
8480 Defining a global register variable in a certain register reserves that
8481 register entirely for this use, at least within the current compilation.
8482 The register is not allocated for any other purpose in the functions
8483 in the current compilation, and is not saved and restored by
8484 these functions. Stores into this register are never deleted even if they
8485 appear to be dead, but references may be deleted or moved or
8488 It is not safe to access the global register variables from signal
8489 handlers, or from more than one thread of control, because the system
8490 library routines may temporarily use the register for other things (unless
8491 you recompile them specially for the task at hand).
8493 @cindex @code{qsort}, and global register variables
8494 It is not safe for one function that uses a global register variable to
8495 call another such function @code{foo} by way of a third function
8496 @code{lose} that is compiled without knowledge of this variable (i.e.@: in a
8497 different source file in which the variable isn't declared). This is
8498 because @code{lose} might save the register and put some other value there.
8499 For example, you can't expect a global register variable to be available in
8500 the comparison-function that you pass to @code{qsort}, since @code{qsort}
8501 might have put something else in that register. (If you are prepared to
8502 recompile @code{qsort} with the same global register variable, you can
8503 solve this problem.)
8505 If you want to recompile @code{qsort} or other source files that do not
8506 actually use your global register variable, so that they do not use that
8507 register for any other purpose, then it suffices to specify the compiler
8508 option @option{-ffixed-@var{reg}}. You need not actually add a global
8509 register declaration to their source code.
8511 A function that can alter the value of a global register variable cannot
8512 safely be called from a function compiled without this variable, because it
8513 could clobber the value the caller expects to find there on return.
8514 Therefore, the function that is the entry point into the part of the
8515 program that uses the global register variable must explicitly save and
8516 restore the value that belongs to its caller.
8518 @cindex register variable after @code{longjmp}
8519 @cindex global register after @code{longjmp}
8520 @cindex value after @code{longjmp}
8523 On most machines, @code{longjmp} restores to each global register
8524 variable the value it had at the time of the @code{setjmp}. On some
8525 machines, however, @code{longjmp} does not change the value of global
8526 register variables. To be portable, the function that called @code{setjmp}
8527 should make other arrangements to save the values of the global register
8528 variables, and to restore them in a @code{longjmp}. This way, the same
8529 thing happens regardless of what @code{longjmp} does.
8531 All global register variable declarations must precede all function
8532 definitions. If such a declaration could appear after function
8533 definitions, the declaration would be too late to prevent the register from
8534 being used for other purposes in the preceding functions.
8536 Global register variables may not have initial values, because an
8537 executable file has no means to supply initial contents for a register.
8539 On the SPARC, there are reports that g3 @dots{} g7 are suitable
8540 registers, but certain library functions, such as @code{getwd}, as well
8541 as the subroutines for division and remainder, modify g3 and g4. g1 and
8542 g2 are local temporaries.
8544 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
8545 Of course, it does not do to use more than a few of those.
8547 @node Local Reg Vars
8548 @subsubsection Specifying Registers for Local Variables
8549 @cindex local variables, specifying registers
8550 @cindex specifying registers for local variables
8551 @cindex registers for local variables
8553 You can define a local register variable with a specified register
8557 register int *foo asm ("a5");
8561 Here @code{a5} is the name of the register that should be used. Note
8562 that this is the same syntax used for defining global register
8563 variables, but for a local variable it appears within a function.
8565 Naturally the register name is CPU-dependent, but this is not a
8566 problem, since specific registers are most often useful with explicit
8567 assembler instructions (@pxref{Extended Asm}). Both of these things
8568 generally require that you conditionalize your program according to
8571 In addition, operating systems on one type of CPU may differ in how they
8572 name the registers; then you need additional conditionals. For
8573 example, some 68000 operating systems call this register @code{%a5}.
8575 Defining such a register variable does not reserve the register; it
8576 remains available for other uses in places where flow control determines
8577 the variable's value is not live.
8579 This option does not guarantee that GCC generates code that has
8580 this variable in the register you specify at all times. You may not
8581 code an explicit reference to this register in the assembler
8582 instruction template part of an @code{asm} statement and assume it
8583 always refers to this variable.
8584 However, using the variable as an input or output operand to the @code{asm}
8585 guarantees that the specified register is used for that operand.
8586 @xref{Extended Asm}, for more information.
8588 Stores into local register variables may be deleted when they appear to be dead
8589 according to dataflow analysis. References to local register variables may
8590 be deleted or moved or simplified.
8592 As with global register variables, it is recommended that you choose a
8593 register that is normally saved and restored by function calls on
8594 your machine, so that library routines will not clobber it.
8596 Sometimes when writing inline @code{asm} code, you need to make an operand be a
8597 specific register, but there's no matching constraint letter for that
8598 register. To force the operand into that register, create a local variable
8599 and specify the register in the variable's declaration. Then use the local
8600 variable for the asm operand and specify any constraint letter that matches
8604 register int *p1 asm ("r0") = @dots{};
8605 register int *p2 asm ("r1") = @dots{};
8606 register int *result asm ("r0");
8607 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8610 @emph{Warning:} In the above example, be aware that a register (for example r0) can be
8611 call-clobbered by subsequent code, including function calls and library calls
8612 for arithmetic operators on other variables (for example the initialization
8613 of p2). In this case, use temporary variables for expressions between the
8614 register assignments:
8618 register int *p1 asm ("r0") = @dots{};
8619 register int *p2 asm ("r1") = t1;
8620 register int *result asm ("r0");
8621 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8624 @node Size of an asm
8625 @subsection Size of an @code{asm}
8627 Some targets require that GCC track the size of each instruction used
8628 in order to generate correct code. Because the final length of the
8629 code produced by an @code{asm} statement is only known by the
8630 assembler, GCC must make an estimate as to how big it will be. It
8631 does this by counting the number of instructions in the pattern of the
8632 @code{asm} and multiplying that by the length of the longest
8633 instruction supported by that processor. (When working out the number
8634 of instructions, it assumes that any occurrence of a newline or of
8635 whatever statement separator character is supported by the assembler --
8636 typically @samp{;} --- indicates the end of an instruction.)
8638 Normally, GCC's estimate is adequate to ensure that correct
8639 code is generated, but it is possible to confuse the compiler if you use
8640 pseudo instructions or assembler macros that expand into multiple real
8641 instructions, or if you use assembler directives that expand to more
8642 space in the object file than is needed for a single instruction.
8643 If this happens then the assembler may produce a diagnostic saying that
8644 a label is unreachable.
8646 @node Alternate Keywords
8647 @section Alternate Keywords
8648 @cindex alternate keywords
8649 @cindex keywords, alternate
8651 @option{-ansi} and the various @option{-std} options disable certain
8652 keywords. This causes trouble when you want to use GNU C extensions, or
8653 a general-purpose header file that should be usable by all programs,
8654 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8655 @code{inline} are not available in programs compiled with
8656 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8657 program compiled with @option{-std=c99} or @option{-std=c11}). The
8659 @code{restrict} is only available when @option{-std=gnu99} (which will
8660 eventually be the default) or @option{-std=c99} (or the equivalent
8661 @option{-std=iso9899:1999}), or an option for a later standard
8664 The way to solve these problems is to put @samp{__} at the beginning and
8665 end of each problematical keyword. For example, use @code{__asm__}
8666 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8668 Other C compilers won't accept these alternative keywords; if you want to
8669 compile with another compiler, you can define the alternate keywords as
8670 macros to replace them with the customary keywords. It looks like this:
8678 @findex __extension__
8680 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8682 prevent such warnings within one expression by writing
8683 @code{__extension__} before the expression. @code{__extension__} has no
8684 effect aside from this.
8686 @node Incomplete Enums
8687 @section Incomplete @code{enum} Types
8689 You can define an @code{enum} tag without specifying its possible values.
8690 This results in an incomplete type, much like what you get if you write
8691 @code{struct foo} without describing the elements. A later declaration
8692 that does specify the possible values completes the type.
8694 You can't allocate variables or storage using the type while it is
8695 incomplete. However, you can work with pointers to that type.
8697 This extension may not be very useful, but it makes the handling of
8698 @code{enum} more consistent with the way @code{struct} and @code{union}
8701 This extension is not supported by GNU C++.
8703 @node Function Names
8704 @section Function Names as Strings
8705 @cindex @code{__func__} identifier
8706 @cindex @code{__FUNCTION__} identifier
8707 @cindex @code{__PRETTY_FUNCTION__} identifier
8709 GCC provides three magic variables that hold the name of the current
8710 function, as a string. The first of these is @code{__func__}, which
8711 is part of the C99 standard:
8713 The identifier @code{__func__} is implicitly declared by the translator
8714 as if, immediately following the opening brace of each function
8715 definition, the declaration
8718 static const char __func__[] = "function-name";
8722 appeared, where function-name is the name of the lexically-enclosing
8723 function. This name is the unadorned name of the function.
8725 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8726 backward compatibility with old versions of GCC.
8728 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8729 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8730 the type signature of the function as well as its bare name. For
8731 example, this program:
8735 extern int printf (char *, ...);
8742 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8743 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8761 __PRETTY_FUNCTION__ = void a::sub(int)
8764 These identifiers are variables, not preprocessor macros, and may not
8765 be used to initialize @code{char} arrays or be concatenated with other string
8768 @node Return Address
8769 @section Getting the Return or Frame Address of a Function
8771 These functions may be used to get information about the callers of a
8774 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8775 This function returns the return address of the current function, or of
8776 one of its callers. The @var{level} argument is number of frames to
8777 scan up the call stack. A value of @code{0} yields the return address
8778 of the current function, a value of @code{1} yields the return address
8779 of the caller of the current function, and so forth. When inlining
8780 the expected behavior is that the function returns the address of
8781 the function that is returned to. To work around this behavior use
8782 the @code{noinline} function attribute.
8784 The @var{level} argument must be a constant integer.
8786 On some machines it may be impossible to determine the return address of
8787 any function other than the current one; in such cases, or when the top
8788 of the stack has been reached, this function returns @code{0} or a
8789 random value. In addition, @code{__builtin_frame_address} may be used
8790 to determine if the top of the stack has been reached.
8792 Additional post-processing of the returned value may be needed, see
8793 @code{__builtin_extract_return_addr}.
8795 Calling this function with a nonzero argument can have unpredictable
8796 effects, including crashing the calling program. As a result, calls
8797 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8798 option is in effect. Such calls should only be made in debugging
8802 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
8803 The address as returned by @code{__builtin_return_address} may have to be fed
8804 through this function to get the actual encoded address. For example, on the
8805 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
8806 platforms an offset has to be added for the true next instruction to be
8809 If no fixup is needed, this function simply passes through @var{addr}.
8812 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
8813 This function does the reverse of @code{__builtin_extract_return_addr}.
8816 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
8817 This function is similar to @code{__builtin_return_address}, but it
8818 returns the address of the function frame rather than the return address
8819 of the function. Calling @code{__builtin_frame_address} with a value of
8820 @code{0} yields the frame address of the current function, a value of
8821 @code{1} yields the frame address of the caller of the current function,
8824 The frame is the area on the stack that holds local variables and saved
8825 registers. The frame address is normally the address of the first word
8826 pushed on to the stack by the function. However, the exact definition
8827 depends upon the processor and the calling convention. If the processor
8828 has a dedicated frame pointer register, and the function has a frame,
8829 then @code{__builtin_frame_address} returns the value of the frame
8832 On some machines it may be impossible to determine the frame address of
8833 any function other than the current one; in such cases, or when the top
8834 of the stack has been reached, this function returns @code{0} if
8835 the first frame pointer is properly initialized by the startup code.
8837 Calling this function with a nonzero argument can have unpredictable
8838 effects, including crashing the calling program. As a result, calls
8839 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8840 option is in effect. Such calls should only be made in debugging
8844 @node Vector Extensions
8845 @section Using Vector Instructions through Built-in Functions
8847 On some targets, the instruction set contains SIMD vector instructions which
8848 operate on multiple values contained in one large register at the same time.
8849 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
8852 The first step in using these extensions is to provide the necessary data
8853 types. This should be done using an appropriate @code{typedef}:
8856 typedef int v4si __attribute__ ((vector_size (16)));
8860 The @code{int} type specifies the base type, while the attribute specifies
8861 the vector size for the variable, measured in bytes. For example, the
8862 declaration above causes the compiler to set the mode for the @code{v4si}
8863 type to be 16 bytes wide and divided into @code{int} sized units. For
8864 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
8865 corresponding mode of @code{foo} is @acronym{V4SI}.
8867 The @code{vector_size} attribute is only applicable to integral and
8868 float scalars, although arrays, pointers, and function return values
8869 are allowed in conjunction with this construct. Only sizes that are
8870 a power of two are currently allowed.
8872 All the basic integer types can be used as base types, both as signed
8873 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
8874 @code{long long}. In addition, @code{float} and @code{double} can be
8875 used to build floating-point vector types.
8877 Specifying a combination that is not valid for the current architecture
8878 causes GCC to synthesize the instructions using a narrower mode.
8879 For example, if you specify a variable of type @code{V4SI} and your
8880 architecture does not allow for this specific SIMD type, GCC
8881 produces code that uses 4 @code{SIs}.
8883 The types defined in this manner can be used with a subset of normal C
8884 operations. Currently, GCC allows using the following operators
8885 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
8887 The operations behave like C++ @code{valarrays}. Addition is defined as
8888 the addition of the corresponding elements of the operands. For
8889 example, in the code below, each of the 4 elements in @var{a} is
8890 added to the corresponding 4 elements in @var{b} and the resulting
8891 vector is stored in @var{c}.
8894 typedef int v4si __attribute__ ((vector_size (16)));
8901 Subtraction, multiplication, division, and the logical operations
8902 operate in a similar manner. Likewise, the result of using the unary
8903 minus or complement operators on a vector type is a vector whose
8904 elements are the negative or complemented values of the corresponding
8905 elements in the operand.
8907 It is possible to use shifting operators @code{<<}, @code{>>} on
8908 integer-type vectors. The operation is defined as following: @code{@{a0,
8909 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
8910 @dots{}, an >> bn@}}@. Vector operands must have the same number of
8913 For convenience, it is allowed to use a binary vector operation
8914 where one operand is a scalar. In that case the compiler transforms
8915 the scalar operand into a vector where each element is the scalar from
8916 the operation. The transformation happens only if the scalar could be
8917 safely converted to the vector-element type.
8918 Consider the following code.
8921 typedef int v4si __attribute__ ((vector_size (16)));
8926 a = b + 1; /* a = b + @{1,1,1,1@}; */
8927 a = 2 * b; /* a = @{2,2,2,2@} * b; */
8929 a = l + a; /* Error, cannot convert long to int. */
8932 Vectors can be subscripted as if the vector were an array with
8933 the same number of elements and base type. Out of bound accesses
8934 invoke undefined behavior at run time. Warnings for out of bound
8935 accesses for vector subscription can be enabled with
8936 @option{-Warray-bounds}.
8938 Vector comparison is supported with standard comparison
8939 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
8940 vector expressions of integer-type or real-type. Comparison between
8941 integer-type vectors and real-type vectors are not supported. The
8942 result of the comparison is a vector of the same width and number of
8943 elements as the comparison operands with a signed integral element
8946 Vectors are compared element-wise producing 0 when comparison is false
8947 and -1 (constant of the appropriate type where all bits are set)
8948 otherwise. Consider the following example.
8951 typedef int v4si __attribute__ ((vector_size (16)));
8953 v4si a = @{1,2,3,4@};
8954 v4si b = @{3,2,1,4@};
8957 c = a > b; /* The result would be @{0, 0,-1, 0@} */
8958 c = a == b; /* The result would be @{0,-1, 0,-1@} */
8961 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
8962 @code{b} and @code{c} are vectors of the same type and @code{a} is an
8963 integer vector with the same number of elements of the same size as @code{b}
8964 and @code{c}, computes all three arguments and creates a vector
8965 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
8966 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
8967 As in the case of binary operations, this syntax is also accepted when
8968 one of @code{b} or @code{c} is a scalar that is then transformed into a
8969 vector. If both @code{b} and @code{c} are scalars and the type of
8970 @code{true?b:c} has the same size as the element type of @code{a}, then
8971 @code{b} and @code{c} are converted to a vector type whose elements have
8972 this type and with the same number of elements as @code{a}.
8974 In C++, the logic operators @code{!, &&, ||} are available for vectors.
8975 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
8976 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
8977 For mixed operations between a scalar @code{s} and a vector @code{v},
8978 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
8979 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
8981 Vector shuffling is available using functions
8982 @code{__builtin_shuffle (vec, mask)} and
8983 @code{__builtin_shuffle (vec0, vec1, mask)}.
8984 Both functions construct a permutation of elements from one or two
8985 vectors and return a vector of the same type as the input vector(s).
8986 The @var{mask} is an integral vector with the same width (@var{W})
8987 and element count (@var{N}) as the output vector.
8989 The elements of the input vectors are numbered in memory ordering of
8990 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
8991 elements of @var{mask} are considered modulo @var{N} in the single-operand
8992 case and modulo @math{2*@var{N}} in the two-operand case.
8994 Consider the following example,
8997 typedef int v4si __attribute__ ((vector_size (16)));
8999 v4si a = @{1,2,3,4@};
9000 v4si b = @{5,6,7,8@};
9001 v4si mask1 = @{0,1,1,3@};
9002 v4si mask2 = @{0,4,2,5@};
9005 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9006 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9009 Note that @code{__builtin_shuffle} is intentionally semantically
9010 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9012 You can declare variables and use them in function calls and returns, as
9013 well as in assignments and some casts. You can specify a vector type as
9014 a return type for a function. Vector types can also be used as function
9015 arguments. It is possible to cast from one vector type to another,
9016 provided they are of the same size (in fact, you can also cast vectors
9017 to and from other datatypes of the same size).
9019 You cannot operate between vectors of different lengths or different
9020 signedness without a cast.
9023 @section Support for @code{offsetof}
9024 @findex __builtin_offsetof
9026 GCC implements for both C and C++ a syntactic extension to implement
9027 the @code{offsetof} macro.
9031 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9033 offsetof_member_designator:
9035 | offsetof_member_designator "." @code{identifier}
9036 | offsetof_member_designator "[" @code{expr} "]"
9039 This extension is sufficient such that
9042 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9046 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9047 may be dependent. In either case, @var{member} may consist of a single
9048 identifier, or a sequence of member accesses and array references.
9050 @node __sync Builtins
9051 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9053 The following built-in functions
9054 are intended to be compatible with those described
9055 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9056 section 7.4. As such, they depart from normal GCC practice by not using
9057 the @samp{__builtin_} prefix and also by being overloaded so that they
9058 work on multiple types.
9060 The definition given in the Intel documentation allows only for the use of
9061 the types @code{int}, @code{long}, @code{long long} or their unsigned
9062 counterparts. GCC allows any integral scalar or pointer type that is
9063 1, 2, 4 or 8 bytes in length.
9065 These functions are implemented in terms of the @samp{__atomic}
9066 builtins (@pxref{__atomic Builtins}). They should not be used for new
9067 code which should use the @samp{__atomic} builtins instead.
9069 Not all operations are supported by all target processors. If a particular
9070 operation cannot be implemented on the target processor, a warning is
9071 generated and a call to an external function is generated. The external
9072 function carries the same name as the built-in version,
9073 with an additional suffix
9074 @samp{_@var{n}} where @var{n} is the size of the data type.
9076 @c ??? Should we have a mechanism to suppress this warning? This is almost
9077 @c useful for implementing the operation under the control of an external
9080 In most cases, these built-in functions are considered a @dfn{full barrier}.
9082 no memory operand is moved across the operation, either forward or
9083 backward. Further, instructions are issued as necessary to prevent the
9084 processor from speculating loads across the operation and from queuing stores
9085 after the operation.
9087 All of the routines are described in the Intel documentation to take
9088 ``an optional list of variables protected by the memory barrier''. It's
9089 not clear what is meant by that; it could mean that @emph{only} the
9090 listed variables are protected, or it could mean a list of additional
9091 variables to be protected. The list is ignored by GCC which treats it as
9092 empty. GCC interprets an empty list as meaning that all globally
9093 accessible variables should be protected.
9096 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9097 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9098 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9099 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9100 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9101 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9102 @findex __sync_fetch_and_add
9103 @findex __sync_fetch_and_sub
9104 @findex __sync_fetch_and_or
9105 @findex __sync_fetch_and_and
9106 @findex __sync_fetch_and_xor
9107 @findex __sync_fetch_and_nand
9108 These built-in functions perform the operation suggested by the name, and
9109 returns the value that had previously been in memory. That is,
9112 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9113 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9116 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9117 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9119 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9120 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9121 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9122 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9123 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9124 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9125 @findex __sync_add_and_fetch
9126 @findex __sync_sub_and_fetch
9127 @findex __sync_or_and_fetch
9128 @findex __sync_and_and_fetch
9129 @findex __sync_xor_and_fetch
9130 @findex __sync_nand_and_fetch
9131 These built-in functions perform the operation suggested by the name, and
9132 return the new value. That is,
9135 @{ *ptr @var{op}= value; return *ptr; @}
9136 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9139 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9140 as @code{*ptr = ~(*ptr & value)} instead of
9141 @code{*ptr = ~*ptr & value}.
9143 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9144 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9145 @findex __sync_bool_compare_and_swap
9146 @findex __sync_val_compare_and_swap
9147 These built-in functions perform an atomic compare and swap.
9148 That is, if the current
9149 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9152 The ``bool'' version returns true if the comparison is successful and
9153 @var{newval} is written. The ``val'' version returns the contents
9154 of @code{*@var{ptr}} before the operation.
9156 @item __sync_synchronize (...)
9157 @findex __sync_synchronize
9158 This built-in function issues a full memory barrier.
9160 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9161 @findex __sync_lock_test_and_set
9162 This built-in function, as described by Intel, is not a traditional test-and-set
9163 operation, but rather an atomic exchange operation. It writes @var{value}
9164 into @code{*@var{ptr}}, and returns the previous contents of
9167 Many targets have only minimal support for such locks, and do not support
9168 a full exchange operation. In this case, a target may support reduced
9169 functionality here by which the @emph{only} valid value to store is the
9170 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9171 is implementation defined.
9173 This built-in function is not a full barrier,
9174 but rather an @dfn{acquire barrier}.
9175 This means that references after the operation cannot move to (or be
9176 speculated to) before the operation, but previous memory stores may not
9177 be globally visible yet, and previous memory loads may not yet be
9180 @item void __sync_lock_release (@var{type} *ptr, ...)
9181 @findex __sync_lock_release
9182 This built-in function releases the lock acquired by
9183 @code{__sync_lock_test_and_set}.
9184 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9186 This built-in function is not a full barrier,
9187 but rather a @dfn{release barrier}.
9188 This means that all previous memory stores are globally visible, and all
9189 previous memory loads have been satisfied, but following memory reads
9190 are not prevented from being speculated to before the barrier.
9193 @node __atomic Builtins
9194 @section Built-in Functions for Memory Model Aware Atomic Operations
9196 The following built-in functions approximately match the requirements
9197 for the C++11 memory model. They are all
9198 identified by being prefixed with @samp{__atomic} and most are
9199 overloaded so that they work with multiple types.
9201 These functions are intended to replace the legacy @samp{__sync}
9202 builtins. The main difference is that the memory order that is requested
9203 is a parameter to the functions. New code should always use the
9204 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9206 Note that the @samp{__atomic} builtins assume that programs will
9207 conform to the C++11 memory model. In particular, they assume
9208 that programs are free of data races. See the C++11 standard for
9209 detailed requirements.
9211 The @samp{__atomic} builtins can be used with any integral scalar or
9212 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9213 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9214 supported by the architecture.
9216 The four non-arithmetic functions (load, store, exchange, and
9217 compare_exchange) all have a generic version as well. This generic
9218 version works on any data type. It uses the lock-free built-in function
9219 if the specific data type size makes that possible; otherwise, an
9220 external call is left to be resolved at run time. This external call is
9221 the same format with the addition of a @samp{size_t} parameter inserted
9222 as the first parameter indicating the size of the object being pointed to.
9223 All objects must be the same size.
9225 There are 6 different memory orders that can be specified. These map
9226 to the C++11 memory orders with the same names, see the C++11 standard
9227 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9228 on atomic synchronization} for detailed definitions. Individual
9229 targets may also support additional memory orders for use on specific
9230 architectures. Refer to the target documentation for details of
9233 An atomic operation can both constrain code motion and
9234 be mapped to hardware instructions for synchronization between threads
9235 (e.g., a fence). To which extent this happens is controlled by the
9236 memory orders, which are listed here in approximately ascending order of
9237 strength. The description of each memory order is only meant to roughly
9238 illustrate the effects and is not a specification; see the C++11
9239 memory model for precise semantics.
9242 @item __ATOMIC_RELAXED
9243 Implies no inter-thread ordering constraints.
9244 @item __ATOMIC_CONSUME
9245 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9246 memory order because of a deficiency in C++11's semantics for
9247 @code{memory_order_consume}.
9248 @item __ATOMIC_ACQUIRE
9249 Creates an inter-thread happens-before constraint from the release (or
9250 stronger) semantic store to this acquire load. Can prevent hoisting
9251 of code to before the operation.
9252 @item __ATOMIC_RELEASE
9253 Creates an inter-thread happens-before constraint to acquire (or stronger)
9254 semantic loads that read from this release store. Can prevent sinking
9255 of code to after the operation.
9256 @item __ATOMIC_ACQ_REL
9257 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9258 @code{__ATOMIC_RELEASE}.
9259 @item __ATOMIC_SEQ_CST
9260 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9263 Note that in the C++11 memory model, @emph{fences} (e.g.,
9264 @samp{__atomic_thread_fence}) take effect in combination with other
9265 atomic operations on specific memory locations (e.g., atomic loads);
9266 operations on specific memory locations do not necessarily affect other
9267 operations in the same way.
9269 Target architectures are encouraged to provide their own patterns for
9270 each of the atomic built-in functions. If no target is provided, the original
9271 non-memory model set of @samp{__sync} atomic built-in functions are
9272 used, along with any required synchronization fences surrounding it in
9273 order to achieve the proper behavior. Execution in this case is subject
9274 to the same restrictions as those built-in functions.
9276 If there is no pattern or mechanism to provide a lock-free instruction
9277 sequence, a call is made to an external routine with the same parameters
9278 to be resolved at run time.
9280 When implementing patterns for these built-in functions, the memory order
9281 parameter can be ignored as long as the pattern implements the most
9282 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9283 orders execute correctly with this memory order but they may not execute as
9284 efficiently as they could with a more appropriate implementation of the
9285 relaxed requirements.
9287 Note that the C++11 standard allows for the memory order parameter to be
9288 determined at run time rather than at compile time. These built-in
9289 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9290 than invoke a runtime library call or inline a switch statement. This is
9291 standard compliant, safe, and the simplest approach for now.
9293 The memory order parameter is a signed int, but only the lower 16 bits are
9294 reserved for the memory order. The remainder of the signed int is reserved
9295 for target use and should be 0. Use of the predefined atomic values
9296 ensures proper usage.
9298 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9299 This built-in function implements an atomic load operation. It returns the
9300 contents of @code{*@var{ptr}}.
9302 The valid memory order variants are
9303 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9304 and @code{__ATOMIC_CONSUME}.
9308 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9309 This is the generic version of an atomic load. It returns the
9310 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9314 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9315 This built-in function implements an atomic store operation. It writes
9316 @code{@var{val}} into @code{*@var{ptr}}.
9318 The valid memory order variants are
9319 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9323 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9324 This is the generic version of an atomic store. It stores the value
9325 of @code{*@var{val}} into @code{*@var{ptr}}.
9329 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9330 This built-in function implements an atomic exchange operation. It writes
9331 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9334 The valid memory order variants are
9335 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9336 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9340 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9341 This is the generic version of an atomic exchange. It stores the
9342 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9343 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9347 @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)
9348 This built-in function implements an atomic compare and exchange operation.
9349 This compares the contents of @code{*@var{ptr}} with the contents of
9350 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9351 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9352 equal, the operation is a @emph{read} and the current contents of
9353 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
9354 for weak compare_exchange, and false for the strong variation. Many targets
9355 only offer the strong variation and ignore the parameter. When in doubt, use
9356 the strong variation.
9358 True is returned if @var{desired} is written into
9359 @code{*@var{ptr}} and the operation is considered to conform to the
9360 memory order specified by @var{success_memorder}. There are no
9361 restrictions on what memory order can be used here.
9363 False is returned otherwise, and the operation is considered to conform
9364 to @var{failure_memorder}. This memory order cannot be
9365 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9366 stronger order than that specified by @var{success_memorder}.
9370 @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)
9371 This built-in function implements the generic version of
9372 @code{__atomic_compare_exchange}. The function is virtually identical to
9373 @code{__atomic_compare_exchange_n}, except the desired value is also a
9378 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9379 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9380 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9381 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9382 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9383 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9384 These built-in functions perform the operation suggested by the name, and
9385 return the result of the operation. That is,
9388 @{ *ptr @var{op}= val; return *ptr; @}
9391 All memory orders are valid.
9395 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9396 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9397 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9398 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9399 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9400 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9401 These built-in functions perform the operation suggested by the name, and
9402 return the value that had previously been in @code{*@var{ptr}}. That is,
9405 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9408 All memory orders are valid.
9412 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9414 This built-in function performs an atomic test-and-set operation on
9415 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9416 defined nonzero ``set'' value and the return value is @code{true} if and only
9417 if the previous contents were ``set''.
9418 It should be only used for operands of type @code{bool} or @code{char}. For
9419 other types only part of the value may be set.
9421 All memory orders are valid.
9425 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9427 This built-in function performs an atomic clear operation on
9428 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9429 It should be only used for operands of type @code{bool} or @code{char} and
9430 in conjunction with @code{__atomic_test_and_set}.
9431 For other types it may only clear partially. If the type is not @code{bool}
9432 prefer using @code{__atomic_store}.
9434 The valid memory order variants are
9435 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9436 @code{__ATOMIC_RELEASE}.
9440 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9442 This built-in function acts as a synchronization fence between threads
9443 based on the specified memory order.
9445 All memory orders are valid.
9449 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9451 This built-in function acts as a synchronization fence between a thread
9452 and signal handlers based in the same thread.
9454 All memory orders are valid.
9458 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9460 This built-in function returns true if objects of @var{size} bytes always
9461 generate lock-free atomic instructions for the target architecture.
9462 @var{size} must resolve to a compile-time constant and the result also
9463 resolves to a compile-time constant.
9465 @var{ptr} is an optional pointer to the object that may be used to determine
9466 alignment. A value of 0 indicates typical alignment should be used. The
9467 compiler may also ignore this parameter.
9470 if (_atomic_always_lock_free (sizeof (long long), 0))
9475 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9477 This built-in function returns true if objects of @var{size} bytes always
9478 generate lock-free atomic instructions for the target architecture. If
9479 the built-in function is not known to be lock-free, a call is made to a
9480 runtime routine named @code{__atomic_is_lock_free}.
9482 @var{ptr} is an optional pointer to the object that may be used to determine
9483 alignment. A value of 0 indicates typical alignment should be used. The
9484 compiler may also ignore this parameter.
9487 @node Integer Overflow Builtins
9488 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9490 The following built-in functions allow performing simple arithmetic operations
9491 together with checking whether the operations overflowed.
9493 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9494 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9495 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9496 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9497 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9498 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9499 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9501 These built-in functions promote the first two operands into infinite precision signed
9502 type and perform addition on those promoted operands. The result is then
9503 cast to the type the third pointer argument points to and stored there.
9504 If the stored result is equal to the infinite precision result, the built-in
9505 functions return false, otherwise they return true. As the addition is
9506 performed in infinite signed precision, these built-in functions have fully defined
9507 behavior for all argument values.
9509 The first built-in function allows arbitrary integral types for operands and
9510 the result type must be pointer to some integer type, the rest of the built-in
9511 functions have explicit integer types.
9513 The compiler will attempt to use hardware instructions to implement
9514 these built-in functions where possible, like conditional jump on overflow
9515 after addition, conditional jump on carry etc.
9519 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9520 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9521 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9522 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9523 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9524 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9525 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9527 These built-in functions are similar to the add overflow checking built-in
9528 functions above, except they perform subtraction, subtract the second argument
9529 from the first one, instead of addition.
9533 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9534 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9535 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9536 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9537 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9538 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9539 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9541 These built-in functions are similar to the add overflow checking built-in
9542 functions above, except they perform multiplication, instead of addition.
9546 @node x86 specific memory model extensions for transactional memory
9547 @section x86-Specific Memory Model Extensions for Transactional Memory
9549 The x86 architecture supports additional memory ordering flags
9550 to mark lock critical sections for hardware lock elision.
9551 These must be specified in addition to an existing memory order to
9555 @item __ATOMIC_HLE_ACQUIRE
9556 Start lock elision on a lock variable.
9557 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9558 @item __ATOMIC_HLE_RELEASE
9559 End lock elision on a lock variable.
9560 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9563 When a lock acquire fails, it is required for good performance to abort
9564 the transaction quickly. This can be done with a @code{_mm_pause}.
9567 #include <immintrin.h> // For _mm_pause
9571 /* Acquire lock with lock elision */
9572 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9573 _mm_pause(); /* Abort failed transaction */
9575 /* Free lock with lock elision */
9576 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9579 @node Object Size Checking
9580 @section Object Size Checking Built-in Functions
9581 @findex __builtin_object_size
9582 @findex __builtin___memcpy_chk
9583 @findex __builtin___mempcpy_chk
9584 @findex __builtin___memmove_chk
9585 @findex __builtin___memset_chk
9586 @findex __builtin___strcpy_chk
9587 @findex __builtin___stpcpy_chk
9588 @findex __builtin___strncpy_chk
9589 @findex __builtin___strcat_chk
9590 @findex __builtin___strncat_chk
9591 @findex __builtin___sprintf_chk
9592 @findex __builtin___snprintf_chk
9593 @findex __builtin___vsprintf_chk
9594 @findex __builtin___vsnprintf_chk
9595 @findex __builtin___printf_chk
9596 @findex __builtin___vprintf_chk
9597 @findex __builtin___fprintf_chk
9598 @findex __builtin___vfprintf_chk
9600 GCC implements a limited buffer overflow protection mechanism
9601 that can prevent some buffer overflow attacks.
9603 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9604 is a built-in construct that returns a constant number of bytes from
9605 @var{ptr} to the end of the object @var{ptr} pointer points to
9606 (if known at compile time). @code{__builtin_object_size} never evaluates
9607 its arguments for side-effects. If there are any side-effects in them, it
9608 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9609 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9610 point to and all of them are known at compile time, the returned number
9611 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9612 0 and minimum if nonzero. If it is not possible to determine which objects
9613 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9614 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9615 for @var{type} 2 or 3.
9617 @var{type} is an integer constant from 0 to 3. If the least significant
9618 bit is clear, objects are whole variables, if it is set, a closest
9619 surrounding subobject is considered the object a pointer points to.
9620 The second bit determines if maximum or minimum of remaining bytes
9624 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9625 char *p = &var.buf1[1], *q = &var.b;
9627 /* Here the object p points to is var. */
9628 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9629 /* The subobject p points to is var.buf1. */
9630 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9631 /* The object q points to is var. */
9632 assert (__builtin_object_size (q, 0)
9633 == (char *) (&var + 1) - (char *) &var.b);
9634 /* The subobject q points to is var.b. */
9635 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9639 There are built-in functions added for many common string operation
9640 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9641 built-in is provided. This built-in has an additional last argument,
9642 which is the number of bytes remaining in object the @var{dest}
9643 argument points to or @code{(size_t) -1} if the size is not known.
9645 The built-in functions are optimized into the normal string functions
9646 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9647 it is known at compile time that the destination object will not
9648 be overflown. If the compiler can determine at compile time the
9649 object will be always overflown, it issues a warning.
9651 The intended use can be e.g.@:
9655 #define bos0(dest) __builtin_object_size (dest, 0)
9656 #define memcpy(dest, src, n) \
9657 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9661 /* It is unknown what object p points to, so this is optimized
9662 into plain memcpy - no checking is possible. */
9663 memcpy (p, "abcde", n);
9664 /* Destination is known and length too. It is known at compile
9665 time there will be no overflow. */
9666 memcpy (&buf[5], "abcde", 5);
9667 /* Destination is known, but the length is not known at compile time.
9668 This will result in __memcpy_chk call that can check for overflow
9670 memcpy (&buf[5], "abcde", n);
9671 /* Destination is known and it is known at compile time there will
9672 be overflow. There will be a warning and __memcpy_chk call that
9673 will abort the program at run time. */
9674 memcpy (&buf[6], "abcde", 5);
9677 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9678 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9679 @code{strcat} and @code{strncat}.
9681 There are also checking built-in functions for formatted output functions.
9683 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9684 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9685 const char *fmt, ...);
9686 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9688 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9689 const char *fmt, va_list ap);
9692 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9693 etc.@: functions and can contain implementation specific flags on what
9694 additional security measures the checking function might take, such as
9695 handling @code{%n} differently.
9697 The @var{os} argument is the object size @var{s} points to, like in the
9698 other built-in functions. There is a small difference in the behavior
9699 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9700 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9701 the checking function is called with @var{os} argument set to
9704 In addition to this, there are checking built-in functions
9705 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9706 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9707 These have just one additional argument, @var{flag}, right before
9708 format string @var{fmt}. If the compiler is able to optimize them to
9709 @code{fputc} etc.@: functions, it does, otherwise the checking function
9710 is called and the @var{flag} argument passed to it.
9712 @node Pointer Bounds Checker builtins
9713 @section Pointer Bounds Checker Built-in Functions
9714 @cindex Pointer Bounds Checker builtins
9715 @findex __builtin___bnd_set_ptr_bounds
9716 @findex __builtin___bnd_narrow_ptr_bounds
9717 @findex __builtin___bnd_copy_ptr_bounds
9718 @findex __builtin___bnd_init_ptr_bounds
9719 @findex __builtin___bnd_null_ptr_bounds
9720 @findex __builtin___bnd_store_ptr_bounds
9721 @findex __builtin___bnd_chk_ptr_lbounds
9722 @findex __builtin___bnd_chk_ptr_ubounds
9723 @findex __builtin___bnd_chk_ptr_bounds
9724 @findex __builtin___bnd_get_ptr_lbound
9725 @findex __builtin___bnd_get_ptr_ubound
9727 GCC provides a set of built-in functions to control Pointer Bounds Checker
9728 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9729 even if you compile with Pointer Bounds Checker off
9730 (@option{-fno-check-pointer-bounds}).
9731 The behavior may differ in such case as documented below.
9733 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9735 This built-in function returns a new pointer with the value of @var{q}, and
9736 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9737 Bounds Checker off, the built-in function just returns the first argument.
9740 extern void *__wrap_malloc (size_t n)
9742 void *p = (void *)__real_malloc (n);
9743 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9744 return __builtin___bnd_set_ptr_bounds (p, n);
9750 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9752 This built-in function returns a new pointer with the value of @var{p}
9753 and associates it with the narrowed bounds formed by the intersection
9754 of bounds associated with @var{q} and the bounds
9755 [@var{p}, @var{p} + @var{size} - 1].
9756 With Pointer Bounds Checker off, the built-in function just returns the first
9760 void init_objects (object *objs, size_t size)
9763 /* Initialize objects one-by-one passing pointers with bounds of
9764 an object, not the full array of objects. */
9765 for (i = 0; i < size; i++)
9766 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
9773 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
9775 This built-in function returns a new pointer with the value of @var{q},
9776 and associates it with the bounds already associated with pointer @var{r}.
9777 With Pointer Bounds Checker off, the built-in function just returns the first
9781 /* Here is a way to get pointer to object's field but
9782 still with the full object's bounds. */
9783 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
9789 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
9791 This built-in function returns a new pointer with the value of @var{q}, and
9792 associates it with INIT (allowing full memory access) bounds. With Pointer
9793 Bounds Checker off, the built-in function just returns the first argument.
9797 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
9799 This built-in function returns a new pointer with the value of @var{q}, and
9800 associates it with NULL (allowing no memory access) bounds. With Pointer
9801 Bounds Checker off, the built-in function just returns the first argument.
9805 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
9807 This built-in function stores the bounds associated with pointer @var{ptr_val}
9808 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
9809 bounds from legacy code without touching the associated pointer's memory when
9810 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
9811 function call is ignored.
9815 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
9817 This built-in function checks if the pointer @var{q} is within the lower
9818 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
9819 function call is ignored.
9822 extern void *__wrap_memset (void *dst, int c, size_t len)
9826 __builtin___bnd_chk_ptr_lbounds (dst);
9827 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
9828 __real_memset (dst, c, len);
9836 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
9838 This built-in function checks if the pointer @var{q} is within the upper
9839 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
9840 function call is ignored.
9844 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
9846 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
9847 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
9848 off, the built-in function call is ignored.
9851 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
9855 __bnd_chk_ptr_bounds (dst, n);
9856 __bnd_chk_ptr_bounds (src, n);
9857 __real_memcpy (dst, src, n);
9865 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
9867 This built-in function returns the lower bound associated
9868 with the pointer @var{q}, as a pointer value.
9869 This is useful for debugging using @code{printf}.
9870 With Pointer Bounds Checker off, the built-in function returns 0.
9873 void *lb = __builtin___bnd_get_ptr_lbound (q);
9874 void *ub = __builtin___bnd_get_ptr_ubound (q);
9875 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
9880 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
9882 This built-in function returns the upper bound (which is a pointer) associated
9883 with the pointer @var{q}. With Pointer Bounds Checker off,
9884 the built-in function returns -1.
9888 @node Cilk Plus Builtins
9889 @section Cilk Plus C/C++ Language Extension Built-in Functions
9891 GCC provides support for the following built-in reduction functions if Cilk Plus
9892 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
9895 @item @code{__sec_implicit_index}
9896 @item @code{__sec_reduce}
9897 @item @code{__sec_reduce_add}
9898 @item @code{__sec_reduce_all_nonzero}
9899 @item @code{__sec_reduce_all_zero}
9900 @item @code{__sec_reduce_any_nonzero}
9901 @item @code{__sec_reduce_any_zero}
9902 @item @code{__sec_reduce_max}
9903 @item @code{__sec_reduce_min}
9904 @item @code{__sec_reduce_max_ind}
9905 @item @code{__sec_reduce_min_ind}
9906 @item @code{__sec_reduce_mul}
9907 @item @code{__sec_reduce_mutating}
9910 Further details and examples about these built-in functions are described
9911 in the Cilk Plus language manual which can be found at
9912 @uref{http://www.cilkplus.org}.
9914 @node Other Builtins
9915 @section Other Built-in Functions Provided by GCC
9916 @cindex built-in functions
9917 @findex __builtin_call_with_static_chain
9918 @findex __builtin_fpclassify
9919 @findex __builtin_isfinite
9920 @findex __builtin_isnormal
9921 @findex __builtin_isgreater
9922 @findex __builtin_isgreaterequal
9923 @findex __builtin_isinf_sign
9924 @findex __builtin_isless
9925 @findex __builtin_islessequal
9926 @findex __builtin_islessgreater
9927 @findex __builtin_isunordered
9928 @findex __builtin_powi
9929 @findex __builtin_powif
9930 @findex __builtin_powil
10088 @findex fprintf_unlocked
10090 @findex fputs_unlocked
10198 @findex nexttowardf
10199 @findex nexttowardl
10207 @findex printf_unlocked
10237 @findex signbitd128
10238 @findex significand
10239 @findex significandf
10240 @findex significandl
10268 @findex strncasecmp
10311 GCC provides a large number of built-in functions other than the ones
10312 mentioned above. Some of these are for internal use in the processing
10313 of exceptions or variable-length argument lists and are not
10314 documented here because they may change from time to time; we do not
10315 recommend general use of these functions.
10317 The remaining functions are provided for optimization purposes.
10319 With the exception of built-ins that have library equivalents such as
10320 the standard C library functions discussed below, or that expand to
10321 library calls, GCC built-in functions are always expanded inline and
10322 thus do not have corresponding entry points and their address cannot
10323 be obtained. Attempting to use them in an expression other than
10324 a function call results in a compile-time error.
10326 @opindex fno-builtin
10327 GCC includes built-in versions of many of the functions in the standard
10328 C library. These functions come in two forms: one whose names start with
10329 the @code{__builtin_} prefix, and the other without. Both forms have the
10330 same type (including prototype), the same address (when their address is
10331 taken), and the same meaning as the C library functions even if you specify
10332 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10333 functions are only optimized in certain cases; if they are not optimized in
10334 a particular case, a call to the library function is emitted.
10338 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10339 @option{-std=c99} or @option{-std=c11}), the functions
10340 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10341 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10342 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10343 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10344 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10345 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10346 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10347 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10348 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10349 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10350 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10351 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10352 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10353 @code{significandl}, @code{significand}, @code{sincosf},
10354 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10355 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10356 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10357 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10359 may be handled as built-in functions.
10360 All these functions have corresponding versions
10361 prefixed with @code{__builtin_}, which may be used even in strict C90
10364 The ISO C99 functions
10365 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10366 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10367 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10368 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10369 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10370 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10371 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10372 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10373 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10374 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10375 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10376 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10377 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10378 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10379 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10380 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10381 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10382 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10383 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10384 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10385 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10386 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10387 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10388 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10389 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10390 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10391 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10392 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10393 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10394 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10395 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10396 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10397 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10398 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10399 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10400 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10401 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10402 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10403 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10404 are handled as built-in functions
10405 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10407 There are also built-in versions of the ISO C99 functions
10408 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10409 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10410 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10411 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10412 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10413 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10414 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10415 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10416 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10417 that are recognized in any mode since ISO C90 reserves these names for
10418 the purpose to which ISO C99 puts them. All these functions have
10419 corresponding versions prefixed with @code{__builtin_}.
10421 The ISO C94 functions
10422 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10423 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10424 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10426 are handled as built-in functions
10427 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10429 The ISO C90 functions
10430 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10431 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10432 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10433 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10434 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10435 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10436 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10437 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10438 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10439 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10440 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10441 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10442 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10443 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10444 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10445 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10446 are all recognized as built-in functions unless
10447 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10448 is specified for an individual function). All of these functions have
10449 corresponding versions prefixed with @code{__builtin_}.
10451 GCC provides built-in versions of the ISO C99 floating-point comparison
10452 macros that avoid raising exceptions for unordered operands. They have
10453 the same names as the standard macros ( @code{isgreater},
10454 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10455 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10456 prefixed. We intend for a library implementor to be able to simply
10457 @code{#define} each standard macro to its built-in equivalent.
10458 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10459 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10460 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10461 built-in functions appear both with and without the @code{__builtin_} prefix.
10463 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10465 You can use the built-in function @code{__builtin_types_compatible_p} to
10466 determine whether two types are the same.
10468 This built-in function returns 1 if the unqualified versions of the
10469 types @var{type1} and @var{type2} (which are types, not expressions) are
10470 compatible, 0 otherwise. The result of this built-in function can be
10471 used in integer constant expressions.
10473 This built-in function ignores top level qualifiers (e.g., @code{const},
10474 @code{volatile}). For example, @code{int} is equivalent to @code{const
10477 The type @code{int[]} and @code{int[5]} are compatible. On the other
10478 hand, @code{int} and @code{char *} are not compatible, even if the size
10479 of their types, on the particular architecture are the same. Also, the
10480 amount of pointer indirection is taken into account when determining
10481 similarity. Consequently, @code{short *} is not similar to
10482 @code{short **}. Furthermore, two types that are typedefed are
10483 considered compatible if their underlying types are compatible.
10485 An @code{enum} type is not considered to be compatible with another
10486 @code{enum} type even if both are compatible with the same integer
10487 type; this is what the C standard specifies.
10488 For example, @code{enum @{foo, bar@}} is not similar to
10489 @code{enum @{hot, dog@}}.
10491 You typically use this function in code whose execution varies
10492 depending on the arguments' types. For example:
10497 typeof (x) tmp = (x); \
10498 if (__builtin_types_compatible_p (typeof (x), long double)) \
10499 tmp = foo_long_double (tmp); \
10500 else if (__builtin_types_compatible_p (typeof (x), double)) \
10501 tmp = foo_double (tmp); \
10502 else if (__builtin_types_compatible_p (typeof (x), float)) \
10503 tmp = foo_float (tmp); \
10510 @emph{Note:} This construct is only available for C@.
10514 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10516 The @var{call_exp} expression must be a function call, and the
10517 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10518 is passed to the function call in the target's static chain location.
10519 The result of builtin is the result of the function call.
10521 @emph{Note:} This builtin is only available for C@.
10522 This builtin can be used to call Go closures from C.
10526 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10528 You can use the built-in function @code{__builtin_choose_expr} to
10529 evaluate code depending on the value of a constant expression. This
10530 built-in function returns @var{exp1} if @var{const_exp}, which is an
10531 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10533 This built-in function is analogous to the @samp{? :} operator in C,
10534 except that the expression returned has its type unaltered by promotion
10535 rules. Also, the built-in function does not evaluate the expression
10536 that is not chosen. For example, if @var{const_exp} evaluates to true,
10537 @var{exp2} is not evaluated even if it has side-effects.
10539 This built-in function can return an lvalue if the chosen argument is an
10542 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10543 type. Similarly, if @var{exp2} is returned, its return type is the same
10550 __builtin_choose_expr ( \
10551 __builtin_types_compatible_p (typeof (x), double), \
10553 __builtin_choose_expr ( \
10554 __builtin_types_compatible_p (typeof (x), float), \
10556 /* @r{The void expression results in a compile-time error} \
10557 @r{when assigning the result to something.} */ \
10561 @emph{Note:} This construct is only available for C@. Furthermore, the
10562 unused expression (@var{exp1} or @var{exp2} depending on the value of
10563 @var{const_exp}) may still generate syntax errors. This may change in
10568 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10570 The built-in function @code{__builtin_complex} is provided for use in
10571 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10572 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10573 real binary floating-point type, and the result has the corresponding
10574 complex type with real and imaginary parts @var{real} and @var{imag}.
10575 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10576 infinities, NaNs and negative zeros are involved.
10580 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10581 You can use the built-in function @code{__builtin_constant_p} to
10582 determine if a value is known to be constant at compile time and hence
10583 that GCC can perform constant-folding on expressions involving that
10584 value. The argument of the function is the value to test. The function
10585 returns the integer 1 if the argument is known to be a compile-time
10586 constant and 0 if it is not known to be a compile-time constant. A
10587 return of 0 does not indicate that the value is @emph{not} a constant,
10588 but merely that GCC cannot prove it is a constant with the specified
10589 value of the @option{-O} option.
10591 You typically use this function in an embedded application where
10592 memory is a critical resource. If you have some complex calculation,
10593 you may want it to be folded if it involves constants, but need to call
10594 a function if it does not. For example:
10597 #define Scale_Value(X) \
10598 (__builtin_constant_p (X) \
10599 ? ((X) * SCALE + OFFSET) : Scale (X))
10602 You may use this built-in function in either a macro or an inline
10603 function. However, if you use it in an inlined function and pass an
10604 argument of the function as the argument to the built-in, GCC
10605 never returns 1 when you call the inline function with a string constant
10606 or compound literal (@pxref{Compound Literals}) and does not return 1
10607 when you pass a constant numeric value to the inline function unless you
10608 specify the @option{-O} option.
10610 You may also use @code{__builtin_constant_p} in initializers for static
10611 data. For instance, you can write
10614 static const int table[] = @{
10615 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10621 This is an acceptable initializer even if @var{EXPRESSION} is not a
10622 constant expression, including the case where
10623 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10624 folded to a constant but @var{EXPRESSION} contains operands that are
10625 not otherwise permitted in a static initializer (for example,
10626 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10627 built-in in this case, because it has no opportunity to perform
10631 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10632 @opindex fprofile-arcs
10633 You may use @code{__builtin_expect} to provide the compiler with
10634 branch prediction information. In general, you should prefer to
10635 use actual profile feedback for this (@option{-fprofile-arcs}), as
10636 programmers are notoriously bad at predicting how their programs
10637 actually perform. However, there are applications in which this
10638 data is hard to collect.
10640 The return value is the value of @var{exp}, which should be an integral
10641 expression. The semantics of the built-in are that it is expected that
10642 @var{exp} == @var{c}. For example:
10645 if (__builtin_expect (x, 0))
10650 indicates that we do not expect to call @code{foo}, since
10651 we expect @code{x} to be zero. Since you are limited to integral
10652 expressions for @var{exp}, you should use constructions such as
10655 if (__builtin_expect (ptr != NULL, 1))
10660 when testing pointer or floating-point values.
10663 @deftypefn {Built-in Function} void __builtin_trap (void)
10664 This function causes the program to exit abnormally. GCC implements
10665 this function by using a target-dependent mechanism (such as
10666 intentionally executing an illegal instruction) or by calling
10667 @code{abort}. The mechanism used may vary from release to release so
10668 you should not rely on any particular implementation.
10671 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10672 If control flow reaches the point of the @code{__builtin_unreachable},
10673 the program is undefined. It is useful in situations where the
10674 compiler cannot deduce the unreachability of the code.
10676 One such case is immediately following an @code{asm} statement that
10677 either never terminates, or one that transfers control elsewhere
10678 and never returns. In this example, without the
10679 @code{__builtin_unreachable}, GCC issues a warning that control
10680 reaches the end of a non-void function. It also generates code
10681 to return after the @code{asm}.
10684 int f (int c, int v)
10692 asm("jmp error_handler");
10693 __builtin_unreachable ();
10699 Because the @code{asm} statement unconditionally transfers control out
10700 of the function, control never reaches the end of the function
10701 body. The @code{__builtin_unreachable} is in fact unreachable and
10702 communicates this fact to the compiler.
10704 Another use for @code{__builtin_unreachable} is following a call a
10705 function that never returns but that is not declared
10706 @code{__attribute__((noreturn))}, as in this example:
10709 void function_that_never_returns (void);
10719 function_that_never_returns ();
10720 __builtin_unreachable ();
10727 @deftypefn {Built-in Function} void *__builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
10728 This function returns its first argument, and allows the compiler
10729 to assume that the returned pointer is at least @var{align} bytes
10730 aligned. This built-in can have either two or three arguments,
10731 if it has three, the third argument should have integer type, and
10732 if it is nonzero means misalignment offset. For example:
10735 void *x = __builtin_assume_aligned (arg, 16);
10739 means that the compiler can assume @code{x}, set to @code{arg}, is at least
10740 16-byte aligned, while:
10743 void *x = __builtin_assume_aligned (arg, 32, 8);
10747 means that the compiler can assume for @code{x}, set to @code{arg}, that
10748 @code{(char *) x - 8} is 32-byte aligned.
10751 @deftypefn {Built-in Function} int __builtin_LINE ()
10752 This function is the equivalent to the preprocessor @code{__LINE__}
10753 macro and returns the line number of the invocation of the built-in.
10754 In a C++ default argument for a function @var{F}, it gets the line number of
10755 the call to @var{F}.
10758 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
10759 This function is the equivalent to the preprocessor @code{__FUNCTION__}
10760 macro and returns the function name the invocation of the built-in is in.
10763 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
10764 This function is the equivalent to the preprocessor @code{__FILE__}
10765 macro and returns the file name the invocation of the built-in is in.
10766 In a C++ default argument for a function @var{F}, it gets the file name of
10767 the call to @var{F}.
10770 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
10771 This function is used to flush the processor's instruction cache for
10772 the region of memory between @var{begin} inclusive and @var{end}
10773 exclusive. Some targets require that the instruction cache be
10774 flushed, after modifying memory containing code, in order to obtain
10775 deterministic behavior.
10777 If the target does not require instruction cache flushes,
10778 @code{__builtin___clear_cache} has no effect. Otherwise either
10779 instructions are emitted in-line to clear the instruction cache or a
10780 call to the @code{__clear_cache} function in libgcc is made.
10783 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
10784 This function is used to minimize cache-miss latency by moving data into
10785 a cache before it is accessed.
10786 You can insert calls to @code{__builtin_prefetch} into code for which
10787 you know addresses of data in memory that is likely to be accessed soon.
10788 If the target supports them, data prefetch instructions are generated.
10789 If the prefetch is done early enough before the access then the data will
10790 be in the cache by the time it is accessed.
10792 The value of @var{addr} is the address of the memory to prefetch.
10793 There are two optional arguments, @var{rw} and @var{locality}.
10794 The value of @var{rw} is a compile-time constant one or zero; one
10795 means that the prefetch is preparing for a write to the memory address
10796 and zero, the default, means that the prefetch is preparing for a read.
10797 The value @var{locality} must be a compile-time constant integer between
10798 zero and three. A value of zero means that the data has no temporal
10799 locality, so it need not be left in the cache after the access. A value
10800 of three means that the data has a high degree of temporal locality and
10801 should be left in all levels of cache possible. Values of one and two
10802 mean, respectively, a low or moderate degree of temporal locality. The
10806 for (i = 0; i < n; i++)
10808 a[i] = a[i] + b[i];
10809 __builtin_prefetch (&a[i+j], 1, 1);
10810 __builtin_prefetch (&b[i+j], 0, 1);
10815 Data prefetch does not generate faults if @var{addr} is invalid, but
10816 the address expression itself must be valid. For example, a prefetch
10817 of @code{p->next} does not fault if @code{p->next} is not a valid
10818 address, but evaluation faults if @code{p} is not a valid address.
10820 If the target does not support data prefetch, the address expression
10821 is evaluated if it includes side effects but no other code is generated
10822 and GCC does not issue a warning.
10825 @deftypefn {Built-in Function} double __builtin_huge_val (void)
10826 Returns a positive infinity, if supported by the floating-point format,
10827 else @code{DBL_MAX}. This function is suitable for implementing the
10828 ISO C macro @code{HUGE_VAL}.
10831 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
10832 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
10835 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
10836 Similar to @code{__builtin_huge_val}, except the return
10837 type is @code{long double}.
10840 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
10841 This built-in implements the C99 fpclassify functionality. The first
10842 five int arguments should be the target library's notion of the
10843 possible FP classes and are used for return values. They must be
10844 constant values and they must appear in this order: @code{FP_NAN},
10845 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
10846 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
10847 to classify. GCC treats the last argument as type-generic, which
10848 means it does not do default promotion from float to double.
10851 @deftypefn {Built-in Function} double __builtin_inf (void)
10852 Similar to @code{__builtin_huge_val}, except a warning is generated
10853 if the target floating-point format does not support infinities.
10856 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
10857 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
10860 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
10861 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
10864 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
10865 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
10868 @deftypefn {Built-in Function} float __builtin_inff (void)
10869 Similar to @code{__builtin_inf}, except the return type is @code{float}.
10870 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
10873 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
10874 Similar to @code{__builtin_inf}, except the return
10875 type is @code{long double}.
10878 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
10879 Similar to @code{isinf}, except the return value is -1 for
10880 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
10881 Note while the parameter list is an
10882 ellipsis, this function only accepts exactly one floating-point
10883 argument. GCC treats this parameter as type-generic, which means it
10884 does not do default promotion from float to double.
10887 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
10888 This is an implementation of the ISO C99 function @code{nan}.
10890 Since ISO C99 defines this function in terms of @code{strtod}, which we
10891 do not implement, a description of the parsing is in order. The string
10892 is parsed as by @code{strtol}; that is, the base is recognized by
10893 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
10894 in the significand such that the least significant bit of the number
10895 is at the least significant bit of the significand. The number is
10896 truncated to fit the significand field provided. The significand is
10897 forced to be a quiet NaN@.
10899 This function, if given a string literal all of which would have been
10900 consumed by @code{strtol}, is evaluated early enough that it is considered a
10901 compile-time constant.
10904 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
10905 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
10908 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
10909 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
10912 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
10913 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
10916 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
10917 Similar to @code{__builtin_nan}, except the return type is @code{float}.
10920 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
10921 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
10924 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
10925 Similar to @code{__builtin_nan}, except the significand is forced
10926 to be a signaling NaN@. The @code{nans} function is proposed by
10927 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
10930 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
10931 Similar to @code{__builtin_nans}, except the return type is @code{float}.
10934 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
10935 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
10938 @deftypefn {Built-in Function} int __builtin_ffs (int x)
10939 Returns one plus the index of the least significant 1-bit of @var{x}, or
10940 if @var{x} is zero, returns zero.
10943 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
10944 Returns the number of leading 0-bits in @var{x}, starting at the most
10945 significant bit position. If @var{x} is 0, the result is undefined.
10948 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
10949 Returns the number of trailing 0-bits in @var{x}, starting at the least
10950 significant bit position. If @var{x} is 0, the result is undefined.
10953 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
10954 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
10955 number of bits following the most significant bit that are identical
10956 to it. There are no special cases for 0 or other values.
10959 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
10960 Returns the number of 1-bits in @var{x}.
10963 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
10964 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
10968 @deftypefn {Built-in Function} int __builtin_ffsl (long)
10969 Similar to @code{__builtin_ffs}, except the argument type is
10973 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
10974 Similar to @code{__builtin_clz}, except the argument type is
10975 @code{unsigned long}.
10978 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
10979 Similar to @code{__builtin_ctz}, except the argument type is
10980 @code{unsigned long}.
10983 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
10984 Similar to @code{__builtin_clrsb}, except the argument type is
10988 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
10989 Similar to @code{__builtin_popcount}, except the argument type is
10990 @code{unsigned long}.
10993 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
10994 Similar to @code{__builtin_parity}, except the argument type is
10995 @code{unsigned long}.
10998 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
10999 Similar to @code{__builtin_ffs}, except the argument type is
11003 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11004 Similar to @code{__builtin_clz}, except the argument type is
11005 @code{unsigned long long}.
11008 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11009 Similar to @code{__builtin_ctz}, except the argument type is
11010 @code{unsigned long long}.
11013 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11014 Similar to @code{__builtin_clrsb}, except the argument type is
11018 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11019 Similar to @code{__builtin_popcount}, except the argument type is
11020 @code{unsigned long long}.
11023 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11024 Similar to @code{__builtin_parity}, except the argument type is
11025 @code{unsigned long long}.
11028 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11029 Returns the first argument raised to the power of the second. Unlike the
11030 @code{pow} function no guarantees about precision and rounding are made.
11033 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11034 Similar to @code{__builtin_powi}, except the argument and return types
11038 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11039 Similar to @code{__builtin_powi}, except the argument and return types
11040 are @code{long double}.
11043 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11044 Returns @var{x} with the order of the bytes reversed; for example,
11045 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11049 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11050 Similar to @code{__builtin_bswap16}, except the argument and return types
11054 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11055 Similar to @code{__builtin_bswap32}, except the argument and return types
11059 @node Target Builtins
11060 @section Built-in Functions Specific to Particular Target Machines
11062 On some target machines, GCC supports many built-in functions specific
11063 to those machines. Generally these generate calls to specific machine
11064 instructions, but allow the compiler to schedule those calls.
11067 * AArch64 Built-in Functions::
11068 * Alpha Built-in Functions::
11069 * Altera Nios II Built-in Functions::
11070 * ARC Built-in Functions::
11071 * ARC SIMD Built-in Functions::
11072 * ARM iWMMXt Built-in Functions::
11073 * ARM C Language Extensions (ACLE)::
11074 * ARM Floating Point Status and Control Intrinsics::
11075 * AVR Built-in Functions::
11076 * Blackfin Built-in Functions::
11077 * FR-V Built-in Functions::
11078 * MIPS DSP Built-in Functions::
11079 * MIPS Paired-Single Support::
11080 * MIPS Loongson Built-in Functions::
11081 * Other MIPS Built-in Functions::
11082 * MSP430 Built-in Functions::
11083 * NDS32 Built-in Functions::
11084 * picoChip Built-in Functions::
11085 * PowerPC Built-in Functions::
11086 * PowerPC AltiVec/VSX Built-in Functions::
11087 * PowerPC Hardware Transactional Memory Built-in Functions::
11088 * RX Built-in Functions::
11089 * S/390 System z Built-in Functions::
11090 * SH Built-in Functions::
11091 * SPARC VIS Built-in Functions::
11092 * SPU Built-in Functions::
11093 * TI C6X Built-in Functions::
11094 * TILE-Gx Built-in Functions::
11095 * TILEPro Built-in Functions::
11096 * x86 Built-in Functions::
11097 * x86 transactional memory intrinsics::
11100 @node AArch64 Built-in Functions
11101 @subsection AArch64 Built-in Functions
11103 These built-in functions are available for the AArch64 family of
11106 unsigned int __builtin_aarch64_get_fpcr ()
11107 void __builtin_aarch64_set_fpcr (unsigned int)
11108 unsigned int __builtin_aarch64_get_fpsr ()
11109 void __builtin_aarch64_set_fpsr (unsigned int)
11112 @node Alpha Built-in Functions
11113 @subsection Alpha Built-in Functions
11115 These built-in functions are available for the Alpha family of
11116 processors, depending on the command-line switches used.
11118 The following built-in functions are always available. They
11119 all generate the machine instruction that is part of the name.
11122 long __builtin_alpha_implver (void)
11123 long __builtin_alpha_rpcc (void)
11124 long __builtin_alpha_amask (long)
11125 long __builtin_alpha_cmpbge (long, long)
11126 long __builtin_alpha_extbl (long, long)
11127 long __builtin_alpha_extwl (long, long)
11128 long __builtin_alpha_extll (long, long)
11129 long __builtin_alpha_extql (long, long)
11130 long __builtin_alpha_extwh (long, long)
11131 long __builtin_alpha_extlh (long, long)
11132 long __builtin_alpha_extqh (long, long)
11133 long __builtin_alpha_insbl (long, long)
11134 long __builtin_alpha_inswl (long, long)
11135 long __builtin_alpha_insll (long, long)
11136 long __builtin_alpha_insql (long, long)
11137 long __builtin_alpha_inswh (long, long)
11138 long __builtin_alpha_inslh (long, long)
11139 long __builtin_alpha_insqh (long, long)
11140 long __builtin_alpha_mskbl (long, long)
11141 long __builtin_alpha_mskwl (long, long)
11142 long __builtin_alpha_mskll (long, long)
11143 long __builtin_alpha_mskql (long, long)
11144 long __builtin_alpha_mskwh (long, long)
11145 long __builtin_alpha_msklh (long, long)
11146 long __builtin_alpha_mskqh (long, long)
11147 long __builtin_alpha_umulh (long, long)
11148 long __builtin_alpha_zap (long, long)
11149 long __builtin_alpha_zapnot (long, long)
11152 The following built-in functions are always with @option{-mmax}
11153 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11154 later. They all generate the machine instruction that is part
11158 long __builtin_alpha_pklb (long)
11159 long __builtin_alpha_pkwb (long)
11160 long __builtin_alpha_unpkbl (long)
11161 long __builtin_alpha_unpkbw (long)
11162 long __builtin_alpha_minub8 (long, long)
11163 long __builtin_alpha_minsb8 (long, long)
11164 long __builtin_alpha_minuw4 (long, long)
11165 long __builtin_alpha_minsw4 (long, long)
11166 long __builtin_alpha_maxub8 (long, long)
11167 long __builtin_alpha_maxsb8 (long, long)
11168 long __builtin_alpha_maxuw4 (long, long)
11169 long __builtin_alpha_maxsw4 (long, long)
11170 long __builtin_alpha_perr (long, long)
11173 The following built-in functions are always with @option{-mcix}
11174 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11175 later. They all generate the machine instruction that is part
11179 long __builtin_alpha_cttz (long)
11180 long __builtin_alpha_ctlz (long)
11181 long __builtin_alpha_ctpop (long)
11184 The following built-in functions are available on systems that use the OSF/1
11185 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11186 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11187 @code{rdval} and @code{wrval}.
11190 void *__builtin_thread_pointer (void)
11191 void __builtin_set_thread_pointer (void *)
11194 @node Altera Nios II Built-in Functions
11195 @subsection Altera Nios II Built-in Functions
11197 These built-in functions are available for the Altera Nios II
11198 family of processors.
11200 The following built-in functions are always available. They
11201 all generate the machine instruction that is part of the name.
11204 int __builtin_ldbio (volatile const void *)
11205 int __builtin_ldbuio (volatile const void *)
11206 int __builtin_ldhio (volatile const void *)
11207 int __builtin_ldhuio (volatile const void *)
11208 int __builtin_ldwio (volatile const void *)
11209 void __builtin_stbio (volatile void *, int)
11210 void __builtin_sthio (volatile void *, int)
11211 void __builtin_stwio (volatile void *, int)
11212 void __builtin_sync (void)
11213 int __builtin_rdctl (int)
11214 int __builtin_rdprs (int, int)
11215 void __builtin_wrctl (int, int)
11216 void __builtin_flushd (volatile void *)
11217 void __builtin_flushda (volatile void *)
11218 int __builtin_wrpie (int);
11219 void __builtin_eni (int);
11220 int __builtin_ldex (volatile const void *)
11221 int __builtin_stex (volatile void *, int)
11222 int __builtin_ldsex (volatile const void *)
11223 int __builtin_stsex (volatile void *, int)
11226 The following built-in functions are always available. They
11227 all generate a Nios II Custom Instruction. The name of the
11228 function represents the types that the function takes and
11229 returns. The letter before the @code{n} is the return type
11230 or void if absent. The @code{n} represents the first parameter
11231 to all the custom instructions, the custom instruction number.
11232 The two letters after the @code{n} represent the up to two
11233 parameters to the function.
11235 The letters represent the following data types:
11238 @code{void} for return type and no parameter for parameter types.
11241 @code{int} for return type and parameter type
11244 @code{float} for return type and parameter type
11247 @code{void *} for return type and parameter type
11251 And the function names are:
11253 void __builtin_custom_n (void)
11254 void __builtin_custom_ni (int)
11255 void __builtin_custom_nf (float)
11256 void __builtin_custom_np (void *)
11257 void __builtin_custom_nii (int, int)
11258 void __builtin_custom_nif (int, float)
11259 void __builtin_custom_nip (int, void *)
11260 void __builtin_custom_nfi (float, int)
11261 void __builtin_custom_nff (float, float)
11262 void __builtin_custom_nfp (float, void *)
11263 void __builtin_custom_npi (void *, int)
11264 void __builtin_custom_npf (void *, float)
11265 void __builtin_custom_npp (void *, void *)
11266 int __builtin_custom_in (void)
11267 int __builtin_custom_ini (int)
11268 int __builtin_custom_inf (float)
11269 int __builtin_custom_inp (void *)
11270 int __builtin_custom_inii (int, int)
11271 int __builtin_custom_inif (int, float)
11272 int __builtin_custom_inip (int, void *)
11273 int __builtin_custom_infi (float, int)
11274 int __builtin_custom_inff (float, float)
11275 int __builtin_custom_infp (float, void *)
11276 int __builtin_custom_inpi (void *, int)
11277 int __builtin_custom_inpf (void *, float)
11278 int __builtin_custom_inpp (void *, void *)
11279 float __builtin_custom_fn (void)
11280 float __builtin_custom_fni (int)
11281 float __builtin_custom_fnf (float)
11282 float __builtin_custom_fnp (void *)
11283 float __builtin_custom_fnii (int, int)
11284 float __builtin_custom_fnif (int, float)
11285 float __builtin_custom_fnip (int, void *)
11286 float __builtin_custom_fnfi (float, int)
11287 float __builtin_custom_fnff (float, float)
11288 float __builtin_custom_fnfp (float, void *)
11289 float __builtin_custom_fnpi (void *, int)
11290 float __builtin_custom_fnpf (void *, float)
11291 float __builtin_custom_fnpp (void *, void *)
11292 void * __builtin_custom_pn (void)
11293 void * __builtin_custom_pni (int)
11294 void * __builtin_custom_pnf (float)
11295 void * __builtin_custom_pnp (void *)
11296 void * __builtin_custom_pnii (int, int)
11297 void * __builtin_custom_pnif (int, float)
11298 void * __builtin_custom_pnip (int, void *)
11299 void * __builtin_custom_pnfi (float, int)
11300 void * __builtin_custom_pnff (float, float)
11301 void * __builtin_custom_pnfp (float, void *)
11302 void * __builtin_custom_pnpi (void *, int)
11303 void * __builtin_custom_pnpf (void *, float)
11304 void * __builtin_custom_pnpp (void *, void *)
11307 @node ARC Built-in Functions
11308 @subsection ARC Built-in Functions
11310 The following built-in functions are provided for ARC targets. The
11311 built-ins generate the corresponding assembly instructions. In the
11312 examples given below, the generated code often requires an operand or
11313 result to be in a register. Where necessary further code will be
11314 generated to ensure this is true, but for brevity this is not
11315 described in each case.
11317 @emph{Note:} Using a built-in to generate an instruction not supported
11318 by a target may cause problems. At present the compiler is not
11319 guaranteed to detect such misuse, and as a result an internal compiler
11320 error may be generated.
11322 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11323 Return 1 if @var{val} is known to have the byte alignment given
11324 by @var{alignval}, otherwise return 0.
11325 Note that this is different from
11327 __alignof__(*(char *)@var{val}) >= alignval
11329 because __alignof__ sees only the type of the dereference, whereas
11330 __builtin_arc_align uses alignment information from the pointer
11331 as well as from the pointed-to type.
11332 The information available will depend on optimization level.
11335 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11342 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11343 The operand is the number of a register to be read. Generates:
11345 mov @var{dest}, r@var{regno}
11347 where the value in @var{dest} will be the result returned from the
11351 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11352 The first operand is the number of a register to be written, the
11353 second operand is a compile time constant to write into that
11354 register. Generates:
11356 mov r@var{regno}, @var{val}
11360 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11361 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11364 divaw @var{dest}, @var{a}, @var{b}
11366 where the value in @var{dest} will be the result returned from the
11370 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11377 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11378 The operand, @var{auxv}, is the address of an auxiliary register and
11379 must be a compile time constant. Generates:
11381 lr @var{dest}, [@var{auxr}]
11383 Where the value in @var{dest} will be the result returned from the
11387 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11388 Only available with @option{-mmul64}. Generates:
11390 mul64 @var{a}, @var{b}
11394 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11395 Only available with @option{-mmul64}. Generates:
11397 mulu64 @var{a}, @var{b}
11401 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11408 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11409 Only valid if the @samp{norm} instruction is available through the
11410 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11413 norm @var{dest}, @var{src}
11415 Where the value in @var{dest} will be the result returned from the
11419 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11420 Only valid if the @samp{normw} instruction is available through the
11421 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11424 normw @var{dest}, @var{src}
11426 Where the value in @var{dest} will be the result returned from the
11430 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11437 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11444 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11445 The first argument, @var{auxv}, is the address of an auxiliary
11446 register, the second argument, @var{val}, is a compile time constant
11447 to be written to the register. Generates:
11449 sr @var{auxr}, [@var{val}]
11453 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11454 Only valid with @option{-mswap}. Generates:
11456 swap @var{dest}, @var{src}
11458 Where the value in @var{dest} will be the result returned from the
11462 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11469 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11470 Only available with @option{-mcpu=ARC700}. Generates:
11476 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11477 Only available with @option{-mcpu=ARC700}. Generates:
11483 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11484 Only available with @option{-mcpu=ARC700}. Generates:
11490 The instructions generated by the following builtins are not
11491 considered as candidates for scheduling. They are not moved around by
11492 the compiler during scheduling, and thus can be expected to appear
11493 where they are put in the C code:
11495 __builtin_arc_brk()
11496 __builtin_arc_core_read()
11497 __builtin_arc_core_write()
11498 __builtin_arc_flag()
11500 __builtin_arc_sleep()
11502 __builtin_arc_swi()
11505 @node ARC SIMD Built-in Functions
11506 @subsection ARC SIMD Built-in Functions
11508 SIMD builtins provided by the compiler can be used to generate the
11509 vector instructions. This section describes the available builtins
11510 and their usage in programs. With the @option{-msimd} option, the
11511 compiler provides 128-bit vector types, which can be specified using
11512 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11513 can be included to use the following predefined types:
11515 typedef int __v4si __attribute__((vector_size(16)));
11516 typedef short __v8hi __attribute__((vector_size(16)));
11519 These types can be used to define 128-bit variables. The built-in
11520 functions listed in the following section can be used on these
11521 variables to generate the vector operations.
11523 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11524 @file{arc-simd.h} also provides equivalent macros called
11525 @code{_@var{someinsn}} that can be used for programming ease and
11526 improved readability. The following macros for DMA control are also
11529 #define _setup_dma_in_channel_reg _vdiwr
11530 #define _setup_dma_out_channel_reg _vdowr
11533 The following is a complete list of all the SIMD built-ins provided
11534 for ARC, grouped by calling signature.
11536 The following take two @code{__v8hi} arguments and return a
11537 @code{__v8hi} result:
11539 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11540 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11541 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11542 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11543 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11544 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11545 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11546 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11547 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11548 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11549 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11550 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11551 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11552 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11553 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11554 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11555 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11556 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11557 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11558 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11559 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11560 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11561 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11562 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11563 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11564 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11565 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11566 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11567 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11568 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11569 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11570 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11571 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11572 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11573 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11574 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11575 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11576 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11577 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11578 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11579 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11580 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11581 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11582 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11583 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11584 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11585 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11586 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11589 The following take one @code{__v8hi} and one @code{int} argument and return a
11590 @code{__v8hi} result:
11593 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11594 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11595 __v8hi __builtin_arc_vbminw (__v8hi, int)
11596 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11597 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11598 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11599 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11600 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11603 The following take one @code{__v8hi} argument and one @code{int} argument which
11604 must be a 3-bit compile time constant indicating a register number
11605 I0-I7. They return a @code{__v8hi} result.
11607 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11608 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11609 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11612 The following take one @code{__v8hi} argument and one @code{int}
11613 argument which must be a 6-bit compile time constant. They return a
11614 @code{__v8hi} result.
11616 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11617 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11618 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11619 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11620 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11621 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11622 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11625 The following take one @code{__v8hi} argument and one @code{int} argument which
11626 must be a 8-bit compile time constant. They return a @code{__v8hi}
11629 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11630 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11631 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11632 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11635 The following take two @code{int} arguments, the second of which which
11636 must be a 8-bit compile time constant. They return a @code{__v8hi}
11639 __v8hi __builtin_arc_vmovaw (int, const int)
11640 __v8hi __builtin_arc_vmovw (int, const int)
11641 __v8hi __builtin_arc_vmovzw (int, const int)
11644 The following take a single @code{__v8hi} argument and return a
11645 @code{__v8hi} result:
11647 __v8hi __builtin_arc_vabsaw (__v8hi)
11648 __v8hi __builtin_arc_vabsw (__v8hi)
11649 __v8hi __builtin_arc_vaddsuw (__v8hi)
11650 __v8hi __builtin_arc_vexch1 (__v8hi)
11651 __v8hi __builtin_arc_vexch2 (__v8hi)
11652 __v8hi __builtin_arc_vexch4 (__v8hi)
11653 __v8hi __builtin_arc_vsignw (__v8hi)
11654 __v8hi __builtin_arc_vupbaw (__v8hi)
11655 __v8hi __builtin_arc_vupbw (__v8hi)
11656 __v8hi __builtin_arc_vupsbaw (__v8hi)
11657 __v8hi __builtin_arc_vupsbw (__v8hi)
11660 The following take two @code{int} arguments and return no result:
11662 void __builtin_arc_vdirun (int, int)
11663 void __builtin_arc_vdorun (int, int)
11666 The following take two @code{int} arguments and return no result. The
11667 first argument must a 3-bit compile time constant indicating one of
11668 the DR0-DR7 DMA setup channels:
11670 void __builtin_arc_vdiwr (const int, int)
11671 void __builtin_arc_vdowr (const int, int)
11674 The following take an @code{int} argument and return no result:
11676 void __builtin_arc_vendrec (int)
11677 void __builtin_arc_vrec (int)
11678 void __builtin_arc_vrecrun (int)
11679 void __builtin_arc_vrun (int)
11682 The following take a @code{__v8hi} argument and two @code{int}
11683 arguments and return a @code{__v8hi} result. The second argument must
11684 be a 3-bit compile time constants, indicating one the registers I0-I7,
11685 and the third argument must be an 8-bit compile time constant.
11687 @emph{Note:} Although the equivalent hardware instructions do not take
11688 an SIMD register as an operand, these builtins overwrite the relevant
11689 bits of the @code{__v8hi} register provided as the first argument with
11690 the value loaded from the @code{[Ib, u8]} location in the SDM.
11693 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
11694 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
11695 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
11696 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
11699 The following take two @code{int} arguments and return a @code{__v8hi}
11700 result. The first argument must be a 3-bit compile time constants,
11701 indicating one the registers I0-I7, and the second argument must be an
11702 8-bit compile time constant.
11705 __v8hi __builtin_arc_vld128 (const int, const int)
11706 __v8hi __builtin_arc_vld64w (const int, const int)
11709 The following take a @code{__v8hi} argument and two @code{int}
11710 arguments and return no result. The second argument must be a 3-bit
11711 compile time constants, indicating one the registers I0-I7, and the
11712 third argument must be an 8-bit compile time constant.
11715 void __builtin_arc_vst128 (__v8hi, const int, const int)
11716 void __builtin_arc_vst64 (__v8hi, const int, const int)
11719 The following take a @code{__v8hi} argument and three @code{int}
11720 arguments and return no result. The second argument must be a 3-bit
11721 compile-time constant, identifying the 16-bit sub-register to be
11722 stored, the third argument must be a 3-bit compile time constants,
11723 indicating one the registers I0-I7, and the fourth argument must be an
11724 8-bit compile time constant.
11727 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
11728 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
11731 @node ARM iWMMXt Built-in Functions
11732 @subsection ARM iWMMXt Built-in Functions
11734 These built-in functions are available for the ARM family of
11735 processors when the @option{-mcpu=iwmmxt} switch is used:
11738 typedef int v2si __attribute__ ((vector_size (8)));
11739 typedef short v4hi __attribute__ ((vector_size (8)));
11740 typedef char v8qi __attribute__ ((vector_size (8)));
11742 int __builtin_arm_getwcgr0 (void)
11743 void __builtin_arm_setwcgr0 (int)
11744 int __builtin_arm_getwcgr1 (void)
11745 void __builtin_arm_setwcgr1 (int)
11746 int __builtin_arm_getwcgr2 (void)
11747 void __builtin_arm_setwcgr2 (int)
11748 int __builtin_arm_getwcgr3 (void)
11749 void __builtin_arm_setwcgr3 (int)
11750 int __builtin_arm_textrmsb (v8qi, int)
11751 int __builtin_arm_textrmsh (v4hi, int)
11752 int __builtin_arm_textrmsw (v2si, int)
11753 int __builtin_arm_textrmub (v8qi, int)
11754 int __builtin_arm_textrmuh (v4hi, int)
11755 int __builtin_arm_textrmuw (v2si, int)
11756 v8qi __builtin_arm_tinsrb (v8qi, int, int)
11757 v4hi __builtin_arm_tinsrh (v4hi, int, int)
11758 v2si __builtin_arm_tinsrw (v2si, int, int)
11759 long long __builtin_arm_tmia (long long, int, int)
11760 long long __builtin_arm_tmiabb (long long, int, int)
11761 long long __builtin_arm_tmiabt (long long, int, int)
11762 long long __builtin_arm_tmiaph (long long, int, int)
11763 long long __builtin_arm_tmiatb (long long, int, int)
11764 long long __builtin_arm_tmiatt (long long, int, int)
11765 int __builtin_arm_tmovmskb (v8qi)
11766 int __builtin_arm_tmovmskh (v4hi)
11767 int __builtin_arm_tmovmskw (v2si)
11768 long long __builtin_arm_waccb (v8qi)
11769 long long __builtin_arm_wacch (v4hi)
11770 long long __builtin_arm_waccw (v2si)
11771 v8qi __builtin_arm_waddb (v8qi, v8qi)
11772 v8qi __builtin_arm_waddbss (v8qi, v8qi)
11773 v8qi __builtin_arm_waddbus (v8qi, v8qi)
11774 v4hi __builtin_arm_waddh (v4hi, v4hi)
11775 v4hi __builtin_arm_waddhss (v4hi, v4hi)
11776 v4hi __builtin_arm_waddhus (v4hi, v4hi)
11777 v2si __builtin_arm_waddw (v2si, v2si)
11778 v2si __builtin_arm_waddwss (v2si, v2si)
11779 v2si __builtin_arm_waddwus (v2si, v2si)
11780 v8qi __builtin_arm_walign (v8qi, v8qi, int)
11781 long long __builtin_arm_wand(long long, long long)
11782 long long __builtin_arm_wandn (long long, long long)
11783 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
11784 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
11785 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
11786 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
11787 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
11788 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
11789 v2si __builtin_arm_wcmpeqw (v2si, v2si)
11790 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
11791 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
11792 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
11793 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
11794 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
11795 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
11796 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
11797 long long __builtin_arm_wmacsz (v4hi, v4hi)
11798 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
11799 long long __builtin_arm_wmacuz (v4hi, v4hi)
11800 v4hi __builtin_arm_wmadds (v4hi, v4hi)
11801 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
11802 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
11803 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
11804 v2si __builtin_arm_wmaxsw (v2si, v2si)
11805 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
11806 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
11807 v2si __builtin_arm_wmaxuw (v2si, v2si)
11808 v8qi __builtin_arm_wminsb (v8qi, v8qi)
11809 v4hi __builtin_arm_wminsh (v4hi, v4hi)
11810 v2si __builtin_arm_wminsw (v2si, v2si)
11811 v8qi __builtin_arm_wminub (v8qi, v8qi)
11812 v4hi __builtin_arm_wminuh (v4hi, v4hi)
11813 v2si __builtin_arm_wminuw (v2si, v2si)
11814 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
11815 v4hi __builtin_arm_wmulul (v4hi, v4hi)
11816 v4hi __builtin_arm_wmulum (v4hi, v4hi)
11817 long long __builtin_arm_wor (long long, long long)
11818 v2si __builtin_arm_wpackdss (long long, long long)
11819 v2si __builtin_arm_wpackdus (long long, long long)
11820 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
11821 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
11822 v4hi __builtin_arm_wpackwss (v2si, v2si)
11823 v4hi __builtin_arm_wpackwus (v2si, v2si)
11824 long long __builtin_arm_wrord (long long, long long)
11825 long long __builtin_arm_wrordi (long long, int)
11826 v4hi __builtin_arm_wrorh (v4hi, long long)
11827 v4hi __builtin_arm_wrorhi (v4hi, int)
11828 v2si __builtin_arm_wrorw (v2si, long long)
11829 v2si __builtin_arm_wrorwi (v2si, int)
11830 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
11831 v2si __builtin_arm_wsadbz (v8qi, v8qi)
11832 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
11833 v2si __builtin_arm_wsadhz (v4hi, v4hi)
11834 v4hi __builtin_arm_wshufh (v4hi, int)
11835 long long __builtin_arm_wslld (long long, long long)
11836 long long __builtin_arm_wslldi (long long, int)
11837 v4hi __builtin_arm_wsllh (v4hi, long long)
11838 v4hi __builtin_arm_wsllhi (v4hi, int)
11839 v2si __builtin_arm_wsllw (v2si, long long)
11840 v2si __builtin_arm_wsllwi (v2si, int)
11841 long long __builtin_arm_wsrad (long long, long long)
11842 long long __builtin_arm_wsradi (long long, int)
11843 v4hi __builtin_arm_wsrah (v4hi, long long)
11844 v4hi __builtin_arm_wsrahi (v4hi, int)
11845 v2si __builtin_arm_wsraw (v2si, long long)
11846 v2si __builtin_arm_wsrawi (v2si, int)
11847 long long __builtin_arm_wsrld (long long, long long)
11848 long long __builtin_arm_wsrldi (long long, int)
11849 v4hi __builtin_arm_wsrlh (v4hi, long long)
11850 v4hi __builtin_arm_wsrlhi (v4hi, int)
11851 v2si __builtin_arm_wsrlw (v2si, long long)
11852 v2si __builtin_arm_wsrlwi (v2si, int)
11853 v8qi __builtin_arm_wsubb (v8qi, v8qi)
11854 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
11855 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
11856 v4hi __builtin_arm_wsubh (v4hi, v4hi)
11857 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
11858 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
11859 v2si __builtin_arm_wsubw (v2si, v2si)
11860 v2si __builtin_arm_wsubwss (v2si, v2si)
11861 v2si __builtin_arm_wsubwus (v2si, v2si)
11862 v4hi __builtin_arm_wunpckehsb (v8qi)
11863 v2si __builtin_arm_wunpckehsh (v4hi)
11864 long long __builtin_arm_wunpckehsw (v2si)
11865 v4hi __builtin_arm_wunpckehub (v8qi)
11866 v2si __builtin_arm_wunpckehuh (v4hi)
11867 long long __builtin_arm_wunpckehuw (v2si)
11868 v4hi __builtin_arm_wunpckelsb (v8qi)
11869 v2si __builtin_arm_wunpckelsh (v4hi)
11870 long long __builtin_arm_wunpckelsw (v2si)
11871 v4hi __builtin_arm_wunpckelub (v8qi)
11872 v2si __builtin_arm_wunpckeluh (v4hi)
11873 long long __builtin_arm_wunpckeluw (v2si)
11874 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
11875 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
11876 v2si __builtin_arm_wunpckihw (v2si, v2si)
11877 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
11878 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
11879 v2si __builtin_arm_wunpckilw (v2si, v2si)
11880 long long __builtin_arm_wxor (long long, long long)
11881 long long __builtin_arm_wzero ()
11885 @node ARM C Language Extensions (ACLE)
11886 @subsection ARM C Language Extensions (ACLE)
11888 GCC implements extensions for C as described in the ARM C Language
11889 Extensions (ACLE) specification, which can be found at
11890 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
11892 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
11893 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
11894 intrinsics can be found at
11895 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
11896 The built-in intrinsics for the Advanced SIMD extension are available when
11899 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
11900 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
11901 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
11902 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
11905 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
11906 availability of extensions.
11908 @node ARM Floating Point Status and Control Intrinsics
11909 @subsection ARM Floating Point Status and Control Intrinsics
11911 These built-in functions are available for the ARM family of
11912 processors with floating-point unit.
11915 unsigned int __builtin_arm_get_fpscr ()
11916 void __builtin_arm_set_fpscr (unsigned int)
11919 @node AVR Built-in Functions
11920 @subsection AVR Built-in Functions
11922 For each built-in function for AVR, there is an equally named,
11923 uppercase built-in macro defined. That way users can easily query if
11924 or if not a specific built-in is implemented or not. For example, if
11925 @code{__builtin_avr_nop} is available the macro
11926 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
11928 The following built-in functions map to the respective machine
11929 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
11930 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
11931 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
11932 as library call if no hardware multiplier is available.
11935 void __builtin_avr_nop (void)
11936 void __builtin_avr_sei (void)
11937 void __builtin_avr_cli (void)
11938 void __builtin_avr_sleep (void)
11939 void __builtin_avr_wdr (void)
11940 unsigned char __builtin_avr_swap (unsigned char)
11941 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
11942 int __builtin_avr_fmuls (char, char)
11943 int __builtin_avr_fmulsu (char, unsigned char)
11946 In order to delay execution for a specific number of cycles, GCC
11949 void __builtin_avr_delay_cycles (unsigned long ticks)
11953 @code{ticks} is the number of ticks to delay execution. Note that this
11954 built-in does not take into account the effect of interrupts that
11955 might increase delay time. @code{ticks} must be a compile-time
11956 integer constant; delays with a variable number of cycles are not supported.
11959 char __builtin_avr_flash_segment (const __memx void*)
11963 This built-in takes a byte address to the 24-bit
11964 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
11965 the number of the flash segment (the 64 KiB chunk) where the address
11966 points to. Counting starts at @code{0}.
11967 If the address does not point to flash memory, return @code{-1}.
11970 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
11974 Insert bits from @var{bits} into @var{val} and return the resulting
11975 value. The nibbles of @var{map} determine how the insertion is
11976 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
11978 @item If @var{X} is @code{0xf},
11979 then the @var{n}-th bit of @var{val} is returned unaltered.
11981 @item If X is in the range 0@dots{}7,
11982 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
11984 @item If X is in the range 8@dots{}@code{0xe},
11985 then the @var{n}-th result bit is undefined.
11989 One typical use case for this built-in is adjusting input and
11990 output values to non-contiguous port layouts. Some examples:
11993 // same as val, bits is unused
11994 __builtin_avr_insert_bits (0xffffffff, bits, val)
11998 // same as bits, val is unused
11999 __builtin_avr_insert_bits (0x76543210, bits, val)
12003 // same as rotating bits by 4
12004 __builtin_avr_insert_bits (0x32107654, bits, 0)
12008 // high nibble of result is the high nibble of val
12009 // low nibble of result is the low nibble of bits
12010 __builtin_avr_insert_bits (0xffff3210, bits, val)
12014 // reverse the bit order of bits
12015 __builtin_avr_insert_bits (0x01234567, bits, 0)
12018 @node Blackfin Built-in Functions
12019 @subsection Blackfin Built-in Functions
12021 Currently, there are two Blackfin-specific built-in functions. These are
12022 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12023 using inline assembly; by using these built-in functions the compiler can
12024 automatically add workarounds for hardware errata involving these
12025 instructions. These functions are named as follows:
12028 void __builtin_bfin_csync (void)
12029 void __builtin_bfin_ssync (void)
12032 @node FR-V Built-in Functions
12033 @subsection FR-V Built-in Functions
12035 GCC provides many FR-V-specific built-in functions. In general,
12036 these functions are intended to be compatible with those described
12037 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12038 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12039 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12040 pointer rather than by value.
12042 Most of the functions are named after specific FR-V instructions.
12043 Such functions are said to be ``directly mapped'' and are summarized
12044 here in tabular form.
12048 * Directly-mapped Integer Functions::
12049 * Directly-mapped Media Functions::
12050 * Raw read/write Functions::
12051 * Other Built-in Functions::
12054 @node Argument Types
12055 @subsubsection Argument Types
12057 The arguments to the built-in functions can be divided into three groups:
12058 register numbers, compile-time constants and run-time values. In order
12059 to make this classification clear at a glance, the arguments and return
12060 values are given the following pseudo types:
12062 @multitable @columnfractions .20 .30 .15 .35
12063 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12064 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12065 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12066 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12067 @item @code{uw2} @tab @code{unsigned long long} @tab No
12068 @tab an unsigned doubleword
12069 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12070 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12071 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12072 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12075 These pseudo types are not defined by GCC, they are simply a notational
12076 convenience used in this manual.
12078 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12079 and @code{sw2} are evaluated at run time. They correspond to
12080 register operands in the underlying FR-V instructions.
12082 @code{const} arguments represent immediate operands in the underlying
12083 FR-V instructions. They must be compile-time constants.
12085 @code{acc} arguments are evaluated at compile time and specify the number
12086 of an accumulator register. For example, an @code{acc} argument of 2
12087 selects the ACC2 register.
12089 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12090 number of an IACC register. See @pxref{Other Built-in Functions}
12093 @node Directly-mapped Integer Functions
12094 @subsubsection Directly-Mapped Integer Functions
12096 The functions listed below map directly to FR-V I-type instructions.
12098 @multitable @columnfractions .45 .32 .23
12099 @item Function prototype @tab Example usage @tab Assembly output
12100 @item @code{sw1 __ADDSS (sw1, sw1)}
12101 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12102 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12103 @item @code{sw1 __SCAN (sw1, sw1)}
12104 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12105 @tab @code{SCAN @var{a},@var{b},@var{c}}
12106 @item @code{sw1 __SCUTSS (sw1)}
12107 @tab @code{@var{b} = __SCUTSS (@var{a})}
12108 @tab @code{SCUTSS @var{a},@var{b}}
12109 @item @code{sw1 __SLASS (sw1, sw1)}
12110 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12111 @tab @code{SLASS @var{a},@var{b},@var{c}}
12112 @item @code{void __SMASS (sw1, sw1)}
12113 @tab @code{__SMASS (@var{a}, @var{b})}
12114 @tab @code{SMASS @var{a},@var{b}}
12115 @item @code{void __SMSSS (sw1, sw1)}
12116 @tab @code{__SMSSS (@var{a}, @var{b})}
12117 @tab @code{SMSSS @var{a},@var{b}}
12118 @item @code{void __SMU (sw1, sw1)}
12119 @tab @code{__SMU (@var{a}, @var{b})}
12120 @tab @code{SMU @var{a},@var{b}}
12121 @item @code{sw2 __SMUL (sw1, sw1)}
12122 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12123 @tab @code{SMUL @var{a},@var{b},@var{c}}
12124 @item @code{sw1 __SUBSS (sw1, sw1)}
12125 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12126 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12127 @item @code{uw2 __UMUL (uw1, uw1)}
12128 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12129 @tab @code{UMUL @var{a},@var{b},@var{c}}
12132 @node Directly-mapped Media Functions
12133 @subsubsection Directly-Mapped Media Functions
12135 The functions listed below map directly to FR-V M-type instructions.
12137 @multitable @columnfractions .45 .32 .23
12138 @item Function prototype @tab Example usage @tab Assembly output
12139 @item @code{uw1 __MABSHS (sw1)}
12140 @tab @code{@var{b} = __MABSHS (@var{a})}
12141 @tab @code{MABSHS @var{a},@var{b}}
12142 @item @code{void __MADDACCS (acc, acc)}
12143 @tab @code{__MADDACCS (@var{b}, @var{a})}
12144 @tab @code{MADDACCS @var{a},@var{b}}
12145 @item @code{sw1 __MADDHSS (sw1, sw1)}
12146 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12147 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12148 @item @code{uw1 __MADDHUS (uw1, uw1)}
12149 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12150 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12151 @item @code{uw1 __MAND (uw1, uw1)}
12152 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12153 @tab @code{MAND @var{a},@var{b},@var{c}}
12154 @item @code{void __MASACCS (acc, acc)}
12155 @tab @code{__MASACCS (@var{b}, @var{a})}
12156 @tab @code{MASACCS @var{a},@var{b}}
12157 @item @code{uw1 __MAVEH (uw1, uw1)}
12158 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12159 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12160 @item @code{uw2 __MBTOH (uw1)}
12161 @tab @code{@var{b} = __MBTOH (@var{a})}
12162 @tab @code{MBTOH @var{a},@var{b}}
12163 @item @code{void __MBTOHE (uw1 *, uw1)}
12164 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12165 @tab @code{MBTOHE @var{a},@var{b}}
12166 @item @code{void __MCLRACC (acc)}
12167 @tab @code{__MCLRACC (@var{a})}
12168 @tab @code{MCLRACC @var{a}}
12169 @item @code{void __MCLRACCA (void)}
12170 @tab @code{__MCLRACCA ()}
12171 @tab @code{MCLRACCA}
12172 @item @code{uw1 __Mcop1 (uw1, uw1)}
12173 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12174 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12175 @item @code{uw1 __Mcop2 (uw1, uw1)}
12176 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12177 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12178 @item @code{uw1 __MCPLHI (uw2, const)}
12179 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12180 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12181 @item @code{uw1 __MCPLI (uw2, const)}
12182 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12183 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12184 @item @code{void __MCPXIS (acc, sw1, sw1)}
12185 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12186 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12187 @item @code{void __MCPXIU (acc, uw1, uw1)}
12188 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12189 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12190 @item @code{void __MCPXRS (acc, sw1, sw1)}
12191 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12192 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12193 @item @code{void __MCPXRU (acc, uw1, uw1)}
12194 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12195 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12196 @item @code{uw1 __MCUT (acc, uw1)}
12197 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12198 @tab @code{MCUT @var{a},@var{b},@var{c}}
12199 @item @code{uw1 __MCUTSS (acc, sw1)}
12200 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12201 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12202 @item @code{void __MDADDACCS (acc, acc)}
12203 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12204 @tab @code{MDADDACCS @var{a},@var{b}}
12205 @item @code{void __MDASACCS (acc, acc)}
12206 @tab @code{__MDASACCS (@var{b}, @var{a})}
12207 @tab @code{MDASACCS @var{a},@var{b}}
12208 @item @code{uw2 __MDCUTSSI (acc, const)}
12209 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12210 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12211 @item @code{uw2 __MDPACKH (uw2, uw2)}
12212 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12213 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12214 @item @code{uw2 __MDROTLI (uw2, const)}
12215 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12216 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12217 @item @code{void __MDSUBACCS (acc, acc)}
12218 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12219 @tab @code{MDSUBACCS @var{a},@var{b}}
12220 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12221 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12222 @tab @code{MDUNPACKH @var{a},@var{b}}
12223 @item @code{uw2 __MEXPDHD (uw1, const)}
12224 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12225 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12226 @item @code{uw1 __MEXPDHW (uw1, const)}
12227 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12228 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12229 @item @code{uw1 __MHDSETH (uw1, const)}
12230 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12231 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12232 @item @code{sw1 __MHDSETS (const)}
12233 @tab @code{@var{b} = __MHDSETS (@var{a})}
12234 @tab @code{MHDSETS #@var{a},@var{b}}
12235 @item @code{uw1 __MHSETHIH (uw1, const)}
12236 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12237 @tab @code{MHSETHIH #@var{a},@var{b}}
12238 @item @code{sw1 __MHSETHIS (sw1, const)}
12239 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12240 @tab @code{MHSETHIS #@var{a},@var{b}}
12241 @item @code{uw1 __MHSETLOH (uw1, const)}
12242 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12243 @tab @code{MHSETLOH #@var{a},@var{b}}
12244 @item @code{sw1 __MHSETLOS (sw1, const)}
12245 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12246 @tab @code{MHSETLOS #@var{a},@var{b}}
12247 @item @code{uw1 __MHTOB (uw2)}
12248 @tab @code{@var{b} = __MHTOB (@var{a})}
12249 @tab @code{MHTOB @var{a},@var{b}}
12250 @item @code{void __MMACHS (acc, sw1, sw1)}
12251 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12252 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12253 @item @code{void __MMACHU (acc, uw1, uw1)}
12254 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12255 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12256 @item @code{void __MMRDHS (acc, sw1, sw1)}
12257 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12258 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12259 @item @code{void __MMRDHU (acc, uw1, uw1)}
12260 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12261 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12262 @item @code{void __MMULHS (acc, sw1, sw1)}
12263 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12264 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12265 @item @code{void __MMULHU (acc, uw1, uw1)}
12266 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12267 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12268 @item @code{void __MMULXHS (acc, sw1, sw1)}
12269 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12270 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12271 @item @code{void __MMULXHU (acc, uw1, uw1)}
12272 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12273 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12274 @item @code{uw1 __MNOT (uw1)}
12275 @tab @code{@var{b} = __MNOT (@var{a})}
12276 @tab @code{MNOT @var{a},@var{b}}
12277 @item @code{uw1 __MOR (uw1, uw1)}
12278 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12279 @tab @code{MOR @var{a},@var{b},@var{c}}
12280 @item @code{uw1 __MPACKH (uh, uh)}
12281 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12282 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12283 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12284 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12285 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12286 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12287 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12288 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12289 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12290 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12291 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12292 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12293 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12294 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12295 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12296 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12297 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12298 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12299 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12300 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12301 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12302 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12303 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12304 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12305 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12306 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12307 @item @code{void __MQMACHS (acc, sw2, sw2)}
12308 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12309 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12310 @item @code{void __MQMACHU (acc, uw2, uw2)}
12311 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12312 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12313 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12314 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12315 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12316 @item @code{void __MQMULHS (acc, sw2, sw2)}
12317 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12318 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12319 @item @code{void __MQMULHU (acc, uw2, uw2)}
12320 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12321 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12322 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12323 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12324 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12325 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12326 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12327 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12328 @item @code{sw2 __MQSATHS (sw2, sw2)}
12329 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12330 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12331 @item @code{uw2 __MQSLLHI (uw2, int)}
12332 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12333 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12334 @item @code{sw2 __MQSRAHI (sw2, int)}
12335 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12336 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12337 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12338 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12339 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12340 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12341 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12342 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12343 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12344 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12345 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12346 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12347 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12348 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12349 @item @code{uw1 __MRDACC (acc)}
12350 @tab @code{@var{b} = __MRDACC (@var{a})}
12351 @tab @code{MRDACC @var{a},@var{b}}
12352 @item @code{uw1 __MRDACCG (acc)}
12353 @tab @code{@var{b} = __MRDACCG (@var{a})}
12354 @tab @code{MRDACCG @var{a},@var{b}}
12355 @item @code{uw1 __MROTLI (uw1, const)}
12356 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12357 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12358 @item @code{uw1 __MROTRI (uw1, const)}
12359 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12360 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12361 @item @code{sw1 __MSATHS (sw1, sw1)}
12362 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12363 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12364 @item @code{uw1 __MSATHU (uw1, uw1)}
12365 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12366 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12367 @item @code{uw1 __MSLLHI (uw1, const)}
12368 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12369 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12370 @item @code{sw1 __MSRAHI (sw1, const)}
12371 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12372 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12373 @item @code{uw1 __MSRLHI (uw1, const)}
12374 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12375 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12376 @item @code{void __MSUBACCS (acc, acc)}
12377 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12378 @tab @code{MSUBACCS @var{a},@var{b}}
12379 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12380 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12381 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12382 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12383 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12384 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12385 @item @code{void __MTRAP (void)}
12386 @tab @code{__MTRAP ()}
12388 @item @code{uw2 __MUNPACKH (uw1)}
12389 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12390 @tab @code{MUNPACKH @var{a},@var{b}}
12391 @item @code{uw1 __MWCUT (uw2, uw1)}
12392 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12393 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12394 @item @code{void __MWTACC (acc, uw1)}
12395 @tab @code{__MWTACC (@var{b}, @var{a})}
12396 @tab @code{MWTACC @var{a},@var{b}}
12397 @item @code{void __MWTACCG (acc, uw1)}
12398 @tab @code{__MWTACCG (@var{b}, @var{a})}
12399 @tab @code{MWTACCG @var{a},@var{b}}
12400 @item @code{uw1 __MXOR (uw1, uw1)}
12401 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12402 @tab @code{MXOR @var{a},@var{b},@var{c}}
12405 @node Raw read/write Functions
12406 @subsubsection Raw Read/Write Functions
12408 This sections describes built-in functions related to read and write
12409 instructions to access memory. These functions generate
12410 @code{membar} instructions to flush the I/O load and stores where
12411 appropriate, as described in Fujitsu's manual described above.
12415 @item unsigned char __builtin_read8 (void *@var{data})
12416 @item unsigned short __builtin_read16 (void *@var{data})
12417 @item unsigned long __builtin_read32 (void *@var{data})
12418 @item unsigned long long __builtin_read64 (void *@var{data})
12420 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12421 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12422 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12423 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12426 @node Other Built-in Functions
12427 @subsubsection Other Built-in Functions
12429 This section describes built-in functions that are not named after
12430 a specific FR-V instruction.
12433 @item sw2 __IACCreadll (iacc @var{reg})
12434 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12435 for future expansion and must be 0.
12437 @item sw1 __IACCreadl (iacc @var{reg})
12438 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12439 Other values of @var{reg} are rejected as invalid.
12441 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12442 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12443 is reserved for future expansion and must be 0.
12445 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12446 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12447 is 1. Other values of @var{reg} are rejected as invalid.
12449 @item void __data_prefetch0 (const void *@var{x})
12450 Use the @code{dcpl} instruction to load the contents of address @var{x}
12451 into the data cache.
12453 @item void __data_prefetch (const void *@var{x})
12454 Use the @code{nldub} instruction to load the contents of address @var{x}
12455 into the data cache. The instruction is issued in slot I1@.
12458 @node MIPS DSP Built-in Functions
12459 @subsection MIPS DSP Built-in Functions
12461 The MIPS DSP Application-Specific Extension (ASE) includes new
12462 instructions that are designed to improve the performance of DSP and
12463 media applications. It provides instructions that operate on packed
12464 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12466 GCC supports MIPS DSP operations using both the generic
12467 vector extensions (@pxref{Vector Extensions}) and a collection of
12468 MIPS-specific built-in functions. Both kinds of support are
12469 enabled by the @option{-mdsp} command-line option.
12471 Revision 2 of the ASE was introduced in the second half of 2006.
12472 This revision adds extra instructions to the original ASE, but is
12473 otherwise backwards-compatible with it. You can select revision 2
12474 using the command-line option @option{-mdspr2}; this option implies
12477 The SCOUNT and POS bits of the DSP control register are global. The
12478 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12479 POS bits. During optimization, the compiler does not delete these
12480 instructions and it does not delete calls to functions containing
12481 these instructions.
12483 At present, GCC only provides support for operations on 32-bit
12484 vectors. The vector type associated with 8-bit integer data is
12485 usually called @code{v4i8}, the vector type associated with Q7
12486 is usually called @code{v4q7}, the vector type associated with 16-bit
12487 integer data is usually called @code{v2i16}, and the vector type
12488 associated with Q15 is usually called @code{v2q15}. They can be
12489 defined in C as follows:
12492 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12493 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12494 typedef short v2i16 __attribute__ ((vector_size(4)));
12495 typedef short v2q15 __attribute__ ((vector_size(4)));
12498 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12499 initialized in the same way as aggregates. For example:
12502 v4i8 a = @{1, 2, 3, 4@};
12504 b = (v4i8) @{5, 6, 7, 8@};
12506 v2q15 c = @{0x0fcb, 0x3a75@};
12508 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12511 @emph{Note:} The CPU's endianness determines the order in which values
12512 are packed. On little-endian targets, the first value is the least
12513 significant and the last value is the most significant. The opposite
12514 order applies to big-endian targets. For example, the code above
12515 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12516 and @code{4} on big-endian targets.
12518 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12519 representation. As shown in this example, the integer representation
12520 of a Q7 value can be obtained by multiplying the fractional value by
12521 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12522 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12525 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12526 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12527 and @code{c} and @code{d} are @code{v2q15} values.
12529 @multitable @columnfractions .50 .50
12530 @item C code @tab MIPS instruction
12531 @item @code{a + b} @tab @code{addu.qb}
12532 @item @code{c + d} @tab @code{addq.ph}
12533 @item @code{a - b} @tab @code{subu.qb}
12534 @item @code{c - d} @tab @code{subq.ph}
12537 The table below lists the @code{v2i16} operation for which
12538 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12539 @code{v2i16} values.
12541 @multitable @columnfractions .50 .50
12542 @item C code @tab MIPS instruction
12543 @item @code{e * f} @tab @code{mul.ph}
12546 It is easier to describe the DSP built-in functions if we first define
12547 the following types:
12552 typedef unsigned int ui32;
12553 typedef long long a64;
12556 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12557 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12558 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12559 @code{long long}, but we use @code{a64} to indicate values that are
12560 placed in one of the four DSP accumulators (@code{$ac0},
12561 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12563 Also, some built-in functions prefer or require immediate numbers as
12564 parameters, because the corresponding DSP instructions accept both immediate
12565 numbers and register operands, or accept immediate numbers only. The
12566 immediate parameters are listed as follows.
12574 imm0_255: 0 to 255.
12575 imm_n32_31: -32 to 31.
12576 imm_n512_511: -512 to 511.
12579 The following built-in functions map directly to a particular MIPS DSP
12580 instruction. Please refer to the architecture specification
12581 for details on what each instruction does.
12584 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12585 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12586 q31 __builtin_mips_addq_s_w (q31, q31)
12587 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12588 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12589 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12590 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12591 q31 __builtin_mips_subq_s_w (q31, q31)
12592 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12593 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12594 i32 __builtin_mips_addsc (i32, i32)
12595 i32 __builtin_mips_addwc (i32, i32)
12596 i32 __builtin_mips_modsub (i32, i32)
12597 i32 __builtin_mips_raddu_w_qb (v4i8)
12598 v2q15 __builtin_mips_absq_s_ph (v2q15)
12599 q31 __builtin_mips_absq_s_w (q31)
12600 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12601 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12602 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12603 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12604 q31 __builtin_mips_preceq_w_phl (v2q15)
12605 q31 __builtin_mips_preceq_w_phr (v2q15)
12606 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12607 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12608 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12609 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12610 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12611 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12612 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12613 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12614 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12615 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12616 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12617 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12618 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12619 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12620 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12621 q31 __builtin_mips_shll_s_w (q31, i32)
12622 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12623 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12624 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12625 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12626 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12627 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12628 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12629 q31 __builtin_mips_shra_r_w (q31, i32)
12630 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12631 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12632 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12633 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12634 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12635 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12636 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12637 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12638 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12639 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12640 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12641 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12642 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12643 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12644 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12645 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12646 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12647 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12648 i32 __builtin_mips_bitrev (i32)
12649 i32 __builtin_mips_insv (i32, i32)
12650 v4i8 __builtin_mips_repl_qb (imm0_255)
12651 v4i8 __builtin_mips_repl_qb (i32)
12652 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12653 v2q15 __builtin_mips_repl_ph (i32)
12654 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12655 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12656 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12657 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12658 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12659 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12660 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12661 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12662 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12663 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12664 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12665 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12666 i32 __builtin_mips_extr_w (a64, imm0_31)
12667 i32 __builtin_mips_extr_w (a64, i32)
12668 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12669 i32 __builtin_mips_extr_s_h (a64, i32)
12670 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12671 i32 __builtin_mips_extr_rs_w (a64, i32)
12672 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12673 i32 __builtin_mips_extr_r_w (a64, i32)
12674 i32 __builtin_mips_extp (a64, imm0_31)
12675 i32 __builtin_mips_extp (a64, i32)
12676 i32 __builtin_mips_extpdp (a64, imm0_31)
12677 i32 __builtin_mips_extpdp (a64, i32)
12678 a64 __builtin_mips_shilo (a64, imm_n32_31)
12679 a64 __builtin_mips_shilo (a64, i32)
12680 a64 __builtin_mips_mthlip (a64, i32)
12681 void __builtin_mips_wrdsp (i32, imm0_63)
12682 i32 __builtin_mips_rddsp (imm0_63)
12683 i32 __builtin_mips_lbux (void *, i32)
12684 i32 __builtin_mips_lhx (void *, i32)
12685 i32 __builtin_mips_lwx (void *, i32)
12686 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12687 i32 __builtin_mips_bposge32 (void)
12688 a64 __builtin_mips_madd (a64, i32, i32);
12689 a64 __builtin_mips_maddu (a64, ui32, ui32);
12690 a64 __builtin_mips_msub (a64, i32, i32);
12691 a64 __builtin_mips_msubu (a64, ui32, ui32);
12692 a64 __builtin_mips_mult (i32, i32);
12693 a64 __builtin_mips_multu (ui32, ui32);
12696 The following built-in functions map directly to a particular MIPS DSP REV 2
12697 instruction. Please refer to the architecture specification
12698 for details on what each instruction does.
12701 v4q7 __builtin_mips_absq_s_qb (v4q7);
12702 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12703 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12704 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12705 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12706 i32 __builtin_mips_append (i32, i32, imm0_31);
12707 i32 __builtin_mips_balign (i32, i32, imm0_3);
12708 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12709 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12710 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12711 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12712 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12713 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12714 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12715 q31 __builtin_mips_mulq_rs_w (q31, q31);
12716 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12717 q31 __builtin_mips_mulq_s_w (q31, q31);
12718 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12719 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12720 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12721 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12722 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12723 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12724 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12725 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12726 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12727 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12728 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12729 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12730 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12731 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12732 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12733 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12734 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12735 q31 __builtin_mips_addqh_w (q31, q31);
12736 q31 __builtin_mips_addqh_r_w (q31, q31);
12737 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12738 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12739 q31 __builtin_mips_subqh_w (q31, q31);
12740 q31 __builtin_mips_subqh_r_w (q31, q31);
12741 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12742 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12743 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12744 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12745 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12746 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12750 @node MIPS Paired-Single Support
12751 @subsection MIPS Paired-Single Support
12753 The MIPS64 architecture includes a number of instructions that
12754 operate on pairs of single-precision floating-point values.
12755 Each pair is packed into a 64-bit floating-point register,
12756 with one element being designated the ``upper half'' and
12757 the other being designated the ``lower half''.
12759 GCC supports paired-single operations using both the generic
12760 vector extensions (@pxref{Vector Extensions}) and a collection of
12761 MIPS-specific built-in functions. Both kinds of support are
12762 enabled by the @option{-mpaired-single} command-line option.
12764 The vector type associated with paired-single values is usually
12765 called @code{v2sf}. It can be defined in C as follows:
12768 typedef float v2sf __attribute__ ((vector_size (8)));
12771 @code{v2sf} values are initialized in the same way as aggregates.
12775 v2sf a = @{1.5, 9.1@};
12778 b = (v2sf) @{e, f@};
12781 @emph{Note:} The CPU's endianness determines which value is stored in
12782 the upper half of a register and which value is stored in the lower half.
12783 On little-endian targets, the first value is the lower one and the second
12784 value is the upper one. The opposite order applies to big-endian targets.
12785 For example, the code above sets the lower half of @code{a} to
12786 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
12788 @node MIPS Loongson Built-in Functions
12789 @subsection MIPS Loongson Built-in Functions
12791 GCC provides intrinsics to access the SIMD instructions provided by the
12792 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
12793 available after inclusion of the @code{loongson.h} header file,
12794 operate on the following 64-bit vector types:
12797 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
12798 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
12799 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
12800 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
12801 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
12802 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
12805 The intrinsics provided are listed below; each is named after the
12806 machine instruction to which it corresponds, with suffixes added as
12807 appropriate to distinguish intrinsics that expand to the same machine
12808 instruction yet have different argument types. Refer to the architecture
12809 documentation for a description of the functionality of each
12813 int16x4_t packsswh (int32x2_t s, int32x2_t t);
12814 int8x8_t packsshb (int16x4_t s, int16x4_t t);
12815 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
12816 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
12817 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
12818 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
12819 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
12820 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
12821 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
12822 uint64_t paddd_u (uint64_t s, uint64_t t);
12823 int64_t paddd_s (int64_t s, int64_t t);
12824 int16x4_t paddsh (int16x4_t s, int16x4_t t);
12825 int8x8_t paddsb (int8x8_t s, int8x8_t t);
12826 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
12827 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
12828 uint64_t pandn_ud (uint64_t s, uint64_t t);
12829 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
12830 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
12831 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
12832 int64_t pandn_sd (int64_t s, int64_t t);
12833 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
12834 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
12835 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
12836 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
12837 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
12838 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
12839 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
12840 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
12841 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
12842 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
12843 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
12844 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
12845 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
12846 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
12847 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
12848 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
12849 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
12850 uint16x4_t pextrh_u (uint16x4_t s, int field);
12851 int16x4_t pextrh_s (int16x4_t s, int field);
12852 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
12853 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
12854 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
12855 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
12856 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
12857 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
12858 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
12859 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
12860 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
12861 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
12862 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
12863 int16x4_t pminsh (int16x4_t s, int16x4_t t);
12864 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
12865 uint8x8_t pmovmskb_u (uint8x8_t s);
12866 int8x8_t pmovmskb_s (int8x8_t s);
12867 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
12868 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
12869 int16x4_t pmullh (int16x4_t s, int16x4_t t);
12870 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
12871 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
12872 uint16x4_t biadd (uint8x8_t s);
12873 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
12874 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
12875 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
12876 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
12877 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
12878 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
12879 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
12880 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
12881 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
12882 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
12883 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
12884 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
12885 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
12886 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
12887 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
12888 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
12889 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
12890 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
12891 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
12892 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
12893 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
12894 uint64_t psubd_u (uint64_t s, uint64_t t);
12895 int64_t psubd_s (int64_t s, int64_t t);
12896 int16x4_t psubsh (int16x4_t s, int16x4_t t);
12897 int8x8_t psubsb (int8x8_t s, int8x8_t t);
12898 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
12899 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
12900 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
12901 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
12902 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
12903 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
12904 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
12905 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
12906 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
12907 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
12908 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
12909 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
12910 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
12911 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
12915 * Paired-Single Arithmetic::
12916 * Paired-Single Built-in Functions::
12917 * MIPS-3D Built-in Functions::
12920 @node Paired-Single Arithmetic
12921 @subsubsection Paired-Single Arithmetic
12923 The table below lists the @code{v2sf} operations for which hardware
12924 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
12925 values and @code{x} is an integral value.
12927 @multitable @columnfractions .50 .50
12928 @item C code @tab MIPS instruction
12929 @item @code{a + b} @tab @code{add.ps}
12930 @item @code{a - b} @tab @code{sub.ps}
12931 @item @code{-a} @tab @code{neg.ps}
12932 @item @code{a * b} @tab @code{mul.ps}
12933 @item @code{a * b + c} @tab @code{madd.ps}
12934 @item @code{a * b - c} @tab @code{msub.ps}
12935 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
12936 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
12937 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
12940 Note that the multiply-accumulate instructions can be disabled
12941 using the command-line option @code{-mno-fused-madd}.
12943 @node Paired-Single Built-in Functions
12944 @subsubsection Paired-Single Built-in Functions
12946 The following paired-single functions map directly to a particular
12947 MIPS instruction. Please refer to the architecture specification
12948 for details on what each instruction does.
12951 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
12952 Pair lower lower (@code{pll.ps}).
12954 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
12955 Pair upper lower (@code{pul.ps}).
12957 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
12958 Pair lower upper (@code{plu.ps}).
12960 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
12961 Pair upper upper (@code{puu.ps}).
12963 @item v2sf __builtin_mips_cvt_ps_s (float, float)
12964 Convert pair to paired single (@code{cvt.ps.s}).
12966 @item float __builtin_mips_cvt_s_pl (v2sf)
12967 Convert pair lower to single (@code{cvt.s.pl}).
12969 @item float __builtin_mips_cvt_s_pu (v2sf)
12970 Convert pair upper to single (@code{cvt.s.pu}).
12972 @item v2sf __builtin_mips_abs_ps (v2sf)
12973 Absolute value (@code{abs.ps}).
12975 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
12976 Align variable (@code{alnv.ps}).
12978 @emph{Note:} The value of the third parameter must be 0 or 4
12979 modulo 8, otherwise the result is unpredictable. Please read the
12980 instruction description for details.
12983 The following multi-instruction functions are also available.
12984 In each case, @var{cond} can be any of the 16 floating-point conditions:
12985 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
12986 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
12987 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
12990 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12991 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12992 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
12993 @code{movt.ps}/@code{movf.ps}).
12995 The @code{movt} functions return the value @var{x} computed by:
12998 c.@var{cond}.ps @var{cc},@var{a},@var{b}
12999 mov.ps @var{x},@var{c}
13000 movt.ps @var{x},@var{d},@var{cc}
13003 The @code{movf} functions are similar but use @code{movf.ps} instead
13006 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13007 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13008 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13009 @code{bc1t}/@code{bc1f}).
13011 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13012 and return either the upper or lower half of the result. For example:
13016 if (__builtin_mips_upper_c_eq_ps (a, b))
13017 upper_halves_are_equal ();
13019 upper_halves_are_unequal ();
13021 if (__builtin_mips_lower_c_eq_ps (a, b))
13022 lower_halves_are_equal ();
13024 lower_halves_are_unequal ();
13028 @node MIPS-3D Built-in Functions
13029 @subsubsection MIPS-3D Built-in Functions
13031 The MIPS-3D Application-Specific Extension (ASE) includes additional
13032 paired-single instructions that are designed to improve the performance
13033 of 3D graphics operations. Support for these instructions is controlled
13034 by the @option{-mips3d} command-line option.
13036 The functions listed below map directly to a particular MIPS-3D
13037 instruction. Please refer to the architecture specification for
13038 more details on what each instruction does.
13041 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13042 Reduction add (@code{addr.ps}).
13044 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13045 Reduction multiply (@code{mulr.ps}).
13047 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13048 Convert paired single to paired word (@code{cvt.pw.ps}).
13050 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13051 Convert paired word to paired single (@code{cvt.ps.pw}).
13053 @item float __builtin_mips_recip1_s (float)
13054 @itemx double __builtin_mips_recip1_d (double)
13055 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13056 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13058 @item float __builtin_mips_recip2_s (float, float)
13059 @itemx double __builtin_mips_recip2_d (double, double)
13060 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13061 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13063 @item float __builtin_mips_rsqrt1_s (float)
13064 @itemx double __builtin_mips_rsqrt1_d (double)
13065 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13066 Reduced-precision reciprocal square root (sequence step 1)
13067 (@code{rsqrt1.@var{fmt}}).
13069 @item float __builtin_mips_rsqrt2_s (float, float)
13070 @itemx double __builtin_mips_rsqrt2_d (double, double)
13071 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13072 Reduced-precision reciprocal square root (sequence step 2)
13073 (@code{rsqrt2.@var{fmt}}).
13076 The following multi-instruction functions are also available.
13077 In each case, @var{cond} can be any of the 16 floating-point conditions:
13078 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13079 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13080 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13083 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13084 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13085 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13086 @code{bc1t}/@code{bc1f}).
13088 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13089 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13094 if (__builtin_mips_cabs_eq_s (a, b))
13100 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13101 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13102 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13103 @code{bc1t}/@code{bc1f}).
13105 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13106 and return either the upper or lower half of the result. For example:
13110 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13111 upper_halves_are_equal ();
13113 upper_halves_are_unequal ();
13115 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13116 lower_halves_are_equal ();
13118 lower_halves_are_unequal ();
13121 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13122 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13123 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13124 @code{movt.ps}/@code{movf.ps}).
13126 The @code{movt} functions return the value @var{x} computed by:
13129 cabs.@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_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13138 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13139 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13140 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13141 Comparison of two paired-single values
13142 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13143 @code{bc1any2t}/@code{bc1any2f}).
13145 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13146 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13147 result is true and the @code{all} forms return true if both results are true.
13152 if (__builtin_mips_any_c_eq_ps (a, b))
13157 if (__builtin_mips_all_c_eq_ps (a, b))
13163 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13164 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13165 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13166 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13167 Comparison of four paired-single values
13168 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13169 @code{bc1any4t}/@code{bc1any4f}).
13171 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13172 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13173 The @code{any} forms return true if any of the four results are true
13174 and the @code{all} forms return true if all four results are true.
13179 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13184 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13191 @node Other MIPS Built-in Functions
13192 @subsection Other MIPS Built-in Functions
13194 GCC provides other MIPS-specific built-in functions:
13197 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13198 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13199 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13200 when this function is available.
13202 @item unsigned int __builtin_mips_get_fcsr (void)
13203 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13204 Get and set the contents of the floating-point control and status register
13205 (FPU control register 31). These functions are only available in hard-float
13206 code but can be called in both MIPS16 and non-MIPS16 contexts.
13208 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13209 register except the condition codes, which GCC assumes are preserved.
13212 @node MSP430 Built-in Functions
13213 @subsection MSP430 Built-in Functions
13215 GCC provides a couple of special builtin functions to aid in the
13216 writing of interrupt handlers in C.
13219 @item __bic_SR_register_on_exit (int @var{mask})
13220 This clears the indicated bits in the saved copy of the status register
13221 currently residing on the stack. This only works inside interrupt
13222 handlers and the changes to the status register will only take affect
13223 once the handler returns.
13225 @item __bis_SR_register_on_exit (int @var{mask})
13226 This sets the indicated bits in the saved copy of the status register
13227 currently residing on the stack. This only works inside interrupt
13228 handlers and the changes to the status register will only take affect
13229 once the handler returns.
13231 @item __delay_cycles (long long @var{cycles})
13232 This inserts an instruction sequence that takes exactly @var{cycles}
13233 cycles (between 0 and about 17E9) to complete. The inserted sequence
13234 may use jumps, loops, or no-ops, and does not interfere with any other
13235 instructions. Note that @var{cycles} must be a compile-time constant
13236 integer - that is, you must pass a number, not a variable that may be
13237 optimized to a constant later. The number of cycles delayed by this
13241 @node NDS32 Built-in Functions
13242 @subsection NDS32 Built-in Functions
13244 These built-in functions are available for the NDS32 target:
13246 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13247 Insert an ISYNC instruction into the instruction stream where
13248 @var{addr} is an instruction address for serialization.
13251 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13252 Insert an ISB instruction into the instruction stream.
13255 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13256 Return the content of a system register which is mapped by @var{sr}.
13259 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13260 Return the content of a user space register which is mapped by @var{usr}.
13263 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13264 Move the @var{value} to a system register which is mapped by @var{sr}.
13267 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13268 Move the @var{value} to a user space register which is mapped by @var{usr}.
13271 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13272 Enable global interrupt.
13275 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13276 Disable global interrupt.
13279 @node picoChip Built-in Functions
13280 @subsection picoChip Built-in Functions
13282 GCC provides an interface to selected machine instructions from the
13283 picoChip instruction set.
13286 @item int __builtin_sbc (int @var{value})
13287 Sign bit count. Return the number of consecutive bits in @var{value}
13288 that have the same value as the sign bit. The result is the number of
13289 leading sign bits minus one, giving the number of redundant sign bits in
13292 @item int __builtin_byteswap (int @var{value})
13293 Byte swap. Return the result of swapping the upper and lower bytes of
13296 @item int __builtin_brev (int @var{value})
13297 Bit reversal. Return the result of reversing the bits in
13298 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13301 @item int __builtin_adds (int @var{x}, int @var{y})
13302 Saturating addition. Return the result of adding @var{x} and @var{y},
13303 storing the value 32767 if the result overflows.
13305 @item int __builtin_subs (int @var{x}, int @var{y})
13306 Saturating subtraction. Return the result of subtracting @var{y} from
13307 @var{x}, storing the value @minus{}32768 if the result overflows.
13309 @item void __builtin_halt (void)
13310 Halt. The processor stops execution. This built-in is useful for
13311 implementing assertions.
13315 @node PowerPC Built-in Functions
13316 @subsection PowerPC Built-in Functions
13318 These built-in functions are available for the PowerPC family of
13321 float __builtin_recipdivf (float, float);
13322 float __builtin_rsqrtf (float);
13323 double __builtin_recipdiv (double, double);
13324 double __builtin_rsqrt (double);
13325 uint64_t __builtin_ppc_get_timebase ();
13326 unsigned long __builtin_ppc_mftb ();
13327 double __builtin_unpack_longdouble (long double, int);
13328 long double __builtin_pack_longdouble (double, double);
13331 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13332 @code{__builtin_rsqrtf} functions generate multiple instructions to
13333 implement the reciprocal sqrt functionality using reciprocal sqrt
13334 estimate instructions.
13336 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13337 functions generate multiple instructions to implement division using
13338 the reciprocal estimate instructions.
13340 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13341 functions generate instructions to read the Time Base Register. The
13342 @code{__builtin_ppc_get_timebase} function may generate multiple
13343 instructions and always returns the 64 bits of the Time Base Register.
13344 The @code{__builtin_ppc_mftb} function always generates one instruction and
13345 returns the Time Base Register value as an unsigned long, throwing away
13346 the most significant word on 32-bit environments.
13348 The following built-in functions are available for the PowerPC family
13349 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13350 or @option{-mpopcntd}):
13352 long __builtin_bpermd (long, long);
13353 int __builtin_divwe (int, int);
13354 int __builtin_divweo (int, int);
13355 unsigned int __builtin_divweu (unsigned int, unsigned int);
13356 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13357 long __builtin_divde (long, long);
13358 long __builtin_divdeo (long, long);
13359 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13360 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13361 unsigned int cdtbcd (unsigned int);
13362 unsigned int cbcdtd (unsigned int);
13363 unsigned int addg6s (unsigned int, unsigned int);
13366 The @code{__builtin_divde}, @code{__builtin_divdeo},
13367 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13368 64-bit environment support ISA 2.06 or later.
13370 The following built-in functions are available for the PowerPC family
13371 of processors when hardware decimal floating point
13372 (@option{-mhard-dfp}) is available:
13374 _Decimal64 __builtin_dxex (_Decimal64);
13375 _Decimal128 __builtin_dxexq (_Decimal128);
13376 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13377 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13378 _Decimal64 __builtin_denbcd (int, _Decimal64);
13379 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13380 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13381 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13382 _Decimal64 __builtin_dscli (_Decimal64, int);
13383 _Decimal128 __builtin_dscliq (_Decimal128, int);
13384 _Decimal64 __builtin_dscri (_Decimal64, int);
13385 _Decimal128 __builtin_dscriq (_Decimal128, int);
13386 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13387 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13390 The following built-in functions are available for the PowerPC family
13391 of processors when the Vector Scalar (vsx) instruction set is
13394 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13395 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13396 unsigned long long);
13399 @node PowerPC AltiVec/VSX Built-in Functions
13400 @subsection PowerPC AltiVec Built-in Functions
13402 GCC provides an interface for the PowerPC family of processors to access
13403 the AltiVec operations described in Motorola's AltiVec Programming
13404 Interface Manual. The interface is made available by including
13405 @code{<altivec.h>} and using @option{-maltivec} and
13406 @option{-mabi=altivec}. The interface supports the following vector
13410 vector unsigned char
13414 vector unsigned short
13415 vector signed short
13419 vector unsigned int
13425 If @option{-mvsx} is used the following additional vector types are
13429 vector unsigned long
13434 The long types are only implemented for 64-bit code generation, and
13435 the long type is only used in the floating point/integer conversion
13438 GCC's implementation of the high-level language interface available from
13439 C and C++ code differs from Motorola's documentation in several ways.
13444 A vector constant is a list of constant expressions within curly braces.
13447 A vector initializer requires no cast if the vector constant is of the
13448 same type as the variable it is initializing.
13451 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13452 vector type is the default signedness of the base type. The default
13453 varies depending on the operating system, so a portable program should
13454 always specify the signedness.
13457 Compiling with @option{-maltivec} adds keywords @code{__vector},
13458 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13459 @code{bool}. When compiling ISO C, the context-sensitive substitution
13460 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13461 disabled. To use them, you must include @code{<altivec.h>} instead.
13464 GCC allows using a @code{typedef} name as the type specifier for a
13468 For C, overloaded functions are implemented with macros so the following
13472 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13476 Since @code{vec_add} is a macro, the vector constant in the example
13477 is treated as four separate arguments. Wrap the entire argument in
13478 parentheses for this to work.
13481 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13482 Internally, GCC uses built-in functions to achieve the functionality in
13483 the aforementioned header file, but they are not supported and are
13484 subject to change without notice.
13486 The following interfaces are supported for the generic and specific
13487 AltiVec operations and the AltiVec predicates. In cases where there
13488 is a direct mapping between generic and specific operations, only the
13489 generic names are shown here, although the specific operations can also
13492 Arguments that are documented as @code{const int} require literal
13493 integral values within the range required for that operation.
13496 vector signed char vec_abs (vector signed char);
13497 vector signed short vec_abs (vector signed short);
13498 vector signed int vec_abs (vector signed int);
13499 vector float vec_abs (vector float);
13501 vector signed char vec_abss (vector signed char);
13502 vector signed short vec_abss (vector signed short);
13503 vector signed int vec_abss (vector signed int);
13505 vector signed char vec_add (vector bool char, vector signed char);
13506 vector signed char vec_add (vector signed char, vector bool char);
13507 vector signed char vec_add (vector signed char, vector signed char);
13508 vector unsigned char vec_add (vector bool char, vector unsigned char);
13509 vector unsigned char vec_add (vector unsigned char, vector bool char);
13510 vector unsigned char vec_add (vector unsigned char,
13511 vector unsigned char);
13512 vector signed short vec_add (vector bool short, vector signed short);
13513 vector signed short vec_add (vector signed short, vector bool short);
13514 vector signed short vec_add (vector signed short, vector signed short);
13515 vector unsigned short vec_add (vector bool short,
13516 vector unsigned short);
13517 vector unsigned short vec_add (vector unsigned short,
13518 vector bool short);
13519 vector unsigned short vec_add (vector unsigned short,
13520 vector unsigned short);
13521 vector signed int vec_add (vector bool int, vector signed int);
13522 vector signed int vec_add (vector signed int, vector bool int);
13523 vector signed int vec_add (vector signed int, vector signed int);
13524 vector unsigned int vec_add (vector bool int, vector unsigned int);
13525 vector unsigned int vec_add (vector unsigned int, vector bool int);
13526 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
13527 vector float vec_add (vector float, vector float);
13529 vector float vec_vaddfp (vector float, vector float);
13531 vector signed int vec_vadduwm (vector bool int, vector signed int);
13532 vector signed int vec_vadduwm (vector signed int, vector bool int);
13533 vector signed int vec_vadduwm (vector signed int, vector signed int);
13534 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
13535 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
13536 vector unsigned int vec_vadduwm (vector unsigned int,
13537 vector unsigned int);
13539 vector signed short vec_vadduhm (vector bool short,
13540 vector signed short);
13541 vector signed short vec_vadduhm (vector signed short,
13542 vector bool short);
13543 vector signed short vec_vadduhm (vector signed short,
13544 vector signed short);
13545 vector unsigned short vec_vadduhm (vector bool short,
13546 vector unsigned short);
13547 vector unsigned short vec_vadduhm (vector unsigned short,
13548 vector bool short);
13549 vector unsigned short vec_vadduhm (vector unsigned short,
13550 vector unsigned short);
13552 vector signed char vec_vaddubm (vector bool char, vector signed char);
13553 vector signed char vec_vaddubm (vector signed char, vector bool char);
13554 vector signed char vec_vaddubm (vector signed char, vector signed char);
13555 vector unsigned char vec_vaddubm (vector bool char,
13556 vector unsigned char);
13557 vector unsigned char vec_vaddubm (vector unsigned char,
13559 vector unsigned char vec_vaddubm (vector unsigned char,
13560 vector unsigned char);
13562 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
13564 vector unsigned char vec_adds (vector bool char, vector unsigned char);
13565 vector unsigned char vec_adds (vector unsigned char, vector bool char);
13566 vector unsigned char vec_adds (vector unsigned char,
13567 vector unsigned char);
13568 vector signed char vec_adds (vector bool char, vector signed char);
13569 vector signed char vec_adds (vector signed char, vector bool char);
13570 vector signed char vec_adds (vector signed char, vector signed char);
13571 vector unsigned short vec_adds (vector bool short,
13572 vector unsigned short);
13573 vector unsigned short vec_adds (vector unsigned short,
13574 vector bool short);
13575 vector unsigned short vec_adds (vector unsigned short,
13576 vector unsigned short);
13577 vector signed short vec_adds (vector bool short, vector signed short);
13578 vector signed short vec_adds (vector signed short, vector bool short);
13579 vector signed short vec_adds (vector signed short, vector signed short);
13580 vector unsigned int vec_adds (vector bool int, vector unsigned int);
13581 vector unsigned int vec_adds (vector unsigned int, vector bool int);
13582 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
13583 vector signed int vec_adds (vector bool int, vector signed int);
13584 vector signed int vec_adds (vector signed int, vector bool int);
13585 vector signed int vec_adds (vector signed int, vector signed int);
13587 vector signed int vec_vaddsws (vector bool int, vector signed int);
13588 vector signed int vec_vaddsws (vector signed int, vector bool int);
13589 vector signed int vec_vaddsws (vector signed int, vector signed int);
13591 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
13592 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
13593 vector unsigned int vec_vadduws (vector unsigned int,
13594 vector unsigned int);
13596 vector signed short vec_vaddshs (vector bool short,
13597 vector signed short);
13598 vector signed short vec_vaddshs (vector signed short,
13599 vector bool short);
13600 vector signed short vec_vaddshs (vector signed short,
13601 vector signed short);
13603 vector unsigned short vec_vadduhs (vector bool short,
13604 vector unsigned short);
13605 vector unsigned short vec_vadduhs (vector unsigned short,
13606 vector bool short);
13607 vector unsigned short vec_vadduhs (vector unsigned short,
13608 vector unsigned short);
13610 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13611 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13612 vector signed char vec_vaddsbs (vector signed char, vector signed char);
13614 vector unsigned char vec_vaddubs (vector bool char,
13615 vector unsigned char);
13616 vector unsigned char vec_vaddubs (vector unsigned char,
13618 vector unsigned char vec_vaddubs (vector unsigned char,
13619 vector unsigned char);
13621 vector float vec_and (vector float, vector float);
13622 vector float vec_and (vector float, vector bool int);
13623 vector float vec_and (vector bool int, vector float);
13624 vector bool int vec_and (vector bool int, vector bool int);
13625 vector signed int vec_and (vector bool int, vector signed int);
13626 vector signed int vec_and (vector signed int, vector bool int);
13627 vector signed int vec_and (vector signed int, vector signed int);
13628 vector unsigned int vec_and (vector bool int, vector unsigned int);
13629 vector unsigned int vec_and (vector unsigned int, vector bool int);
13630 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
13631 vector bool short vec_and (vector bool short, vector bool short);
13632 vector signed short vec_and (vector bool short, vector signed short);
13633 vector signed short vec_and (vector signed short, vector bool short);
13634 vector signed short vec_and (vector signed short, vector signed short);
13635 vector unsigned short vec_and (vector bool short,
13636 vector unsigned short);
13637 vector unsigned short vec_and (vector unsigned short,
13638 vector bool short);
13639 vector unsigned short vec_and (vector unsigned short,
13640 vector unsigned short);
13641 vector signed char vec_and (vector bool char, vector signed char);
13642 vector bool char vec_and (vector bool char, vector bool char);
13643 vector signed char vec_and (vector signed char, vector bool char);
13644 vector signed char vec_and (vector signed char, vector signed char);
13645 vector unsigned char vec_and (vector bool char, vector unsigned char);
13646 vector unsigned char vec_and (vector unsigned char, vector bool char);
13647 vector unsigned char vec_and (vector unsigned char,
13648 vector unsigned char);
13650 vector float vec_andc (vector float, vector float);
13651 vector float vec_andc (vector float, vector bool int);
13652 vector float vec_andc (vector bool int, vector float);
13653 vector bool int vec_andc (vector bool int, vector bool int);
13654 vector signed int vec_andc (vector bool int, vector signed int);
13655 vector signed int vec_andc (vector signed int, vector bool int);
13656 vector signed int vec_andc (vector signed int, vector signed int);
13657 vector unsigned int vec_andc (vector bool int, vector unsigned int);
13658 vector unsigned int vec_andc (vector unsigned int, vector bool int);
13659 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
13660 vector bool short vec_andc (vector bool short, vector bool short);
13661 vector signed short vec_andc (vector bool short, vector signed short);
13662 vector signed short vec_andc (vector signed short, vector bool short);
13663 vector signed short vec_andc (vector signed short, vector signed short);
13664 vector unsigned short vec_andc (vector bool short,
13665 vector unsigned short);
13666 vector unsigned short vec_andc (vector unsigned short,
13667 vector bool short);
13668 vector unsigned short vec_andc (vector unsigned short,
13669 vector unsigned short);
13670 vector signed char vec_andc (vector bool char, vector signed char);
13671 vector bool char vec_andc (vector bool char, vector bool char);
13672 vector signed char vec_andc (vector signed char, vector bool char);
13673 vector signed char vec_andc (vector signed char, vector signed char);
13674 vector unsigned char vec_andc (vector bool char, vector unsigned char);
13675 vector unsigned char vec_andc (vector unsigned char, vector bool char);
13676 vector unsigned char vec_andc (vector unsigned char,
13677 vector unsigned char);
13679 vector unsigned char vec_avg (vector unsigned char,
13680 vector unsigned char);
13681 vector signed char vec_avg (vector signed char, vector signed char);
13682 vector unsigned short vec_avg (vector unsigned short,
13683 vector unsigned short);
13684 vector signed short vec_avg (vector signed short, vector signed short);
13685 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
13686 vector signed int vec_avg (vector signed int, vector signed int);
13688 vector signed int vec_vavgsw (vector signed int, vector signed int);
13690 vector unsigned int vec_vavguw (vector unsigned int,
13691 vector unsigned int);
13693 vector signed short vec_vavgsh (vector signed short,
13694 vector signed short);
13696 vector unsigned short vec_vavguh (vector unsigned short,
13697 vector unsigned short);
13699 vector signed char vec_vavgsb (vector signed char, vector signed char);
13701 vector unsigned char vec_vavgub (vector unsigned char,
13702 vector unsigned char);
13704 vector float vec_copysign (vector float);
13706 vector float vec_ceil (vector float);
13708 vector signed int vec_cmpb (vector float, vector float);
13710 vector bool char vec_cmpeq (vector signed char, vector signed char);
13711 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
13712 vector bool short vec_cmpeq (vector signed short, vector signed short);
13713 vector bool short vec_cmpeq (vector unsigned short,
13714 vector unsigned short);
13715 vector bool int vec_cmpeq (vector signed int, vector signed int);
13716 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
13717 vector bool int vec_cmpeq (vector float, vector float);
13719 vector bool int vec_vcmpeqfp (vector float, vector float);
13721 vector bool int vec_vcmpequw (vector signed int, vector signed int);
13722 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
13724 vector bool short vec_vcmpequh (vector signed short,
13725 vector signed short);
13726 vector bool short vec_vcmpequh (vector unsigned short,
13727 vector unsigned short);
13729 vector bool char vec_vcmpequb (vector signed char, vector signed char);
13730 vector bool char vec_vcmpequb (vector unsigned char,
13731 vector unsigned char);
13733 vector bool int vec_cmpge (vector float, vector float);
13735 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
13736 vector bool char vec_cmpgt (vector signed char, vector signed char);
13737 vector bool short vec_cmpgt (vector unsigned short,
13738 vector unsigned short);
13739 vector bool short vec_cmpgt (vector signed short, vector signed short);
13740 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
13741 vector bool int vec_cmpgt (vector signed int, vector signed int);
13742 vector bool int vec_cmpgt (vector float, vector float);
13744 vector bool int vec_vcmpgtfp (vector float, vector float);
13746 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
13748 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
13750 vector bool short vec_vcmpgtsh (vector signed short,
13751 vector signed short);
13753 vector bool short vec_vcmpgtuh (vector unsigned short,
13754 vector unsigned short);
13756 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
13758 vector bool char vec_vcmpgtub (vector unsigned char,
13759 vector unsigned char);
13761 vector bool int vec_cmple (vector float, vector float);
13763 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
13764 vector bool char vec_cmplt (vector signed char, vector signed char);
13765 vector bool short vec_cmplt (vector unsigned short,
13766 vector unsigned short);
13767 vector bool short vec_cmplt (vector signed short, vector signed short);
13768 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
13769 vector bool int vec_cmplt (vector signed int, vector signed int);
13770 vector bool int vec_cmplt (vector float, vector float);
13772 vector float vec_cpsgn (vector float, vector float);
13774 vector float vec_ctf (vector unsigned int, const int);
13775 vector float vec_ctf (vector signed int, const int);
13776 vector double vec_ctf (vector unsigned long, const int);
13777 vector double vec_ctf (vector signed long, const int);
13779 vector float vec_vcfsx (vector signed int, const int);
13781 vector float vec_vcfux (vector unsigned int, const int);
13783 vector signed int vec_cts (vector float, const int);
13784 vector signed long vec_cts (vector double, const int);
13786 vector unsigned int vec_ctu (vector float, const int);
13787 vector unsigned long vec_ctu (vector double, const int);
13789 void vec_dss (const int);
13791 void vec_dssall (void);
13793 void vec_dst (const vector unsigned char *, int, const int);
13794 void vec_dst (const vector signed char *, int, const int);
13795 void vec_dst (const vector bool char *, int, const int);
13796 void vec_dst (const vector unsigned short *, int, const int);
13797 void vec_dst (const vector signed short *, int, const int);
13798 void vec_dst (const vector bool short *, int, const int);
13799 void vec_dst (const vector pixel *, int, const int);
13800 void vec_dst (const vector unsigned int *, int, const int);
13801 void vec_dst (const vector signed int *, int, const int);
13802 void vec_dst (const vector bool int *, int, const int);
13803 void vec_dst (const vector float *, int, const int);
13804 void vec_dst (const unsigned char *, int, const int);
13805 void vec_dst (const signed char *, int, const int);
13806 void vec_dst (const unsigned short *, int, const int);
13807 void vec_dst (const short *, int, const int);
13808 void vec_dst (const unsigned int *, int, const int);
13809 void vec_dst (const int *, int, const int);
13810 void vec_dst (const unsigned long *, int, const int);
13811 void vec_dst (const long *, int, const int);
13812 void vec_dst (const float *, int, const int);
13814 void vec_dstst (const vector unsigned char *, int, const int);
13815 void vec_dstst (const vector signed char *, int, const int);
13816 void vec_dstst (const vector bool char *, int, const int);
13817 void vec_dstst (const vector unsigned short *, int, const int);
13818 void vec_dstst (const vector signed short *, int, const int);
13819 void vec_dstst (const vector bool short *, int, const int);
13820 void vec_dstst (const vector pixel *, int, const int);
13821 void vec_dstst (const vector unsigned int *, int, const int);
13822 void vec_dstst (const vector signed int *, int, const int);
13823 void vec_dstst (const vector bool int *, int, const int);
13824 void vec_dstst (const vector float *, int, const int);
13825 void vec_dstst (const unsigned char *, int, const int);
13826 void vec_dstst (const signed char *, int, const int);
13827 void vec_dstst (const unsigned short *, int, const int);
13828 void vec_dstst (const short *, int, const int);
13829 void vec_dstst (const unsigned int *, int, const int);
13830 void vec_dstst (const int *, int, const int);
13831 void vec_dstst (const unsigned long *, int, const int);
13832 void vec_dstst (const long *, int, const int);
13833 void vec_dstst (const float *, int, const int);
13835 void vec_dststt (const vector unsigned char *, int, const int);
13836 void vec_dststt (const vector signed char *, int, const int);
13837 void vec_dststt (const vector bool char *, int, const int);
13838 void vec_dststt (const vector unsigned short *, int, const int);
13839 void vec_dststt (const vector signed short *, int, const int);
13840 void vec_dststt (const vector bool short *, int, const int);
13841 void vec_dststt (const vector pixel *, int, const int);
13842 void vec_dststt (const vector unsigned int *, int, const int);
13843 void vec_dststt (const vector signed int *, int, const int);
13844 void vec_dststt (const vector bool int *, int, const int);
13845 void vec_dststt (const vector float *, int, const int);
13846 void vec_dststt (const unsigned char *, int, const int);
13847 void vec_dststt (const signed char *, int, const int);
13848 void vec_dststt (const unsigned short *, int, const int);
13849 void vec_dststt (const short *, int, const int);
13850 void vec_dststt (const unsigned int *, int, const int);
13851 void vec_dststt (const int *, int, const int);
13852 void vec_dststt (const unsigned long *, int, const int);
13853 void vec_dststt (const long *, int, const int);
13854 void vec_dststt (const float *, int, const int);
13856 void vec_dstt (const vector unsigned char *, int, const int);
13857 void vec_dstt (const vector signed char *, int, const int);
13858 void vec_dstt (const vector bool char *, int, const int);
13859 void vec_dstt (const vector unsigned short *, int, const int);
13860 void vec_dstt (const vector signed short *, int, const int);
13861 void vec_dstt (const vector bool short *, int, const int);
13862 void vec_dstt (const vector pixel *, int, const int);
13863 void vec_dstt (const vector unsigned int *, int, const int);
13864 void vec_dstt (const vector signed int *, int, const int);
13865 void vec_dstt (const vector bool int *, int, const int);
13866 void vec_dstt (const vector float *, int, const int);
13867 void vec_dstt (const unsigned char *, int, const int);
13868 void vec_dstt (const signed char *, int, const int);
13869 void vec_dstt (const unsigned short *, int, const int);
13870 void vec_dstt (const short *, int, const int);
13871 void vec_dstt (const unsigned int *, int, const int);
13872 void vec_dstt (const int *, int, const int);
13873 void vec_dstt (const unsigned long *, int, const int);
13874 void vec_dstt (const long *, int, const int);
13875 void vec_dstt (const float *, int, const int);
13877 vector float vec_expte (vector float);
13879 vector float vec_floor (vector float);
13881 vector float vec_ld (int, const vector float *);
13882 vector float vec_ld (int, const float *);
13883 vector bool int vec_ld (int, const vector bool int *);
13884 vector signed int vec_ld (int, const vector signed int *);
13885 vector signed int vec_ld (int, const int *);
13886 vector signed int vec_ld (int, const long *);
13887 vector unsigned int vec_ld (int, const vector unsigned int *);
13888 vector unsigned int vec_ld (int, const unsigned int *);
13889 vector unsigned int vec_ld (int, const unsigned long *);
13890 vector bool short vec_ld (int, const vector bool short *);
13891 vector pixel vec_ld (int, const vector pixel *);
13892 vector signed short vec_ld (int, const vector signed short *);
13893 vector signed short vec_ld (int, const short *);
13894 vector unsigned short vec_ld (int, const vector unsigned short *);
13895 vector unsigned short vec_ld (int, const unsigned short *);
13896 vector bool char vec_ld (int, const vector bool char *);
13897 vector signed char vec_ld (int, const vector signed char *);
13898 vector signed char vec_ld (int, const signed char *);
13899 vector unsigned char vec_ld (int, const vector unsigned char *);
13900 vector unsigned char vec_ld (int, const unsigned char *);
13902 vector signed char vec_lde (int, const signed char *);
13903 vector unsigned char vec_lde (int, const unsigned char *);
13904 vector signed short vec_lde (int, const short *);
13905 vector unsigned short vec_lde (int, const unsigned short *);
13906 vector float vec_lde (int, const float *);
13907 vector signed int vec_lde (int, const int *);
13908 vector unsigned int vec_lde (int, const unsigned int *);
13909 vector signed int vec_lde (int, const long *);
13910 vector unsigned int vec_lde (int, const unsigned long *);
13912 vector float vec_lvewx (int, float *);
13913 vector signed int vec_lvewx (int, int *);
13914 vector unsigned int vec_lvewx (int, unsigned int *);
13915 vector signed int vec_lvewx (int, long *);
13916 vector unsigned int vec_lvewx (int, unsigned long *);
13918 vector signed short vec_lvehx (int, short *);
13919 vector unsigned short vec_lvehx (int, unsigned short *);
13921 vector signed char vec_lvebx (int, char *);
13922 vector unsigned char vec_lvebx (int, unsigned char *);
13924 vector float vec_ldl (int, const vector float *);
13925 vector float vec_ldl (int, const float *);
13926 vector bool int vec_ldl (int, const vector bool int *);
13927 vector signed int vec_ldl (int, const vector signed int *);
13928 vector signed int vec_ldl (int, const int *);
13929 vector signed int vec_ldl (int, const long *);
13930 vector unsigned int vec_ldl (int, const vector unsigned int *);
13931 vector unsigned int vec_ldl (int, const unsigned int *);
13932 vector unsigned int vec_ldl (int, const unsigned long *);
13933 vector bool short vec_ldl (int, const vector bool short *);
13934 vector pixel vec_ldl (int, const vector pixel *);
13935 vector signed short vec_ldl (int, const vector signed short *);
13936 vector signed short vec_ldl (int, const short *);
13937 vector unsigned short vec_ldl (int, const vector unsigned short *);
13938 vector unsigned short vec_ldl (int, const unsigned short *);
13939 vector bool char vec_ldl (int, const vector bool char *);
13940 vector signed char vec_ldl (int, const vector signed char *);
13941 vector signed char vec_ldl (int, const signed char *);
13942 vector unsigned char vec_ldl (int, const vector unsigned char *);
13943 vector unsigned char vec_ldl (int, const unsigned char *);
13945 vector float vec_loge (vector float);
13947 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
13948 vector unsigned char vec_lvsl (int, const volatile signed char *);
13949 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
13950 vector unsigned char vec_lvsl (int, const volatile short *);
13951 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
13952 vector unsigned char vec_lvsl (int, const volatile int *);
13953 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
13954 vector unsigned char vec_lvsl (int, const volatile long *);
13955 vector unsigned char vec_lvsl (int, const volatile float *);
13957 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
13958 vector unsigned char vec_lvsr (int, const volatile signed char *);
13959 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
13960 vector unsigned char vec_lvsr (int, const volatile short *);
13961 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
13962 vector unsigned char vec_lvsr (int, const volatile int *);
13963 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
13964 vector unsigned char vec_lvsr (int, const volatile long *);
13965 vector unsigned char vec_lvsr (int, const volatile float *);
13967 vector float vec_madd (vector float, vector float, vector float);
13969 vector signed short vec_madds (vector signed short,
13970 vector signed short,
13971 vector signed short);
13973 vector unsigned char vec_max (vector bool char, vector unsigned char);
13974 vector unsigned char vec_max (vector unsigned char, vector bool char);
13975 vector unsigned char vec_max (vector unsigned char,
13976 vector unsigned char);
13977 vector signed char vec_max (vector bool char, vector signed char);
13978 vector signed char vec_max (vector signed char, vector bool char);
13979 vector signed char vec_max (vector signed char, vector signed char);
13980 vector unsigned short vec_max (vector bool short,
13981 vector unsigned short);
13982 vector unsigned short vec_max (vector unsigned short,
13983 vector bool short);
13984 vector unsigned short vec_max (vector unsigned short,
13985 vector unsigned short);
13986 vector signed short vec_max (vector bool short, vector signed short);
13987 vector signed short vec_max (vector signed short, vector bool short);
13988 vector signed short vec_max (vector signed short, vector signed short);
13989 vector unsigned int vec_max (vector bool int, vector unsigned int);
13990 vector unsigned int vec_max (vector unsigned int, vector bool int);
13991 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
13992 vector signed int vec_max (vector bool int, vector signed int);
13993 vector signed int vec_max (vector signed int, vector bool int);
13994 vector signed int vec_max (vector signed int, vector signed int);
13995 vector float vec_max (vector float, vector float);
13997 vector float vec_vmaxfp (vector float, vector float);
13999 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14000 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14001 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14003 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14004 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14005 vector unsigned int vec_vmaxuw (vector unsigned int,
14006 vector unsigned int);
14008 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14009 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14010 vector signed short vec_vmaxsh (vector signed short,
14011 vector signed short);
14013 vector unsigned short vec_vmaxuh (vector bool short,
14014 vector unsigned short);
14015 vector unsigned short vec_vmaxuh (vector unsigned short,
14016 vector bool short);
14017 vector unsigned short vec_vmaxuh (vector unsigned short,
14018 vector unsigned short);
14020 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14021 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14022 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14024 vector unsigned char vec_vmaxub (vector bool char,
14025 vector unsigned char);
14026 vector unsigned char vec_vmaxub (vector unsigned char,
14028 vector unsigned char vec_vmaxub (vector unsigned char,
14029 vector unsigned char);
14031 vector bool char vec_mergeh (vector bool char, vector bool char);
14032 vector signed char vec_mergeh (vector signed char, vector signed char);
14033 vector unsigned char vec_mergeh (vector unsigned char,
14034 vector unsigned char);
14035 vector bool short vec_mergeh (vector bool short, vector bool short);
14036 vector pixel vec_mergeh (vector pixel, vector pixel);
14037 vector signed short vec_mergeh (vector signed short,
14038 vector signed short);
14039 vector unsigned short vec_mergeh (vector unsigned short,
14040 vector unsigned short);
14041 vector float vec_mergeh (vector float, vector float);
14042 vector bool int vec_mergeh (vector bool int, vector bool int);
14043 vector signed int vec_mergeh (vector signed int, vector signed int);
14044 vector unsigned int vec_mergeh (vector unsigned int,
14045 vector unsigned int);
14047 vector float vec_vmrghw (vector float, vector float);
14048 vector bool int vec_vmrghw (vector bool int, vector bool int);
14049 vector signed int vec_vmrghw (vector signed int, vector signed int);
14050 vector unsigned int vec_vmrghw (vector unsigned int,
14051 vector unsigned int);
14053 vector bool short vec_vmrghh (vector bool short, vector bool short);
14054 vector signed short vec_vmrghh (vector signed short,
14055 vector signed short);
14056 vector unsigned short vec_vmrghh (vector unsigned short,
14057 vector unsigned short);
14058 vector pixel vec_vmrghh (vector pixel, vector pixel);
14060 vector bool char vec_vmrghb (vector bool char, vector bool char);
14061 vector signed char vec_vmrghb (vector signed char, vector signed char);
14062 vector unsigned char vec_vmrghb (vector unsigned char,
14063 vector unsigned char);
14065 vector bool char vec_mergel (vector bool char, vector bool char);
14066 vector signed char vec_mergel (vector signed char, vector signed char);
14067 vector unsigned char vec_mergel (vector unsigned char,
14068 vector unsigned char);
14069 vector bool short vec_mergel (vector bool short, vector bool short);
14070 vector pixel vec_mergel (vector pixel, vector pixel);
14071 vector signed short vec_mergel (vector signed short,
14072 vector signed short);
14073 vector unsigned short vec_mergel (vector unsigned short,
14074 vector unsigned short);
14075 vector float vec_mergel (vector float, vector float);
14076 vector bool int vec_mergel (vector bool int, vector bool int);
14077 vector signed int vec_mergel (vector signed int, vector signed int);
14078 vector unsigned int vec_mergel (vector unsigned int,
14079 vector unsigned int);
14081 vector float vec_vmrglw (vector float, vector float);
14082 vector signed int vec_vmrglw (vector signed int, vector signed int);
14083 vector unsigned int vec_vmrglw (vector unsigned int,
14084 vector unsigned int);
14085 vector bool int vec_vmrglw (vector bool int, vector bool int);
14087 vector bool short vec_vmrglh (vector bool short, vector bool short);
14088 vector signed short vec_vmrglh (vector signed short,
14089 vector signed short);
14090 vector unsigned short vec_vmrglh (vector unsigned short,
14091 vector unsigned short);
14092 vector pixel vec_vmrglh (vector pixel, vector pixel);
14094 vector bool char vec_vmrglb (vector bool char, vector bool char);
14095 vector signed char vec_vmrglb (vector signed char, vector signed char);
14096 vector unsigned char vec_vmrglb (vector unsigned char,
14097 vector unsigned char);
14099 vector unsigned short vec_mfvscr (void);
14101 vector unsigned char vec_min (vector bool char, vector unsigned char);
14102 vector unsigned char vec_min (vector unsigned char, vector bool char);
14103 vector unsigned char vec_min (vector unsigned char,
14104 vector unsigned char);
14105 vector signed char vec_min (vector bool char, vector signed char);
14106 vector signed char vec_min (vector signed char, vector bool char);
14107 vector signed char vec_min (vector signed char, vector signed char);
14108 vector unsigned short vec_min (vector bool short,
14109 vector unsigned short);
14110 vector unsigned short vec_min (vector unsigned short,
14111 vector bool short);
14112 vector unsigned short vec_min (vector unsigned short,
14113 vector unsigned short);
14114 vector signed short vec_min (vector bool short, vector signed short);
14115 vector signed short vec_min (vector signed short, vector bool short);
14116 vector signed short vec_min (vector signed short, vector signed short);
14117 vector unsigned int vec_min (vector bool int, vector unsigned int);
14118 vector unsigned int vec_min (vector unsigned int, vector bool int);
14119 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14120 vector signed int vec_min (vector bool int, vector signed int);
14121 vector signed int vec_min (vector signed int, vector bool int);
14122 vector signed int vec_min (vector signed int, vector signed int);
14123 vector float vec_min (vector float, vector float);
14125 vector float vec_vminfp (vector float, vector float);
14127 vector signed int vec_vminsw (vector bool int, vector signed int);
14128 vector signed int vec_vminsw (vector signed int, vector bool int);
14129 vector signed int vec_vminsw (vector signed int, vector signed int);
14131 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14132 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14133 vector unsigned int vec_vminuw (vector unsigned int,
14134 vector unsigned int);
14136 vector signed short vec_vminsh (vector bool short, vector signed short);
14137 vector signed short vec_vminsh (vector signed short, vector bool short);
14138 vector signed short vec_vminsh (vector signed short,
14139 vector signed short);
14141 vector unsigned short vec_vminuh (vector bool short,
14142 vector unsigned short);
14143 vector unsigned short vec_vminuh (vector unsigned short,
14144 vector bool short);
14145 vector unsigned short vec_vminuh (vector unsigned short,
14146 vector unsigned short);
14148 vector signed char vec_vminsb (vector bool char, vector signed char);
14149 vector signed char vec_vminsb (vector signed char, vector bool char);
14150 vector signed char vec_vminsb (vector signed char, vector signed char);
14152 vector unsigned char vec_vminub (vector bool char,
14153 vector unsigned char);
14154 vector unsigned char vec_vminub (vector unsigned char,
14156 vector unsigned char vec_vminub (vector unsigned char,
14157 vector unsigned char);
14159 vector signed short vec_mladd (vector signed short,
14160 vector signed short,
14161 vector signed short);
14162 vector signed short vec_mladd (vector signed short,
14163 vector unsigned short,
14164 vector unsigned short);
14165 vector signed short vec_mladd (vector unsigned short,
14166 vector signed short,
14167 vector signed short);
14168 vector unsigned short vec_mladd (vector unsigned short,
14169 vector unsigned short,
14170 vector unsigned short);
14172 vector signed short vec_mradds (vector signed short,
14173 vector signed short,
14174 vector signed short);
14176 vector unsigned int vec_msum (vector unsigned char,
14177 vector unsigned char,
14178 vector unsigned int);
14179 vector signed int vec_msum (vector signed char,
14180 vector unsigned char,
14181 vector signed int);
14182 vector unsigned int vec_msum (vector unsigned short,
14183 vector unsigned short,
14184 vector unsigned int);
14185 vector signed int vec_msum (vector signed short,
14186 vector signed short,
14187 vector signed int);
14189 vector signed int vec_vmsumshm (vector signed short,
14190 vector signed short,
14191 vector signed int);
14193 vector unsigned int vec_vmsumuhm (vector unsigned short,
14194 vector unsigned short,
14195 vector unsigned int);
14197 vector signed int vec_vmsummbm (vector signed char,
14198 vector unsigned char,
14199 vector signed int);
14201 vector unsigned int vec_vmsumubm (vector unsigned char,
14202 vector unsigned char,
14203 vector unsigned int);
14205 vector unsigned int vec_msums (vector unsigned short,
14206 vector unsigned short,
14207 vector unsigned int);
14208 vector signed int vec_msums (vector signed short,
14209 vector signed short,
14210 vector signed int);
14212 vector signed int vec_vmsumshs (vector signed short,
14213 vector signed short,
14214 vector signed int);
14216 vector unsigned int vec_vmsumuhs (vector unsigned short,
14217 vector unsigned short,
14218 vector unsigned int);
14220 void vec_mtvscr (vector signed int);
14221 void vec_mtvscr (vector unsigned int);
14222 void vec_mtvscr (vector bool int);
14223 void vec_mtvscr (vector signed short);
14224 void vec_mtvscr (vector unsigned short);
14225 void vec_mtvscr (vector bool short);
14226 void vec_mtvscr (vector pixel);
14227 void vec_mtvscr (vector signed char);
14228 void vec_mtvscr (vector unsigned char);
14229 void vec_mtvscr (vector bool char);
14231 vector unsigned short vec_mule (vector unsigned char,
14232 vector unsigned char);
14233 vector signed short vec_mule (vector signed char,
14234 vector signed char);
14235 vector unsigned int vec_mule (vector unsigned short,
14236 vector unsigned short);
14237 vector signed int vec_mule (vector signed short, vector signed short);
14239 vector signed int vec_vmulesh (vector signed short,
14240 vector signed short);
14242 vector unsigned int vec_vmuleuh (vector unsigned short,
14243 vector unsigned short);
14245 vector signed short vec_vmulesb (vector signed char,
14246 vector signed char);
14248 vector unsigned short vec_vmuleub (vector unsigned char,
14249 vector unsigned char);
14251 vector unsigned short vec_mulo (vector unsigned char,
14252 vector unsigned char);
14253 vector signed short vec_mulo (vector signed char, vector signed char);
14254 vector unsigned int vec_mulo (vector unsigned short,
14255 vector unsigned short);
14256 vector signed int vec_mulo (vector signed short, vector signed short);
14258 vector signed int vec_vmulosh (vector signed short,
14259 vector signed short);
14261 vector unsigned int vec_vmulouh (vector unsigned short,
14262 vector unsigned short);
14264 vector signed short vec_vmulosb (vector signed char,
14265 vector signed char);
14267 vector unsigned short vec_vmuloub (vector unsigned char,
14268 vector unsigned char);
14270 vector float vec_nmsub (vector float, vector float, vector float);
14272 vector float vec_nor (vector float, vector float);
14273 vector signed int vec_nor (vector signed int, vector signed int);
14274 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14275 vector bool int vec_nor (vector bool int, vector bool int);
14276 vector signed short vec_nor (vector signed short, vector signed short);
14277 vector unsigned short vec_nor (vector unsigned short,
14278 vector unsigned short);
14279 vector bool short vec_nor (vector bool short, vector bool short);
14280 vector signed char vec_nor (vector signed char, vector signed char);
14281 vector unsigned char vec_nor (vector unsigned char,
14282 vector unsigned char);
14283 vector bool char vec_nor (vector bool char, vector bool char);
14285 vector float vec_or (vector float, vector float);
14286 vector float vec_or (vector float, vector bool int);
14287 vector float vec_or (vector bool int, vector float);
14288 vector bool int vec_or (vector bool int, vector bool int);
14289 vector signed int vec_or (vector bool int, vector signed int);
14290 vector signed int vec_or (vector signed int, vector bool int);
14291 vector signed int vec_or (vector signed int, vector signed int);
14292 vector unsigned int vec_or (vector bool int, vector unsigned int);
14293 vector unsigned int vec_or (vector unsigned int, vector bool int);
14294 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14295 vector bool short vec_or (vector bool short, vector bool short);
14296 vector signed short vec_or (vector bool short, vector signed short);
14297 vector signed short vec_or (vector signed short, vector bool short);
14298 vector signed short vec_or (vector signed short, vector signed short);
14299 vector unsigned short vec_or (vector bool short, vector unsigned short);
14300 vector unsigned short vec_or (vector unsigned short, vector bool short);
14301 vector unsigned short vec_or (vector unsigned short,
14302 vector unsigned short);
14303 vector signed char vec_or (vector bool char, vector signed char);
14304 vector bool char vec_or (vector bool char, vector bool char);
14305 vector signed char vec_or (vector signed char, vector bool char);
14306 vector signed char vec_or (vector signed char, vector signed char);
14307 vector unsigned char vec_or (vector bool char, vector unsigned char);
14308 vector unsigned char vec_or (vector unsigned char, vector bool char);
14309 vector unsigned char vec_or (vector unsigned char,
14310 vector unsigned char);
14312 vector signed char vec_pack (vector signed short, vector signed short);
14313 vector unsigned char vec_pack (vector unsigned short,
14314 vector unsigned short);
14315 vector bool char vec_pack (vector bool short, vector bool short);
14316 vector signed short vec_pack (vector signed int, vector signed int);
14317 vector unsigned short vec_pack (vector unsigned int,
14318 vector unsigned int);
14319 vector bool short vec_pack (vector bool int, vector bool int);
14321 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14322 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14323 vector unsigned short vec_vpkuwum (vector unsigned int,
14324 vector unsigned int);
14326 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14327 vector signed char vec_vpkuhum (vector signed short,
14328 vector signed short);
14329 vector unsigned char vec_vpkuhum (vector unsigned short,
14330 vector unsigned short);
14332 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14334 vector unsigned char vec_packs (vector unsigned short,
14335 vector unsigned short);
14336 vector signed char vec_packs (vector signed short, vector signed short);
14337 vector unsigned short vec_packs (vector unsigned int,
14338 vector unsigned int);
14339 vector signed short vec_packs (vector signed int, vector signed int);
14341 vector signed short vec_vpkswss (vector signed int, vector signed int);
14343 vector unsigned short vec_vpkuwus (vector unsigned int,
14344 vector unsigned int);
14346 vector signed char vec_vpkshss (vector signed short,
14347 vector signed short);
14349 vector unsigned char vec_vpkuhus (vector unsigned short,
14350 vector unsigned short);
14352 vector unsigned char vec_packsu (vector unsigned short,
14353 vector unsigned short);
14354 vector unsigned char vec_packsu (vector signed short,
14355 vector signed short);
14356 vector unsigned short vec_packsu (vector unsigned int,
14357 vector unsigned int);
14358 vector unsigned short vec_packsu (vector signed int, vector signed int);
14360 vector unsigned short vec_vpkswus (vector signed int,
14361 vector signed int);
14363 vector unsigned char vec_vpkshus (vector signed short,
14364 vector signed short);
14366 vector float vec_perm (vector float,
14368 vector unsigned char);
14369 vector signed int vec_perm (vector signed int,
14371 vector unsigned char);
14372 vector unsigned int vec_perm (vector unsigned int,
14373 vector unsigned int,
14374 vector unsigned char);
14375 vector bool int vec_perm (vector bool int,
14377 vector unsigned char);
14378 vector signed short vec_perm (vector signed short,
14379 vector signed short,
14380 vector unsigned char);
14381 vector unsigned short vec_perm (vector unsigned short,
14382 vector unsigned short,
14383 vector unsigned char);
14384 vector bool short vec_perm (vector bool short,
14386 vector unsigned char);
14387 vector pixel vec_perm (vector pixel,
14389 vector unsigned char);
14390 vector signed char vec_perm (vector signed char,
14391 vector signed char,
14392 vector unsigned char);
14393 vector unsigned char vec_perm (vector unsigned char,
14394 vector unsigned char,
14395 vector unsigned char);
14396 vector bool char vec_perm (vector bool char,
14398 vector unsigned char);
14400 vector float vec_re (vector float);
14402 vector signed char vec_rl (vector signed char,
14403 vector unsigned char);
14404 vector unsigned char vec_rl (vector unsigned char,
14405 vector unsigned char);
14406 vector signed short vec_rl (vector signed short, vector unsigned short);
14407 vector unsigned short vec_rl (vector unsigned short,
14408 vector unsigned short);
14409 vector signed int vec_rl (vector signed int, vector unsigned int);
14410 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14412 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14413 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14415 vector signed short vec_vrlh (vector signed short,
14416 vector unsigned short);
14417 vector unsigned short vec_vrlh (vector unsigned short,
14418 vector unsigned short);
14420 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14421 vector unsigned char vec_vrlb (vector unsigned char,
14422 vector unsigned char);
14424 vector float vec_round (vector float);
14426 vector float vec_recip (vector float, vector float);
14428 vector float vec_rsqrt (vector float);
14430 vector float vec_rsqrte (vector float);
14432 vector float vec_sel (vector float, vector float, vector bool int);
14433 vector float vec_sel (vector float, vector float, vector unsigned int);
14434 vector signed int vec_sel (vector signed int,
14437 vector signed int vec_sel (vector signed int,
14439 vector unsigned int);
14440 vector unsigned int vec_sel (vector unsigned int,
14441 vector unsigned int,
14443 vector unsigned int vec_sel (vector unsigned int,
14444 vector unsigned int,
14445 vector unsigned int);
14446 vector bool int vec_sel (vector bool int,
14449 vector bool int vec_sel (vector bool int,
14451 vector unsigned int);
14452 vector signed short vec_sel (vector signed short,
14453 vector signed short,
14454 vector bool short);
14455 vector signed short vec_sel (vector signed short,
14456 vector signed short,
14457 vector unsigned short);
14458 vector unsigned short vec_sel (vector unsigned short,
14459 vector unsigned short,
14460 vector bool short);
14461 vector unsigned short vec_sel (vector unsigned short,
14462 vector unsigned short,
14463 vector unsigned short);
14464 vector bool short vec_sel (vector bool short,
14466 vector bool short);
14467 vector bool short vec_sel (vector bool short,
14469 vector unsigned short);
14470 vector signed char vec_sel (vector signed char,
14471 vector signed char,
14473 vector signed char vec_sel (vector signed char,
14474 vector signed char,
14475 vector unsigned char);
14476 vector unsigned char vec_sel (vector unsigned char,
14477 vector unsigned char,
14479 vector unsigned char vec_sel (vector unsigned char,
14480 vector unsigned char,
14481 vector unsigned char);
14482 vector bool char vec_sel (vector bool char,
14485 vector bool char vec_sel (vector bool char,
14487 vector unsigned char);
14489 vector signed char vec_sl (vector signed char,
14490 vector unsigned char);
14491 vector unsigned char vec_sl (vector unsigned char,
14492 vector unsigned char);
14493 vector signed short vec_sl (vector signed short, vector unsigned short);
14494 vector unsigned short vec_sl (vector unsigned short,
14495 vector unsigned short);
14496 vector signed int vec_sl (vector signed int, vector unsigned int);
14497 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14499 vector signed int vec_vslw (vector signed int, vector unsigned int);
14500 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14502 vector signed short vec_vslh (vector signed short,
14503 vector unsigned short);
14504 vector unsigned short vec_vslh (vector unsigned short,
14505 vector unsigned short);
14507 vector signed char vec_vslb (vector signed char, vector unsigned char);
14508 vector unsigned char vec_vslb (vector unsigned char,
14509 vector unsigned char);
14511 vector float vec_sld (vector float, vector float, const int);
14512 vector signed int vec_sld (vector signed int,
14515 vector unsigned int vec_sld (vector unsigned int,
14516 vector unsigned int,
14518 vector bool int vec_sld (vector bool int,
14521 vector signed short vec_sld (vector signed short,
14522 vector signed short,
14524 vector unsigned short vec_sld (vector unsigned short,
14525 vector unsigned short,
14527 vector bool short vec_sld (vector bool short,
14530 vector pixel vec_sld (vector pixel,
14533 vector signed char vec_sld (vector signed char,
14534 vector signed char,
14536 vector unsigned char vec_sld (vector unsigned char,
14537 vector unsigned char,
14539 vector bool char vec_sld (vector bool char,
14543 vector signed int vec_sll (vector signed int,
14544 vector unsigned int);
14545 vector signed int vec_sll (vector signed int,
14546 vector unsigned short);
14547 vector signed int vec_sll (vector signed int,
14548 vector unsigned char);
14549 vector unsigned int vec_sll (vector unsigned int,
14550 vector unsigned int);
14551 vector unsigned int vec_sll (vector unsigned int,
14552 vector unsigned short);
14553 vector unsigned int vec_sll (vector unsigned int,
14554 vector unsigned char);
14555 vector bool int vec_sll (vector bool int,
14556 vector unsigned int);
14557 vector bool int vec_sll (vector bool int,
14558 vector unsigned short);
14559 vector bool int vec_sll (vector bool int,
14560 vector unsigned char);
14561 vector signed short vec_sll (vector signed short,
14562 vector unsigned int);
14563 vector signed short vec_sll (vector signed short,
14564 vector unsigned short);
14565 vector signed short vec_sll (vector signed short,
14566 vector unsigned char);
14567 vector unsigned short vec_sll (vector unsigned short,
14568 vector unsigned int);
14569 vector unsigned short vec_sll (vector unsigned short,
14570 vector unsigned short);
14571 vector unsigned short vec_sll (vector unsigned short,
14572 vector unsigned char);
14573 vector bool short vec_sll (vector bool short, vector unsigned int);
14574 vector bool short vec_sll (vector bool short, vector unsigned short);
14575 vector bool short vec_sll (vector bool short, vector unsigned char);
14576 vector pixel vec_sll (vector pixel, vector unsigned int);
14577 vector pixel vec_sll (vector pixel, vector unsigned short);
14578 vector pixel vec_sll (vector pixel, vector unsigned char);
14579 vector signed char vec_sll (vector signed char, vector unsigned int);
14580 vector signed char vec_sll (vector signed char, vector unsigned short);
14581 vector signed char vec_sll (vector signed char, vector unsigned char);
14582 vector unsigned char vec_sll (vector unsigned char,
14583 vector unsigned int);
14584 vector unsigned char vec_sll (vector unsigned char,
14585 vector unsigned short);
14586 vector unsigned char vec_sll (vector unsigned char,
14587 vector unsigned char);
14588 vector bool char vec_sll (vector bool char, vector unsigned int);
14589 vector bool char vec_sll (vector bool char, vector unsigned short);
14590 vector bool char vec_sll (vector bool char, vector unsigned char);
14592 vector float vec_slo (vector float, vector signed char);
14593 vector float vec_slo (vector float, vector unsigned char);
14594 vector signed int vec_slo (vector signed int, vector signed char);
14595 vector signed int vec_slo (vector signed int, vector unsigned char);
14596 vector unsigned int vec_slo (vector unsigned int, vector signed char);
14597 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
14598 vector signed short vec_slo (vector signed short, vector signed char);
14599 vector signed short vec_slo (vector signed short, vector unsigned char);
14600 vector unsigned short vec_slo (vector unsigned short,
14601 vector signed char);
14602 vector unsigned short vec_slo (vector unsigned short,
14603 vector unsigned char);
14604 vector pixel vec_slo (vector pixel, vector signed char);
14605 vector pixel vec_slo (vector pixel, vector unsigned char);
14606 vector signed char vec_slo (vector signed char, vector signed char);
14607 vector signed char vec_slo (vector signed char, vector unsigned char);
14608 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14609 vector unsigned char vec_slo (vector unsigned char,
14610 vector unsigned char);
14612 vector signed char vec_splat (vector signed char, const int);
14613 vector unsigned char vec_splat (vector unsigned char, const int);
14614 vector bool char vec_splat (vector bool char, const int);
14615 vector signed short vec_splat (vector signed short, const int);
14616 vector unsigned short vec_splat (vector unsigned short, const int);
14617 vector bool short vec_splat (vector bool short, const int);
14618 vector pixel vec_splat (vector pixel, const int);
14619 vector float vec_splat (vector float, const int);
14620 vector signed int vec_splat (vector signed int, const int);
14621 vector unsigned int vec_splat (vector unsigned int, const int);
14622 vector bool int vec_splat (vector bool int, const int);
14623 vector signed long vec_splat (vector signed long, const int);
14624 vector unsigned long vec_splat (vector unsigned long, const int);
14626 vector signed char vec_splats (signed char);
14627 vector unsigned char vec_splats (unsigned char);
14628 vector signed short vec_splats (signed short);
14629 vector unsigned short vec_splats (unsigned short);
14630 vector signed int vec_splats (signed int);
14631 vector unsigned int vec_splats (unsigned int);
14632 vector float vec_splats (float);
14634 vector float vec_vspltw (vector float, const int);
14635 vector signed int vec_vspltw (vector signed int, const int);
14636 vector unsigned int vec_vspltw (vector unsigned int, const int);
14637 vector bool int vec_vspltw (vector bool int, const int);
14639 vector bool short vec_vsplth (vector bool short, const int);
14640 vector signed short vec_vsplth (vector signed short, const int);
14641 vector unsigned short vec_vsplth (vector unsigned short, const int);
14642 vector pixel vec_vsplth (vector pixel, const int);
14644 vector signed char vec_vspltb (vector signed char, const int);
14645 vector unsigned char vec_vspltb (vector unsigned char, const int);
14646 vector bool char vec_vspltb (vector bool char, const int);
14648 vector signed char vec_splat_s8 (const int);
14650 vector signed short vec_splat_s16 (const int);
14652 vector signed int vec_splat_s32 (const int);
14654 vector unsigned char vec_splat_u8 (const int);
14656 vector unsigned short vec_splat_u16 (const int);
14658 vector unsigned int vec_splat_u32 (const int);
14660 vector signed char vec_sr (vector signed char, vector unsigned char);
14661 vector unsigned char vec_sr (vector unsigned char,
14662 vector unsigned char);
14663 vector signed short vec_sr (vector signed short,
14664 vector unsigned short);
14665 vector unsigned short vec_sr (vector unsigned short,
14666 vector unsigned short);
14667 vector signed int vec_sr (vector signed int, vector unsigned int);
14668 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
14670 vector signed int vec_vsrw (vector signed int, vector unsigned int);
14671 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
14673 vector signed short vec_vsrh (vector signed short,
14674 vector unsigned short);
14675 vector unsigned short vec_vsrh (vector unsigned short,
14676 vector unsigned short);
14678 vector signed char vec_vsrb (vector signed char, vector unsigned char);
14679 vector unsigned char vec_vsrb (vector unsigned char,
14680 vector unsigned char);
14682 vector signed char vec_sra (vector signed char, vector unsigned char);
14683 vector unsigned char vec_sra (vector unsigned char,
14684 vector unsigned char);
14685 vector signed short vec_sra (vector signed short,
14686 vector unsigned short);
14687 vector unsigned short vec_sra (vector unsigned short,
14688 vector unsigned short);
14689 vector signed int vec_sra (vector signed int, vector unsigned int);
14690 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
14692 vector signed int vec_vsraw (vector signed int, vector unsigned int);
14693 vector unsigned int vec_vsraw (vector unsigned int,
14694 vector unsigned int);
14696 vector signed short vec_vsrah (vector signed short,
14697 vector unsigned short);
14698 vector unsigned short vec_vsrah (vector unsigned short,
14699 vector unsigned short);
14701 vector signed char vec_vsrab (vector signed char, vector unsigned char);
14702 vector unsigned char vec_vsrab (vector unsigned char,
14703 vector unsigned char);
14705 vector signed int vec_srl (vector signed int, vector unsigned int);
14706 vector signed int vec_srl (vector signed int, vector unsigned short);
14707 vector signed int vec_srl (vector signed int, vector unsigned char);
14708 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
14709 vector unsigned int vec_srl (vector unsigned int,
14710 vector unsigned short);
14711 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
14712 vector bool int vec_srl (vector bool int, vector unsigned int);
14713 vector bool int vec_srl (vector bool int, vector unsigned short);
14714 vector bool int vec_srl (vector bool int, vector unsigned char);
14715 vector signed short vec_srl (vector signed short, vector unsigned int);
14716 vector signed short vec_srl (vector signed short,
14717 vector unsigned short);
14718 vector signed short vec_srl (vector signed short, vector unsigned char);
14719 vector unsigned short vec_srl (vector unsigned short,
14720 vector unsigned int);
14721 vector unsigned short vec_srl (vector unsigned short,
14722 vector unsigned short);
14723 vector unsigned short vec_srl (vector unsigned short,
14724 vector unsigned char);
14725 vector bool short vec_srl (vector bool short, vector unsigned int);
14726 vector bool short vec_srl (vector bool short, vector unsigned short);
14727 vector bool short vec_srl (vector bool short, vector unsigned char);
14728 vector pixel vec_srl (vector pixel, vector unsigned int);
14729 vector pixel vec_srl (vector pixel, vector unsigned short);
14730 vector pixel vec_srl (vector pixel, vector unsigned char);
14731 vector signed char vec_srl (vector signed char, vector unsigned int);
14732 vector signed char vec_srl (vector signed char, vector unsigned short);
14733 vector signed char vec_srl (vector signed char, vector unsigned char);
14734 vector unsigned char vec_srl (vector unsigned char,
14735 vector unsigned int);
14736 vector unsigned char vec_srl (vector unsigned char,
14737 vector unsigned short);
14738 vector unsigned char vec_srl (vector unsigned char,
14739 vector unsigned char);
14740 vector bool char vec_srl (vector bool char, vector unsigned int);
14741 vector bool char vec_srl (vector bool char, vector unsigned short);
14742 vector bool char vec_srl (vector bool char, vector unsigned char);
14744 vector float vec_sro (vector float, vector signed char);
14745 vector float vec_sro (vector float, vector unsigned char);
14746 vector signed int vec_sro (vector signed int, vector signed char);
14747 vector signed int vec_sro (vector signed int, vector unsigned char);
14748 vector unsigned int vec_sro (vector unsigned int, vector signed char);
14749 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
14750 vector signed short vec_sro (vector signed short, vector signed char);
14751 vector signed short vec_sro (vector signed short, vector unsigned char);
14752 vector unsigned short vec_sro (vector unsigned short,
14753 vector signed char);
14754 vector unsigned short vec_sro (vector unsigned short,
14755 vector unsigned char);
14756 vector pixel vec_sro (vector pixel, vector signed char);
14757 vector pixel vec_sro (vector pixel, vector unsigned char);
14758 vector signed char vec_sro (vector signed char, vector signed char);
14759 vector signed char vec_sro (vector signed char, vector unsigned char);
14760 vector unsigned char vec_sro (vector unsigned char, vector signed char);
14761 vector unsigned char vec_sro (vector unsigned char,
14762 vector unsigned char);
14764 void vec_st (vector float, int, vector float *);
14765 void vec_st (vector float, int, float *);
14766 void vec_st (vector signed int, int, vector signed int *);
14767 void vec_st (vector signed int, int, int *);
14768 void vec_st (vector unsigned int, int, vector unsigned int *);
14769 void vec_st (vector unsigned int, int, unsigned int *);
14770 void vec_st (vector bool int, int, vector bool int *);
14771 void vec_st (vector bool int, int, unsigned int *);
14772 void vec_st (vector bool int, int, int *);
14773 void vec_st (vector signed short, int, vector signed short *);
14774 void vec_st (vector signed short, int, short *);
14775 void vec_st (vector unsigned short, int, vector unsigned short *);
14776 void vec_st (vector unsigned short, int, unsigned short *);
14777 void vec_st (vector bool short, int, vector bool short *);
14778 void vec_st (vector bool short, int, unsigned short *);
14779 void vec_st (vector pixel, int, vector pixel *);
14780 void vec_st (vector pixel, int, unsigned short *);
14781 void vec_st (vector pixel, int, short *);
14782 void vec_st (vector bool short, int, short *);
14783 void vec_st (vector signed char, int, vector signed char *);
14784 void vec_st (vector signed char, int, signed char *);
14785 void vec_st (vector unsigned char, int, vector unsigned char *);
14786 void vec_st (vector unsigned char, int, unsigned char *);
14787 void vec_st (vector bool char, int, vector bool char *);
14788 void vec_st (vector bool char, int, unsigned char *);
14789 void vec_st (vector bool char, int, signed char *);
14791 void vec_ste (vector signed char, int, signed char *);
14792 void vec_ste (vector unsigned char, int, unsigned char *);
14793 void vec_ste (vector bool char, int, signed char *);
14794 void vec_ste (vector bool char, int, unsigned char *);
14795 void vec_ste (vector signed short, int, short *);
14796 void vec_ste (vector unsigned short, int, unsigned short *);
14797 void vec_ste (vector bool short, int, short *);
14798 void vec_ste (vector bool short, int, unsigned short *);
14799 void vec_ste (vector pixel, int, short *);
14800 void vec_ste (vector pixel, int, unsigned short *);
14801 void vec_ste (vector float, int, float *);
14802 void vec_ste (vector signed int, int, int *);
14803 void vec_ste (vector unsigned int, int, unsigned int *);
14804 void vec_ste (vector bool int, int, int *);
14805 void vec_ste (vector bool int, int, unsigned int *);
14807 void vec_stvewx (vector float, int, float *);
14808 void vec_stvewx (vector signed int, int, int *);
14809 void vec_stvewx (vector unsigned int, int, unsigned int *);
14810 void vec_stvewx (vector bool int, int, int *);
14811 void vec_stvewx (vector bool int, int, unsigned int *);
14813 void vec_stvehx (vector signed short, int, short *);
14814 void vec_stvehx (vector unsigned short, int, unsigned short *);
14815 void vec_stvehx (vector bool short, int, short *);
14816 void vec_stvehx (vector bool short, int, unsigned short *);
14817 void vec_stvehx (vector pixel, int, short *);
14818 void vec_stvehx (vector pixel, int, unsigned short *);
14820 void vec_stvebx (vector signed char, int, signed char *);
14821 void vec_stvebx (vector unsigned char, int, unsigned char *);
14822 void vec_stvebx (vector bool char, int, signed char *);
14823 void vec_stvebx (vector bool char, int, unsigned char *);
14825 void vec_stl (vector float, int, vector float *);
14826 void vec_stl (vector float, int, float *);
14827 void vec_stl (vector signed int, int, vector signed int *);
14828 void vec_stl (vector signed int, int, int *);
14829 void vec_stl (vector unsigned int, int, vector unsigned int *);
14830 void vec_stl (vector unsigned int, int, unsigned int *);
14831 void vec_stl (vector bool int, int, vector bool int *);
14832 void vec_stl (vector bool int, int, unsigned int *);
14833 void vec_stl (vector bool int, int, int *);
14834 void vec_stl (vector signed short, int, vector signed short *);
14835 void vec_stl (vector signed short, int, short *);
14836 void vec_stl (vector unsigned short, int, vector unsigned short *);
14837 void vec_stl (vector unsigned short, int, unsigned short *);
14838 void vec_stl (vector bool short, int, vector bool short *);
14839 void vec_stl (vector bool short, int, unsigned short *);
14840 void vec_stl (vector bool short, int, short *);
14841 void vec_stl (vector pixel, int, vector pixel *);
14842 void vec_stl (vector pixel, int, unsigned short *);
14843 void vec_stl (vector pixel, int, short *);
14844 void vec_stl (vector signed char, int, vector signed char *);
14845 void vec_stl (vector signed char, int, signed char *);
14846 void vec_stl (vector unsigned char, int, vector unsigned char *);
14847 void vec_stl (vector unsigned char, int, unsigned char *);
14848 void vec_stl (vector bool char, int, vector bool char *);
14849 void vec_stl (vector bool char, int, unsigned char *);
14850 void vec_stl (vector bool char, int, signed char *);
14852 vector signed char vec_sub (vector bool char, vector signed char);
14853 vector signed char vec_sub (vector signed char, vector bool char);
14854 vector signed char vec_sub (vector signed char, vector signed char);
14855 vector unsigned char vec_sub (vector bool char, vector unsigned char);
14856 vector unsigned char vec_sub (vector unsigned char, vector bool char);
14857 vector unsigned char vec_sub (vector unsigned char,
14858 vector unsigned char);
14859 vector signed short vec_sub (vector bool short, vector signed short);
14860 vector signed short vec_sub (vector signed short, vector bool short);
14861 vector signed short vec_sub (vector signed short, vector signed short);
14862 vector unsigned short vec_sub (vector bool short,
14863 vector unsigned short);
14864 vector unsigned short vec_sub (vector unsigned short,
14865 vector bool short);
14866 vector unsigned short vec_sub (vector unsigned short,
14867 vector unsigned short);
14868 vector signed int vec_sub (vector bool int, vector signed int);
14869 vector signed int vec_sub (vector signed int, vector bool int);
14870 vector signed int vec_sub (vector signed int, vector signed int);
14871 vector unsigned int vec_sub (vector bool int, vector unsigned int);
14872 vector unsigned int vec_sub (vector unsigned int, vector bool int);
14873 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
14874 vector float vec_sub (vector float, vector float);
14876 vector float vec_vsubfp (vector float, vector float);
14878 vector signed int vec_vsubuwm (vector bool int, vector signed int);
14879 vector signed int vec_vsubuwm (vector signed int, vector bool int);
14880 vector signed int vec_vsubuwm (vector signed int, vector signed int);
14881 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
14882 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
14883 vector unsigned int vec_vsubuwm (vector unsigned int,
14884 vector unsigned int);
14886 vector signed short vec_vsubuhm (vector bool short,
14887 vector signed short);
14888 vector signed short vec_vsubuhm (vector signed short,
14889 vector bool short);
14890 vector signed short vec_vsubuhm (vector signed short,
14891 vector signed short);
14892 vector unsigned short vec_vsubuhm (vector bool short,
14893 vector unsigned short);
14894 vector unsigned short vec_vsubuhm (vector unsigned short,
14895 vector bool short);
14896 vector unsigned short vec_vsubuhm (vector unsigned short,
14897 vector unsigned short);
14899 vector signed char vec_vsububm (vector bool char, vector signed char);
14900 vector signed char vec_vsububm (vector signed char, vector bool char);
14901 vector signed char vec_vsububm (vector signed char, vector signed char);
14902 vector unsigned char vec_vsububm (vector bool char,
14903 vector unsigned char);
14904 vector unsigned char vec_vsububm (vector unsigned char,
14906 vector unsigned char vec_vsububm (vector unsigned char,
14907 vector unsigned char);
14909 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
14911 vector unsigned char vec_subs (vector bool char, vector unsigned char);
14912 vector unsigned char vec_subs (vector unsigned char, vector bool char);
14913 vector unsigned char vec_subs (vector unsigned char,
14914 vector unsigned char);
14915 vector signed char vec_subs (vector bool char, vector signed char);
14916 vector signed char vec_subs (vector signed char, vector bool char);
14917 vector signed char vec_subs (vector signed char, vector signed char);
14918 vector unsigned short vec_subs (vector bool short,
14919 vector unsigned short);
14920 vector unsigned short vec_subs (vector unsigned short,
14921 vector bool short);
14922 vector unsigned short vec_subs (vector unsigned short,
14923 vector unsigned short);
14924 vector signed short vec_subs (vector bool short, vector signed short);
14925 vector signed short vec_subs (vector signed short, vector bool short);
14926 vector signed short vec_subs (vector signed short, vector signed short);
14927 vector unsigned int vec_subs (vector bool int, vector unsigned int);
14928 vector unsigned int vec_subs (vector unsigned int, vector bool int);
14929 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
14930 vector signed int vec_subs (vector bool int, vector signed int);
14931 vector signed int vec_subs (vector signed int, vector bool int);
14932 vector signed int vec_subs (vector signed int, vector signed int);
14934 vector signed int vec_vsubsws (vector bool int, vector signed int);
14935 vector signed int vec_vsubsws (vector signed int, vector bool int);
14936 vector signed int vec_vsubsws (vector signed int, vector signed int);
14938 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
14939 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
14940 vector unsigned int vec_vsubuws (vector unsigned int,
14941 vector unsigned int);
14943 vector signed short vec_vsubshs (vector bool short,
14944 vector signed short);
14945 vector signed short vec_vsubshs (vector signed short,
14946 vector bool short);
14947 vector signed short vec_vsubshs (vector signed short,
14948 vector signed short);
14950 vector unsigned short vec_vsubuhs (vector bool short,
14951 vector unsigned short);
14952 vector unsigned short vec_vsubuhs (vector unsigned short,
14953 vector bool short);
14954 vector unsigned short vec_vsubuhs (vector unsigned short,
14955 vector unsigned short);
14957 vector signed char vec_vsubsbs (vector bool char, vector signed char);
14958 vector signed char vec_vsubsbs (vector signed char, vector bool char);
14959 vector signed char vec_vsubsbs (vector signed char, vector signed char);
14961 vector unsigned char vec_vsububs (vector bool char,
14962 vector unsigned char);
14963 vector unsigned char vec_vsububs (vector unsigned char,
14965 vector unsigned char vec_vsububs (vector unsigned char,
14966 vector unsigned char);
14968 vector unsigned int vec_sum4s (vector unsigned char,
14969 vector unsigned int);
14970 vector signed int vec_sum4s (vector signed char, vector signed int);
14971 vector signed int vec_sum4s (vector signed short, vector signed int);
14973 vector signed int vec_vsum4shs (vector signed short, vector signed int);
14975 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
14977 vector unsigned int vec_vsum4ubs (vector unsigned char,
14978 vector unsigned int);
14980 vector signed int vec_sum2s (vector signed int, vector signed int);
14982 vector signed int vec_sums (vector signed int, vector signed int);
14984 vector float vec_trunc (vector float);
14986 vector signed short vec_unpackh (vector signed char);
14987 vector bool short vec_unpackh (vector bool char);
14988 vector signed int vec_unpackh (vector signed short);
14989 vector bool int vec_unpackh (vector bool short);
14990 vector unsigned int vec_unpackh (vector pixel);
14992 vector bool int vec_vupkhsh (vector bool short);
14993 vector signed int vec_vupkhsh (vector signed short);
14995 vector unsigned int vec_vupkhpx (vector pixel);
14997 vector bool short vec_vupkhsb (vector bool char);
14998 vector signed short vec_vupkhsb (vector signed char);
15000 vector signed short vec_unpackl (vector signed char);
15001 vector bool short vec_unpackl (vector bool char);
15002 vector unsigned int vec_unpackl (vector pixel);
15003 vector signed int vec_unpackl (vector signed short);
15004 vector bool int vec_unpackl (vector bool short);
15006 vector unsigned int vec_vupklpx (vector pixel);
15008 vector bool int vec_vupklsh (vector bool short);
15009 vector signed int vec_vupklsh (vector signed short);
15011 vector bool short vec_vupklsb (vector bool char);
15012 vector signed short vec_vupklsb (vector signed char);
15014 vector float vec_xor (vector float, vector float);
15015 vector float vec_xor (vector float, vector bool int);
15016 vector float vec_xor (vector bool int, vector float);
15017 vector bool int vec_xor (vector bool int, vector bool int);
15018 vector signed int vec_xor (vector bool int, vector signed int);
15019 vector signed int vec_xor (vector signed int, vector bool int);
15020 vector signed int vec_xor (vector signed int, vector signed int);
15021 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15022 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15023 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15024 vector bool short vec_xor (vector bool short, vector bool short);
15025 vector signed short vec_xor (vector bool short, vector signed short);
15026 vector signed short vec_xor (vector signed short, vector bool short);
15027 vector signed short vec_xor (vector signed short, vector signed short);
15028 vector unsigned short vec_xor (vector bool short,
15029 vector unsigned short);
15030 vector unsigned short vec_xor (vector unsigned short,
15031 vector bool short);
15032 vector unsigned short vec_xor (vector unsigned short,
15033 vector unsigned short);
15034 vector signed char vec_xor (vector bool char, vector signed char);
15035 vector bool char vec_xor (vector bool char, vector bool char);
15036 vector signed char vec_xor (vector signed char, vector bool char);
15037 vector signed char vec_xor (vector signed char, vector signed char);
15038 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15039 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15040 vector unsigned char vec_xor (vector unsigned char,
15041 vector unsigned char);
15043 int vec_all_eq (vector signed char, vector bool char);
15044 int vec_all_eq (vector signed char, vector signed char);
15045 int vec_all_eq (vector unsigned char, vector bool char);
15046 int vec_all_eq (vector unsigned char, vector unsigned char);
15047 int vec_all_eq (vector bool char, vector bool char);
15048 int vec_all_eq (vector bool char, vector unsigned char);
15049 int vec_all_eq (vector bool char, vector signed char);
15050 int vec_all_eq (vector signed short, vector bool short);
15051 int vec_all_eq (vector signed short, vector signed short);
15052 int vec_all_eq (vector unsigned short, vector bool short);
15053 int vec_all_eq (vector unsigned short, vector unsigned short);
15054 int vec_all_eq (vector bool short, vector bool short);
15055 int vec_all_eq (vector bool short, vector unsigned short);
15056 int vec_all_eq (vector bool short, vector signed short);
15057 int vec_all_eq (vector pixel, vector pixel);
15058 int vec_all_eq (vector signed int, vector bool int);
15059 int vec_all_eq (vector signed int, vector signed int);
15060 int vec_all_eq (vector unsigned int, vector bool int);
15061 int vec_all_eq (vector unsigned int, vector unsigned int);
15062 int vec_all_eq (vector bool int, vector bool int);
15063 int vec_all_eq (vector bool int, vector unsigned int);
15064 int vec_all_eq (vector bool int, vector signed int);
15065 int vec_all_eq (vector float, vector float);
15067 int vec_all_ge (vector bool char, vector unsigned char);
15068 int vec_all_ge (vector unsigned char, vector bool char);
15069 int vec_all_ge (vector unsigned char, vector unsigned char);
15070 int vec_all_ge (vector bool char, vector signed char);
15071 int vec_all_ge (vector signed char, vector bool char);
15072 int vec_all_ge (vector signed char, vector signed char);
15073 int vec_all_ge (vector bool short, vector unsigned short);
15074 int vec_all_ge (vector unsigned short, vector bool short);
15075 int vec_all_ge (vector unsigned short, vector unsigned short);
15076 int vec_all_ge (vector signed short, vector signed short);
15077 int vec_all_ge (vector bool short, vector signed short);
15078 int vec_all_ge (vector signed short, vector bool short);
15079 int vec_all_ge (vector bool int, vector unsigned int);
15080 int vec_all_ge (vector unsigned int, vector bool int);
15081 int vec_all_ge (vector unsigned int, vector unsigned int);
15082 int vec_all_ge (vector bool int, vector signed int);
15083 int vec_all_ge (vector signed int, vector bool int);
15084 int vec_all_ge (vector signed int, vector signed int);
15085 int vec_all_ge (vector float, vector float);
15087 int vec_all_gt (vector bool char, vector unsigned char);
15088 int vec_all_gt (vector unsigned char, vector bool char);
15089 int vec_all_gt (vector unsigned char, vector unsigned char);
15090 int vec_all_gt (vector bool char, vector signed char);
15091 int vec_all_gt (vector signed char, vector bool char);
15092 int vec_all_gt (vector signed char, vector signed char);
15093 int vec_all_gt (vector bool short, vector unsigned short);
15094 int vec_all_gt (vector unsigned short, vector bool short);
15095 int vec_all_gt (vector unsigned short, vector unsigned short);
15096 int vec_all_gt (vector bool short, vector signed short);
15097 int vec_all_gt (vector signed short, vector bool short);
15098 int vec_all_gt (vector signed short, vector signed short);
15099 int vec_all_gt (vector bool int, vector unsigned int);
15100 int vec_all_gt (vector unsigned int, vector bool int);
15101 int vec_all_gt (vector unsigned int, vector unsigned int);
15102 int vec_all_gt (vector bool int, vector signed int);
15103 int vec_all_gt (vector signed int, vector bool int);
15104 int vec_all_gt (vector signed int, vector signed int);
15105 int vec_all_gt (vector float, vector float);
15107 int vec_all_in (vector float, vector float);
15109 int vec_all_le (vector bool char, vector unsigned char);
15110 int vec_all_le (vector unsigned char, vector bool char);
15111 int vec_all_le (vector unsigned char, vector unsigned char);
15112 int vec_all_le (vector bool char, vector signed char);
15113 int vec_all_le (vector signed char, vector bool char);
15114 int vec_all_le (vector signed char, vector signed char);
15115 int vec_all_le (vector bool short, vector unsigned short);
15116 int vec_all_le (vector unsigned short, vector bool short);
15117 int vec_all_le (vector unsigned short, vector unsigned short);
15118 int vec_all_le (vector bool short, vector signed short);
15119 int vec_all_le (vector signed short, vector bool short);
15120 int vec_all_le (vector signed short, vector signed short);
15121 int vec_all_le (vector bool int, vector unsigned int);
15122 int vec_all_le (vector unsigned int, vector bool int);
15123 int vec_all_le (vector unsigned int, vector unsigned int);
15124 int vec_all_le (vector bool int, vector signed int);
15125 int vec_all_le (vector signed int, vector bool int);
15126 int vec_all_le (vector signed int, vector signed int);
15127 int vec_all_le (vector float, vector float);
15129 int vec_all_lt (vector bool char, vector unsigned char);
15130 int vec_all_lt (vector unsigned char, vector bool char);
15131 int vec_all_lt (vector unsigned char, vector unsigned char);
15132 int vec_all_lt (vector bool char, vector signed char);
15133 int vec_all_lt (vector signed char, vector bool char);
15134 int vec_all_lt (vector signed char, vector signed char);
15135 int vec_all_lt (vector bool short, vector unsigned short);
15136 int vec_all_lt (vector unsigned short, vector bool short);
15137 int vec_all_lt (vector unsigned short, vector unsigned short);
15138 int vec_all_lt (vector bool short, vector signed short);
15139 int vec_all_lt (vector signed short, vector bool short);
15140 int vec_all_lt (vector signed short, vector signed short);
15141 int vec_all_lt (vector bool int, vector unsigned int);
15142 int vec_all_lt (vector unsigned int, vector bool int);
15143 int vec_all_lt (vector unsigned int, vector unsigned int);
15144 int vec_all_lt (vector bool int, vector signed int);
15145 int vec_all_lt (vector signed int, vector bool int);
15146 int vec_all_lt (vector signed int, vector signed int);
15147 int vec_all_lt (vector float, vector float);
15149 int vec_all_nan (vector float);
15151 int vec_all_ne (vector signed char, vector bool char);
15152 int vec_all_ne (vector signed char, vector signed char);
15153 int vec_all_ne (vector unsigned char, vector bool char);
15154 int vec_all_ne (vector unsigned char, vector unsigned char);
15155 int vec_all_ne (vector bool char, vector bool char);
15156 int vec_all_ne (vector bool char, vector unsigned char);
15157 int vec_all_ne (vector bool char, vector signed char);
15158 int vec_all_ne (vector signed short, vector bool short);
15159 int vec_all_ne (vector signed short, vector signed short);
15160 int vec_all_ne (vector unsigned short, vector bool short);
15161 int vec_all_ne (vector unsigned short, vector unsigned short);
15162 int vec_all_ne (vector bool short, vector bool short);
15163 int vec_all_ne (vector bool short, vector unsigned short);
15164 int vec_all_ne (vector bool short, vector signed short);
15165 int vec_all_ne (vector pixel, vector pixel);
15166 int vec_all_ne (vector signed int, vector bool int);
15167 int vec_all_ne (vector signed int, vector signed int);
15168 int vec_all_ne (vector unsigned int, vector bool int);
15169 int vec_all_ne (vector unsigned int, vector unsigned int);
15170 int vec_all_ne (vector bool int, vector bool int);
15171 int vec_all_ne (vector bool int, vector unsigned int);
15172 int vec_all_ne (vector bool int, vector signed int);
15173 int vec_all_ne (vector float, vector float);
15175 int vec_all_nge (vector float, vector float);
15177 int vec_all_ngt (vector float, vector float);
15179 int vec_all_nle (vector float, vector float);
15181 int vec_all_nlt (vector float, vector float);
15183 int vec_all_numeric (vector float);
15185 int vec_any_eq (vector signed char, vector bool char);
15186 int vec_any_eq (vector signed char, vector signed char);
15187 int vec_any_eq (vector unsigned char, vector bool char);
15188 int vec_any_eq (vector unsigned char, vector unsigned char);
15189 int vec_any_eq (vector bool char, vector bool char);
15190 int vec_any_eq (vector bool char, vector unsigned char);
15191 int vec_any_eq (vector bool char, vector signed char);
15192 int vec_any_eq (vector signed short, vector bool short);
15193 int vec_any_eq (vector signed short, vector signed short);
15194 int vec_any_eq (vector unsigned short, vector bool short);
15195 int vec_any_eq (vector unsigned short, vector unsigned short);
15196 int vec_any_eq (vector bool short, vector bool short);
15197 int vec_any_eq (vector bool short, vector unsigned short);
15198 int vec_any_eq (vector bool short, vector signed short);
15199 int vec_any_eq (vector pixel, vector pixel);
15200 int vec_any_eq (vector signed int, vector bool int);
15201 int vec_any_eq (vector signed int, vector signed int);
15202 int vec_any_eq (vector unsigned int, vector bool int);
15203 int vec_any_eq (vector unsigned int, vector unsigned int);
15204 int vec_any_eq (vector bool int, vector bool int);
15205 int vec_any_eq (vector bool int, vector unsigned int);
15206 int vec_any_eq (vector bool int, vector signed int);
15207 int vec_any_eq (vector float, vector float);
15209 int vec_any_ge (vector signed char, vector bool char);
15210 int vec_any_ge (vector unsigned char, vector bool char);
15211 int vec_any_ge (vector unsigned char, vector unsigned char);
15212 int vec_any_ge (vector signed char, vector signed char);
15213 int vec_any_ge (vector bool char, vector unsigned char);
15214 int vec_any_ge (vector bool char, vector signed char);
15215 int vec_any_ge (vector unsigned short, vector bool short);
15216 int vec_any_ge (vector unsigned short, vector unsigned short);
15217 int vec_any_ge (vector signed short, vector signed short);
15218 int vec_any_ge (vector signed short, vector bool short);
15219 int vec_any_ge (vector bool short, vector unsigned short);
15220 int vec_any_ge (vector bool short, vector signed short);
15221 int vec_any_ge (vector signed int, vector bool int);
15222 int vec_any_ge (vector unsigned int, vector bool int);
15223 int vec_any_ge (vector unsigned int, vector unsigned int);
15224 int vec_any_ge (vector signed int, vector signed int);
15225 int vec_any_ge (vector bool int, vector unsigned int);
15226 int vec_any_ge (vector bool int, vector signed int);
15227 int vec_any_ge (vector float, vector float);
15229 int vec_any_gt (vector bool char, vector unsigned char);
15230 int vec_any_gt (vector unsigned char, vector bool char);
15231 int vec_any_gt (vector unsigned char, vector unsigned char);
15232 int vec_any_gt (vector bool char, vector signed char);
15233 int vec_any_gt (vector signed char, vector bool char);
15234 int vec_any_gt (vector signed char, vector signed char);
15235 int vec_any_gt (vector bool short, vector unsigned short);
15236 int vec_any_gt (vector unsigned short, vector bool short);
15237 int vec_any_gt (vector unsigned short, vector unsigned short);
15238 int vec_any_gt (vector bool short, vector signed short);
15239 int vec_any_gt (vector signed short, vector bool short);
15240 int vec_any_gt (vector signed short, vector signed short);
15241 int vec_any_gt (vector bool int, vector unsigned int);
15242 int vec_any_gt (vector unsigned int, vector bool int);
15243 int vec_any_gt (vector unsigned int, vector unsigned int);
15244 int vec_any_gt (vector bool int, vector signed int);
15245 int vec_any_gt (vector signed int, vector bool int);
15246 int vec_any_gt (vector signed int, vector signed int);
15247 int vec_any_gt (vector float, vector float);
15249 int vec_any_le (vector bool char, vector unsigned char);
15250 int vec_any_le (vector unsigned char, vector bool char);
15251 int vec_any_le (vector unsigned char, vector unsigned char);
15252 int vec_any_le (vector bool char, vector signed char);
15253 int vec_any_le (vector signed char, vector bool char);
15254 int vec_any_le (vector signed char, vector signed char);
15255 int vec_any_le (vector bool short, vector unsigned short);
15256 int vec_any_le (vector unsigned short, vector bool short);
15257 int vec_any_le (vector unsigned short, vector unsigned short);
15258 int vec_any_le (vector bool short, vector signed short);
15259 int vec_any_le (vector signed short, vector bool short);
15260 int vec_any_le (vector signed short, vector signed short);
15261 int vec_any_le (vector bool int, vector unsigned int);
15262 int vec_any_le (vector unsigned int, vector bool int);
15263 int vec_any_le (vector unsigned int, vector unsigned int);
15264 int vec_any_le (vector bool int, vector signed int);
15265 int vec_any_le (vector signed int, vector bool int);
15266 int vec_any_le (vector signed int, vector signed int);
15267 int vec_any_le (vector float, vector float);
15269 int vec_any_lt (vector bool char, vector unsigned char);
15270 int vec_any_lt (vector unsigned char, vector bool char);
15271 int vec_any_lt (vector unsigned char, vector unsigned char);
15272 int vec_any_lt (vector bool char, vector signed char);
15273 int vec_any_lt (vector signed char, vector bool char);
15274 int vec_any_lt (vector signed char, vector signed char);
15275 int vec_any_lt (vector bool short, vector unsigned short);
15276 int vec_any_lt (vector unsigned short, vector bool short);
15277 int vec_any_lt (vector unsigned short, vector unsigned short);
15278 int vec_any_lt (vector bool short, vector signed short);
15279 int vec_any_lt (vector signed short, vector bool short);
15280 int vec_any_lt (vector signed short, vector signed short);
15281 int vec_any_lt (vector bool int, vector unsigned int);
15282 int vec_any_lt (vector unsigned int, vector bool int);
15283 int vec_any_lt (vector unsigned int, vector unsigned int);
15284 int vec_any_lt (vector bool int, vector signed int);
15285 int vec_any_lt (vector signed int, vector bool int);
15286 int vec_any_lt (vector signed int, vector signed int);
15287 int vec_any_lt (vector float, vector float);
15289 int vec_any_nan (vector float);
15291 int vec_any_ne (vector signed char, vector bool char);
15292 int vec_any_ne (vector signed char, vector signed char);
15293 int vec_any_ne (vector unsigned char, vector bool char);
15294 int vec_any_ne (vector unsigned char, vector unsigned char);
15295 int vec_any_ne (vector bool char, vector bool char);
15296 int vec_any_ne (vector bool char, vector unsigned char);
15297 int vec_any_ne (vector bool char, vector signed char);
15298 int vec_any_ne (vector signed short, vector bool short);
15299 int vec_any_ne (vector signed short, vector signed short);
15300 int vec_any_ne (vector unsigned short, vector bool short);
15301 int vec_any_ne (vector unsigned short, vector unsigned short);
15302 int vec_any_ne (vector bool short, vector bool short);
15303 int vec_any_ne (vector bool short, vector unsigned short);
15304 int vec_any_ne (vector bool short, vector signed short);
15305 int vec_any_ne (vector pixel, vector pixel);
15306 int vec_any_ne (vector signed int, vector bool int);
15307 int vec_any_ne (vector signed int, vector signed int);
15308 int vec_any_ne (vector unsigned int, vector bool int);
15309 int vec_any_ne (vector unsigned int, vector unsigned int);
15310 int vec_any_ne (vector bool int, vector bool int);
15311 int vec_any_ne (vector bool int, vector unsigned int);
15312 int vec_any_ne (vector bool int, vector signed int);
15313 int vec_any_ne (vector float, vector float);
15315 int vec_any_nge (vector float, vector float);
15317 int vec_any_ngt (vector float, vector float);
15319 int vec_any_nle (vector float, vector float);
15321 int vec_any_nlt (vector float, vector float);
15323 int vec_any_numeric (vector float);
15325 int vec_any_out (vector float, vector float);
15328 If the vector/scalar (VSX) instruction set is available, the following
15329 additional functions are available:
15332 vector double vec_abs (vector double);
15333 vector double vec_add (vector double, vector double);
15334 vector double vec_and (vector double, vector double);
15335 vector double vec_and (vector double, vector bool long);
15336 vector double vec_and (vector bool long, vector double);
15337 vector long vec_and (vector long, vector long);
15338 vector long vec_and (vector long, vector bool long);
15339 vector long vec_and (vector bool long, vector long);
15340 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15341 vector unsigned long vec_and (vector unsigned long, vector bool long);
15342 vector unsigned long vec_and (vector bool long, vector unsigned long);
15343 vector double vec_andc (vector double, vector double);
15344 vector double vec_andc (vector double, vector bool long);
15345 vector double vec_andc (vector bool long, vector double);
15346 vector long vec_andc (vector long, vector long);
15347 vector long vec_andc (vector long, vector bool long);
15348 vector long vec_andc (vector bool long, vector long);
15349 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15350 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15351 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15352 vector double vec_ceil (vector double);
15353 vector bool long vec_cmpeq (vector double, vector double);
15354 vector bool long vec_cmpge (vector double, vector double);
15355 vector bool long vec_cmpgt (vector double, vector double);
15356 vector bool long vec_cmple (vector double, vector double);
15357 vector bool long vec_cmplt (vector double, vector double);
15358 vector double vec_cpsgn (vector double, vector double);
15359 vector float vec_div (vector float, vector float);
15360 vector double vec_div (vector double, vector double);
15361 vector long vec_div (vector long, vector long);
15362 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15363 vector double vec_floor (vector double);
15364 vector double vec_ld (int, const vector double *);
15365 vector double vec_ld (int, const double *);
15366 vector double vec_ldl (int, const vector double *);
15367 vector double vec_ldl (int, const double *);
15368 vector unsigned char vec_lvsl (int, const volatile double *);
15369 vector unsigned char vec_lvsr (int, const volatile double *);
15370 vector double vec_madd (vector double, vector double, vector double);
15371 vector double vec_max (vector double, vector double);
15372 vector signed long vec_mergeh (vector signed long, vector signed long);
15373 vector signed long vec_mergeh (vector signed long, vector bool long);
15374 vector signed long vec_mergeh (vector bool long, vector signed long);
15375 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15376 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15377 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15378 vector signed long vec_mergel (vector signed long, vector signed long);
15379 vector signed long vec_mergel (vector signed long, vector bool long);
15380 vector signed long vec_mergel (vector bool long, vector signed long);
15381 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15382 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15383 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15384 vector double vec_min (vector double, vector double);
15385 vector float vec_msub (vector float, vector float, vector float);
15386 vector double vec_msub (vector double, vector double, vector double);
15387 vector float vec_mul (vector float, vector float);
15388 vector double vec_mul (vector double, vector double);
15389 vector long vec_mul (vector long, vector long);
15390 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15391 vector float vec_nearbyint (vector float);
15392 vector double vec_nearbyint (vector double);
15393 vector float vec_nmadd (vector float, vector float, vector float);
15394 vector double vec_nmadd (vector double, vector double, vector double);
15395 vector double vec_nmsub (vector double, vector double, vector double);
15396 vector double vec_nor (vector double, vector double);
15397 vector long vec_nor (vector long, vector long);
15398 vector long vec_nor (vector long, vector bool long);
15399 vector long vec_nor (vector bool long, vector long);
15400 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15401 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15402 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15403 vector double vec_or (vector double, vector double);
15404 vector double vec_or (vector double, vector bool long);
15405 vector double vec_or (vector bool long, vector double);
15406 vector long vec_or (vector long, vector long);
15407 vector long vec_or (vector long, vector bool long);
15408 vector long vec_or (vector bool long, vector long);
15409 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15410 vector unsigned long vec_or (vector unsigned long, vector bool long);
15411 vector unsigned long vec_or (vector bool long, vector unsigned long);
15412 vector double vec_perm (vector double, vector double, vector unsigned char);
15413 vector long vec_perm (vector long, vector long, vector unsigned char);
15414 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15415 vector unsigned char);
15416 vector double vec_rint (vector double);
15417 vector double vec_recip (vector double, vector double);
15418 vector double vec_rsqrt (vector double);
15419 vector double vec_rsqrte (vector double);
15420 vector double vec_sel (vector double, vector double, vector bool long);
15421 vector double vec_sel (vector double, vector double, vector unsigned long);
15422 vector long vec_sel (vector long, vector long, vector long);
15423 vector long vec_sel (vector long, vector long, vector unsigned long);
15424 vector long vec_sel (vector long, vector long, vector bool long);
15425 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15427 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15428 vector unsigned long);
15429 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15431 vector double vec_splats (double);
15432 vector signed long vec_splats (signed long);
15433 vector unsigned long vec_splats (unsigned long);
15434 vector float vec_sqrt (vector float);
15435 vector double vec_sqrt (vector double);
15436 void vec_st (vector double, int, vector double *);
15437 void vec_st (vector double, int, double *);
15438 vector double vec_sub (vector double, vector double);
15439 vector double vec_trunc (vector double);
15440 vector double vec_xor (vector double, vector double);
15441 vector double vec_xor (vector double, vector bool long);
15442 vector double vec_xor (vector bool long, vector double);
15443 vector long vec_xor (vector long, vector long);
15444 vector long vec_xor (vector long, vector bool long);
15445 vector long vec_xor (vector bool long, vector long);
15446 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15447 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15448 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15449 int vec_all_eq (vector double, vector double);
15450 int vec_all_ge (vector double, vector double);
15451 int vec_all_gt (vector double, vector double);
15452 int vec_all_le (vector double, vector double);
15453 int vec_all_lt (vector double, vector double);
15454 int vec_all_nan (vector double);
15455 int vec_all_ne (vector double, vector double);
15456 int vec_all_nge (vector double, vector double);
15457 int vec_all_ngt (vector double, vector double);
15458 int vec_all_nle (vector double, vector double);
15459 int vec_all_nlt (vector double, vector double);
15460 int vec_all_numeric (vector double);
15461 int vec_any_eq (vector double, vector double);
15462 int vec_any_ge (vector double, vector double);
15463 int vec_any_gt (vector double, vector double);
15464 int vec_any_le (vector double, vector double);
15465 int vec_any_lt (vector double, vector double);
15466 int vec_any_nan (vector double);
15467 int vec_any_ne (vector double, vector double);
15468 int vec_any_nge (vector double, vector double);
15469 int vec_any_ngt (vector double, vector double);
15470 int vec_any_nle (vector double, vector double);
15471 int vec_any_nlt (vector double, vector double);
15472 int vec_any_numeric (vector double);
15474 vector double vec_vsx_ld (int, const vector double *);
15475 vector double vec_vsx_ld (int, const double *);
15476 vector float vec_vsx_ld (int, const vector float *);
15477 vector float vec_vsx_ld (int, const float *);
15478 vector bool int vec_vsx_ld (int, const vector bool int *);
15479 vector signed int vec_vsx_ld (int, const vector signed int *);
15480 vector signed int vec_vsx_ld (int, const int *);
15481 vector signed int vec_vsx_ld (int, const long *);
15482 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15483 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15484 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15485 vector bool short vec_vsx_ld (int, const vector bool short *);
15486 vector pixel vec_vsx_ld (int, const vector pixel *);
15487 vector signed short vec_vsx_ld (int, const vector signed short *);
15488 vector signed short vec_vsx_ld (int, const short *);
15489 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15490 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15491 vector bool char vec_vsx_ld (int, const vector bool char *);
15492 vector signed char vec_vsx_ld (int, const vector signed char *);
15493 vector signed char vec_vsx_ld (int, const signed char *);
15494 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15495 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15497 void vec_vsx_st (vector double, int, vector double *);
15498 void vec_vsx_st (vector double, int, double *);
15499 void vec_vsx_st (vector float, int, vector float *);
15500 void vec_vsx_st (vector float, int, float *);
15501 void vec_vsx_st (vector signed int, int, vector signed int *);
15502 void vec_vsx_st (vector signed int, int, int *);
15503 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15504 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15505 void vec_vsx_st (vector bool int, int, vector bool int *);
15506 void vec_vsx_st (vector bool int, int, unsigned int *);
15507 void vec_vsx_st (vector bool int, int, int *);
15508 void vec_vsx_st (vector signed short, int, vector signed short *);
15509 void vec_vsx_st (vector signed short, int, short *);
15510 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15511 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15512 void vec_vsx_st (vector bool short, int, vector bool short *);
15513 void vec_vsx_st (vector bool short, int, unsigned short *);
15514 void vec_vsx_st (vector pixel, int, vector pixel *);
15515 void vec_vsx_st (vector pixel, int, unsigned short *);
15516 void vec_vsx_st (vector pixel, int, short *);
15517 void vec_vsx_st (vector bool short, int, short *);
15518 void vec_vsx_st (vector signed char, int, vector signed char *);
15519 void vec_vsx_st (vector signed char, int, signed char *);
15520 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15521 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15522 void vec_vsx_st (vector bool char, int, vector bool char *);
15523 void vec_vsx_st (vector bool char, int, unsigned char *);
15524 void vec_vsx_st (vector bool char, int, signed char *);
15526 vector double vec_xxpermdi (vector double, vector double, int);
15527 vector float vec_xxpermdi (vector float, vector float, int);
15528 vector long long vec_xxpermdi (vector long long, vector long long, int);
15529 vector unsigned long long vec_xxpermdi (vector unsigned long long,
15530 vector unsigned long long, int);
15531 vector int vec_xxpermdi (vector int, vector int, int);
15532 vector unsigned int vec_xxpermdi (vector unsigned int,
15533 vector unsigned int, int);
15534 vector short vec_xxpermdi (vector short, vector short, int);
15535 vector unsigned short vec_xxpermdi (vector unsigned short,
15536 vector unsigned short, int);
15537 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
15538 vector unsigned char vec_xxpermdi (vector unsigned char,
15539 vector unsigned char, int);
15541 vector double vec_xxsldi (vector double, vector double, int);
15542 vector float vec_xxsldi (vector float, vector float, int);
15543 vector long long vec_xxsldi (vector long long, vector long long, int);
15544 vector unsigned long long vec_xxsldi (vector unsigned long long,
15545 vector unsigned long long, int);
15546 vector int vec_xxsldi (vector int, vector int, int);
15547 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
15548 vector short vec_xxsldi (vector short, vector short, int);
15549 vector unsigned short vec_xxsldi (vector unsigned short,
15550 vector unsigned short, int);
15551 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
15552 vector unsigned char vec_xxsldi (vector unsigned char,
15553 vector unsigned char, int);
15556 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
15557 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
15558 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
15559 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
15560 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
15562 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15563 instruction set is available, the following additional functions are
15564 available for both 32-bit and 64-bit targets. For 64-bit targets, you
15565 can use @var{vector long} instead of @var{vector long long},
15566 @var{vector bool long} instead of @var{vector bool long long}, and
15567 @var{vector unsigned long} instead of @var{vector unsigned long long}.
15570 vector long long vec_abs (vector long long);
15572 vector long long vec_add (vector long long, vector long long);
15573 vector unsigned long long vec_add (vector unsigned long long,
15574 vector unsigned long long);
15576 int vec_all_eq (vector long long, vector long long);
15577 int vec_all_eq (vector unsigned long long, vector unsigned long long);
15578 int vec_all_ge (vector long long, vector long long);
15579 int vec_all_ge (vector unsigned long long, vector unsigned long long);
15580 int vec_all_gt (vector long long, vector long long);
15581 int vec_all_gt (vector unsigned long long, vector unsigned long long);
15582 int vec_all_le (vector long long, vector long long);
15583 int vec_all_le (vector unsigned long long, vector unsigned long long);
15584 int vec_all_lt (vector long long, vector long long);
15585 int vec_all_lt (vector unsigned long long, vector unsigned long long);
15586 int vec_all_ne (vector long long, vector long long);
15587 int vec_all_ne (vector unsigned long long, vector unsigned long long);
15589 int vec_any_eq (vector long long, vector long long);
15590 int vec_any_eq (vector unsigned long long, vector unsigned long long);
15591 int vec_any_ge (vector long long, vector long long);
15592 int vec_any_ge (vector unsigned long long, vector unsigned long long);
15593 int vec_any_gt (vector long long, vector long long);
15594 int vec_any_gt (vector unsigned long long, vector unsigned long long);
15595 int vec_any_le (vector long long, vector long long);
15596 int vec_any_le (vector unsigned long long, vector unsigned long long);
15597 int vec_any_lt (vector long long, vector long long);
15598 int vec_any_lt (vector unsigned long long, vector unsigned long long);
15599 int vec_any_ne (vector long long, vector long long);
15600 int vec_any_ne (vector unsigned long long, vector unsigned long long);
15602 vector long long vec_eqv (vector long long, vector long long);
15603 vector long long vec_eqv (vector bool long long, vector long long);
15604 vector long long vec_eqv (vector long long, vector bool long long);
15605 vector unsigned long long vec_eqv (vector unsigned long long,
15606 vector unsigned long long);
15607 vector unsigned long long vec_eqv (vector bool long long,
15608 vector unsigned long long);
15609 vector unsigned long long vec_eqv (vector unsigned long long,
15610 vector bool long long);
15611 vector int vec_eqv (vector int, vector int);
15612 vector int vec_eqv (vector bool int, vector int);
15613 vector int vec_eqv (vector int, vector bool int);
15614 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
15615 vector unsigned int vec_eqv (vector bool unsigned int,
15616 vector unsigned int);
15617 vector unsigned int vec_eqv (vector unsigned int,
15618 vector bool unsigned int);
15619 vector short vec_eqv (vector short, vector short);
15620 vector short vec_eqv (vector bool short, vector short);
15621 vector short vec_eqv (vector short, vector bool short);
15622 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
15623 vector unsigned short vec_eqv (vector bool unsigned short,
15624 vector unsigned short);
15625 vector unsigned short vec_eqv (vector unsigned short,
15626 vector bool unsigned short);
15627 vector signed char vec_eqv (vector signed char, vector signed char);
15628 vector signed char vec_eqv (vector bool signed char, vector signed char);
15629 vector signed char vec_eqv (vector signed char, vector bool signed char);
15630 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
15631 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
15632 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
15634 vector long long vec_max (vector long long, vector long long);
15635 vector unsigned long long vec_max (vector unsigned long long,
15636 vector unsigned long long);
15638 vector signed int vec_mergee (vector signed int, vector signed int);
15639 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
15640 vector bool int vec_mergee (vector bool int, vector bool int);
15642 vector signed int vec_mergeo (vector signed int, vector signed int);
15643 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
15644 vector bool int vec_mergeo (vector bool int, vector bool int);
15646 vector long long vec_min (vector long long, vector long long);
15647 vector unsigned long long vec_min (vector unsigned long long,
15648 vector unsigned long long);
15650 vector long long vec_nand (vector long long, vector long long);
15651 vector long long vec_nand (vector bool long long, vector long long);
15652 vector long long vec_nand (vector long long, vector bool long long);
15653 vector unsigned long long vec_nand (vector unsigned long long,
15654 vector unsigned long long);
15655 vector unsigned long long vec_nand (vector bool long long,
15656 vector unsigned long long);
15657 vector unsigned long long vec_nand (vector unsigned long long,
15658 vector bool long long);
15659 vector int vec_nand (vector int, vector int);
15660 vector int vec_nand (vector bool int, vector int);
15661 vector int vec_nand (vector int, vector bool int);
15662 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
15663 vector unsigned int vec_nand (vector bool unsigned int,
15664 vector unsigned int);
15665 vector unsigned int vec_nand (vector unsigned int,
15666 vector bool unsigned int);
15667 vector short vec_nand (vector short, vector short);
15668 vector short vec_nand (vector bool short, vector short);
15669 vector short vec_nand (vector short, vector bool short);
15670 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
15671 vector unsigned short vec_nand (vector bool unsigned short,
15672 vector unsigned short);
15673 vector unsigned short vec_nand (vector unsigned short,
15674 vector bool unsigned short);
15675 vector signed char vec_nand (vector signed char, vector signed char);
15676 vector signed char vec_nand (vector bool signed char, vector signed char);
15677 vector signed char vec_nand (vector signed char, vector bool signed char);
15678 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
15679 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
15680 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
15682 vector long long vec_orc (vector long long, vector long long);
15683 vector long long vec_orc (vector bool long long, vector long long);
15684 vector long long vec_orc (vector long long, vector bool long long);
15685 vector unsigned long long vec_orc (vector unsigned long long,
15686 vector unsigned long long);
15687 vector unsigned long long vec_orc (vector bool long long,
15688 vector unsigned long long);
15689 vector unsigned long long vec_orc (vector unsigned long long,
15690 vector bool long long);
15691 vector int vec_orc (vector int, vector int);
15692 vector int vec_orc (vector bool int, vector int);
15693 vector int vec_orc (vector int, vector bool int);
15694 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
15695 vector unsigned int vec_orc (vector bool unsigned int,
15696 vector unsigned int);
15697 vector unsigned int vec_orc (vector unsigned int,
15698 vector bool unsigned int);
15699 vector short vec_orc (vector short, vector short);
15700 vector short vec_orc (vector bool short, vector short);
15701 vector short vec_orc (vector short, vector bool short);
15702 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
15703 vector unsigned short vec_orc (vector bool unsigned short,
15704 vector unsigned short);
15705 vector unsigned short vec_orc (vector unsigned short,
15706 vector bool unsigned short);
15707 vector signed char vec_orc (vector signed char, vector signed char);
15708 vector signed char vec_orc (vector bool signed char, vector signed char);
15709 vector signed char vec_orc (vector signed char, vector bool signed char);
15710 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
15711 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
15712 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
15714 vector int vec_pack (vector long long, vector long long);
15715 vector unsigned int vec_pack (vector unsigned long long,
15716 vector unsigned long long);
15717 vector bool int vec_pack (vector bool long long, vector bool long long);
15719 vector int vec_packs (vector long long, vector long long);
15720 vector unsigned int vec_packs (vector unsigned long long,
15721 vector unsigned long long);
15723 vector unsigned int vec_packsu (vector long long, vector long long);
15724 vector unsigned int vec_packsu (vector unsigned long long,
15725 vector unsigned long long);
15727 vector long long vec_rl (vector long long,
15728 vector unsigned long long);
15729 vector long long vec_rl (vector unsigned long long,
15730 vector unsigned long long);
15732 vector long long vec_sl (vector long long, vector unsigned long long);
15733 vector long long vec_sl (vector unsigned long long,
15734 vector unsigned long long);
15736 vector long long vec_sr (vector long long, vector unsigned long long);
15737 vector unsigned long long char vec_sr (vector unsigned long long,
15738 vector unsigned long long);
15740 vector long long vec_sra (vector long long, vector unsigned long long);
15741 vector unsigned long long vec_sra (vector unsigned long long,
15742 vector unsigned long long);
15744 vector long long vec_sub (vector long long, vector long long);
15745 vector unsigned long long vec_sub (vector unsigned long long,
15746 vector unsigned long long);
15748 vector long long vec_unpackh (vector int);
15749 vector unsigned long long vec_unpackh (vector unsigned int);
15751 vector long long vec_unpackl (vector int);
15752 vector unsigned long long vec_unpackl (vector unsigned int);
15754 vector long long vec_vaddudm (vector long long, vector long long);
15755 vector long long vec_vaddudm (vector bool long long, vector long long);
15756 vector long long vec_vaddudm (vector long long, vector bool long long);
15757 vector unsigned long long vec_vaddudm (vector unsigned long long,
15758 vector unsigned long long);
15759 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
15760 vector unsigned long long);
15761 vector unsigned long long vec_vaddudm (vector unsigned long long,
15762 vector bool unsigned long long);
15764 vector long long vec_vbpermq (vector signed char, vector signed char);
15765 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
15767 vector long long vec_cntlz (vector long long);
15768 vector unsigned long long vec_cntlz (vector unsigned long long);
15769 vector int vec_cntlz (vector int);
15770 vector unsigned int vec_cntlz (vector int);
15771 vector short vec_cntlz (vector short);
15772 vector unsigned short vec_cntlz (vector unsigned short);
15773 vector signed char vec_cntlz (vector signed char);
15774 vector unsigned char vec_cntlz (vector unsigned char);
15776 vector long long vec_vclz (vector long long);
15777 vector unsigned long long vec_vclz (vector unsigned long long);
15778 vector int vec_vclz (vector int);
15779 vector unsigned int vec_vclz (vector int);
15780 vector short vec_vclz (vector short);
15781 vector unsigned short vec_vclz (vector unsigned short);
15782 vector signed char vec_vclz (vector signed char);
15783 vector unsigned char vec_vclz (vector unsigned char);
15785 vector signed char vec_vclzb (vector signed char);
15786 vector unsigned char vec_vclzb (vector unsigned char);
15788 vector long long vec_vclzd (vector long long);
15789 vector unsigned long long vec_vclzd (vector unsigned long long);
15791 vector short vec_vclzh (vector short);
15792 vector unsigned short vec_vclzh (vector unsigned short);
15794 vector int vec_vclzw (vector int);
15795 vector unsigned int vec_vclzw (vector int);
15797 vector signed char vec_vgbbd (vector signed char);
15798 vector unsigned char vec_vgbbd (vector unsigned char);
15800 vector long long vec_vmaxsd (vector long long, vector long long);
15802 vector unsigned long long vec_vmaxud (vector unsigned long long,
15803 unsigned vector long long);
15805 vector long long vec_vminsd (vector long long, vector long long);
15807 vector unsigned long long vec_vminud (vector long long,
15810 vector int vec_vpksdss (vector long long, vector long long);
15811 vector unsigned int vec_vpksdss (vector long long, vector long long);
15813 vector unsigned int vec_vpkudus (vector unsigned long long,
15814 vector unsigned long long);
15816 vector int vec_vpkudum (vector long long, vector long long);
15817 vector unsigned int vec_vpkudum (vector unsigned long long,
15818 vector unsigned long long);
15819 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
15821 vector long long vec_vpopcnt (vector long long);
15822 vector unsigned long long vec_vpopcnt (vector unsigned long long);
15823 vector int vec_vpopcnt (vector int);
15824 vector unsigned int vec_vpopcnt (vector int);
15825 vector short vec_vpopcnt (vector short);
15826 vector unsigned short vec_vpopcnt (vector unsigned short);
15827 vector signed char vec_vpopcnt (vector signed char);
15828 vector unsigned char vec_vpopcnt (vector unsigned char);
15830 vector signed char vec_vpopcntb (vector signed char);
15831 vector unsigned char vec_vpopcntb (vector unsigned char);
15833 vector long long vec_vpopcntd (vector long long);
15834 vector unsigned long long vec_vpopcntd (vector unsigned long long);
15836 vector short vec_vpopcnth (vector short);
15837 vector unsigned short vec_vpopcnth (vector unsigned short);
15839 vector int vec_vpopcntw (vector int);
15840 vector unsigned int vec_vpopcntw (vector int);
15842 vector long long vec_vrld (vector long long, vector unsigned long long);
15843 vector unsigned long long vec_vrld (vector unsigned long long,
15844 vector unsigned long long);
15846 vector long long vec_vsld (vector long long, vector unsigned long long);
15847 vector long long vec_vsld (vector unsigned long long,
15848 vector unsigned long long);
15850 vector long long vec_vsrad (vector long long, vector unsigned long long);
15851 vector unsigned long long vec_vsrad (vector unsigned long long,
15852 vector unsigned long long);
15854 vector long long vec_vsrd (vector long long, vector unsigned long long);
15855 vector unsigned long long char vec_vsrd (vector unsigned long long,
15856 vector unsigned long long);
15858 vector long long vec_vsubudm (vector long long, vector long long);
15859 vector long long vec_vsubudm (vector bool long long, vector long long);
15860 vector long long vec_vsubudm (vector long long, vector bool long long);
15861 vector unsigned long long vec_vsubudm (vector unsigned long long,
15862 vector unsigned long long);
15863 vector unsigned long long vec_vsubudm (vector bool long long,
15864 vector unsigned long long);
15865 vector unsigned long long vec_vsubudm (vector unsigned long long,
15866 vector bool long long);
15868 vector long long vec_vupkhsw (vector int);
15869 vector unsigned long long vec_vupkhsw (vector unsigned int);
15871 vector long long vec_vupklsw (vector int);
15872 vector unsigned long long vec_vupklsw (vector int);
15875 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15876 instruction set is available, the following additional functions are
15877 available for 64-bit targets. New vector types
15878 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
15879 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
15882 The normal vector extract, and set operations work on
15883 @var{vector __int128_t} and @var{vector __uint128_t} types,
15884 but the index value must be 0.
15887 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
15888 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
15890 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
15891 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
15893 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
15894 vector __int128_t);
15895 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
15896 vector __uint128_t);
15898 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
15899 vector __int128_t);
15900 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
15901 vector __uint128_t);
15903 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
15904 vector __int128_t);
15905 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
15906 vector __uint128_t);
15908 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
15909 vector __int128_t);
15910 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
15911 vector __uint128_t);
15913 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
15914 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
15916 __int128_t vec_vsubuqm (__int128_t, __int128_t);
15917 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
15919 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
15920 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
15921 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
15922 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
15923 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
15924 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
15925 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
15926 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
15927 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
15928 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
15931 If the cryptographic instructions are enabled (@option{-mcrypto} or
15932 @option{-mcpu=power8}), the following builtins are enabled.
15935 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
15937 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
15938 vector unsigned long long);
15940 vector unsigned long long __builtin_crypto_vcipherlast
15941 (vector unsigned long long,
15942 vector unsigned long long);
15944 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
15945 vector unsigned long long);
15947 vector unsigned long long __builtin_crypto_vncipherlast
15948 (vector unsigned long long,
15949 vector unsigned long long);
15951 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
15952 vector unsigned char,
15953 vector unsigned char);
15955 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
15956 vector unsigned short,
15957 vector unsigned short);
15959 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
15960 vector unsigned int,
15961 vector unsigned int);
15963 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
15964 vector unsigned long long,
15965 vector unsigned long long);
15967 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
15968 vector unsigned char);
15970 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
15971 vector unsigned short);
15973 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
15974 vector unsigned int);
15976 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
15977 vector unsigned long long);
15979 vector unsigned long long __builtin_crypto_vshasigmad
15980 (vector unsigned long long, int, int);
15982 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
15986 The second argument to the @var{__builtin_crypto_vshasigmad} and
15987 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
15988 integer that is 0 or 1. The third argument to these builtin functions
15989 must be a constant integer in the range of 0 to 15.
15991 @node PowerPC Hardware Transactional Memory Built-in Functions
15992 @subsection PowerPC Hardware Transactional Memory Built-in Functions
15993 GCC provides two interfaces for accessing the Hardware Transactional
15994 Memory (HTM) instructions available on some of the PowerPC family
15995 of processors (eg, POWER8). The two interfaces come in a low level
15996 interface, consisting of built-in functions specific to PowerPC and a
15997 higher level interface consisting of inline functions that are common
15998 between PowerPC and S/390.
16000 @subsubsection PowerPC HTM Low Level Built-in Functions
16002 The following low level built-in functions are available with
16003 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16004 They all generate the machine instruction that is part of the name.
16006 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16007 the full 4-bit condition register value set by their associated hardware
16008 instruction. The header file @code{htmintrin.h} defines some macros that can
16009 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16010 returns a simple true or false value depending on whether a transaction was
16011 successfully started or not. The arguments of the builtins match exactly the
16012 type and order of the associated hardware instruction's operands, except for
16013 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16014 Refer to the ISA manual for a description of each instruction's operands.
16017 unsigned int __builtin_tbegin (unsigned int)
16018 unsigned int __builtin_tend (unsigned int)
16020 unsigned int __builtin_tabort (unsigned int)
16021 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16022 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16023 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16024 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16026 unsigned int __builtin_tcheck (void)
16027 unsigned int __builtin_treclaim (unsigned int)
16028 unsigned int __builtin_trechkpt (void)
16029 unsigned int __builtin_tsr (unsigned int)
16032 In addition to the above HTM built-ins, we have added built-ins for
16033 some common extended mnemonics of the HTM instructions:
16036 unsigned int __builtin_tendall (void)
16037 unsigned int __builtin_tresume (void)
16038 unsigned int __builtin_tsuspend (void)
16041 The following set of built-in functions are available to gain access
16042 to the HTM specific special purpose registers.
16045 unsigned long __builtin_get_texasr (void)
16046 unsigned long __builtin_get_texasru (void)
16047 unsigned long __builtin_get_tfhar (void)
16048 unsigned long __builtin_get_tfiar (void)
16050 void __builtin_set_texasr (unsigned long);
16051 void __builtin_set_texasru (unsigned long);
16052 void __builtin_set_tfhar (unsigned long);
16053 void __builtin_set_tfiar (unsigned long);
16056 Example usage of these low level built-in functions may look like:
16059 #include <htmintrin.h>
16061 int num_retries = 10;
16065 if (__builtin_tbegin (0))
16067 /* Transaction State Initiated. */
16068 if (is_locked (lock))
16069 __builtin_tabort (0);
16070 ... transaction code...
16071 __builtin_tend (0);
16076 /* Transaction State Failed. Use locks if the transaction
16077 failure is "persistent" or we've tried too many times. */
16078 if (num_retries-- <= 0
16079 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16081 acquire_lock (lock);
16082 ... non transactional fallback path...
16083 release_lock (lock);
16090 One final built-in function has been added that returns the value of
16091 the 2-bit Transaction State field of the Machine Status Register (MSR)
16092 as stored in @code{CR0}.
16095 unsigned long __builtin_ttest (void)
16098 This built-in can be used to determine the current transaction state
16099 using the following code example:
16102 #include <htmintrin.h>
16104 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16106 if (tx_state == _HTM_TRANSACTIONAL)
16108 /* Code to use in transactional state. */
16110 else if (tx_state == _HTM_NONTRANSACTIONAL)
16112 /* Code to use in non-transactional state. */
16114 else if (tx_state == _HTM_SUSPENDED)
16116 /* Code to use in transaction suspended state. */
16120 @subsubsection PowerPC HTM High Level Inline Functions
16122 The following high level HTM interface is made available by including
16123 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16124 where CPU is `power8' or later. This interface is common between PowerPC
16125 and S/390, allowing users to write one HTM source implementation that
16126 can be compiled and executed on either system.
16129 long __TM_simple_begin (void)
16130 long __TM_begin (void* const TM_buff)
16131 long __TM_end (void)
16132 void __TM_abort (void)
16133 void __TM_named_abort (unsigned char const code)
16134 void __TM_resume (void)
16135 void __TM_suspend (void)
16137 long __TM_is_user_abort (void* const TM_buff)
16138 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16139 long __TM_is_illegal (void* const TM_buff)
16140 long __TM_is_footprint_exceeded (void* const TM_buff)
16141 long __TM_nesting_depth (void* const TM_buff)
16142 long __TM_is_nested_too_deep(void* const TM_buff)
16143 long __TM_is_conflict(void* const TM_buff)
16144 long __TM_is_failure_persistent(void* const TM_buff)
16145 long __TM_failure_address(void* const TM_buff)
16146 long long __TM_failure_code(void* const TM_buff)
16149 Using these common set of HTM inline functions, we can create
16150 a more portable version of the HTM example in the previous
16151 section that will work on either PowerPC or S/390:
16154 #include <htmxlintrin.h>
16156 int num_retries = 10;
16157 TM_buff_type TM_buff;
16161 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16163 /* Transaction State Initiated. */
16164 if (is_locked (lock))
16166 ... transaction code...
16172 /* Transaction State Failed. Use locks if the transaction
16173 failure is "persistent" or we've tried too many times. */
16174 if (num_retries-- <= 0
16175 || __TM_is_failure_persistent (TM_buff))
16177 acquire_lock (lock);
16178 ... non transactional fallback path...
16179 release_lock (lock);
16186 @node RX Built-in Functions
16187 @subsection RX Built-in Functions
16188 GCC supports some of the RX instructions which cannot be expressed in
16189 the C programming language via the use of built-in functions. The
16190 following functions are supported:
16192 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16193 Generates the @code{brk} machine instruction.
16196 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16197 Generates the @code{clrpsw} machine instruction to clear the specified
16198 bit in the processor status word.
16201 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16202 Generates the @code{int} machine instruction to generate an interrupt
16203 with the specified value.
16206 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16207 Generates the @code{machi} machine instruction to add the result of
16208 multiplying the top 16 bits of the two arguments into the
16212 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16213 Generates the @code{maclo} machine instruction to add the result of
16214 multiplying the bottom 16 bits of the two arguments into the
16218 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16219 Generates the @code{mulhi} machine instruction to place the result of
16220 multiplying the top 16 bits of the two arguments into the
16224 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16225 Generates the @code{mullo} machine instruction to place the result of
16226 multiplying the bottom 16 bits of the two arguments into the
16230 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16231 Generates the @code{mvfachi} machine instruction to read the top
16232 32 bits of the accumulator.
16235 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16236 Generates the @code{mvfacmi} machine instruction to read the middle
16237 32 bits of the accumulator.
16240 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16241 Generates the @code{mvfc} machine instruction which reads the control
16242 register specified in its argument and returns its value.
16245 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16246 Generates the @code{mvtachi} machine instruction to set the top
16247 32 bits of the accumulator.
16250 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16251 Generates the @code{mvtaclo} machine instruction to set the bottom
16252 32 bits of the accumulator.
16255 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16256 Generates the @code{mvtc} machine instruction which sets control
16257 register number @code{reg} to @code{val}.
16260 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16261 Generates the @code{mvtipl} machine instruction set the interrupt
16265 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16266 Generates the @code{racw} machine instruction to round the accumulator
16267 according to the specified mode.
16270 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16271 Generates the @code{revw} machine instruction which swaps the bytes in
16272 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16273 and also bits 16--23 occupy bits 24--31 and vice versa.
16276 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16277 Generates the @code{rmpa} machine instruction which initiates a
16278 repeated multiply and accumulate sequence.
16281 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16282 Generates the @code{round} machine instruction which returns the
16283 floating-point argument rounded according to the current rounding mode
16284 set in the floating-point status word register.
16287 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16288 Generates the @code{sat} machine instruction which returns the
16289 saturated value of the argument.
16292 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16293 Generates the @code{setpsw} machine instruction to set the specified
16294 bit in the processor status word.
16297 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16298 Generates the @code{wait} machine instruction.
16301 @node S/390 System z Built-in Functions
16302 @subsection S/390 System z Built-in Functions
16303 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16304 Generates the @code{tbegin} machine instruction starting a
16305 non-constraint hardware transaction. If the parameter is non-NULL the
16306 memory area is used to store the transaction diagnostic buffer and
16307 will be passed as first operand to @code{tbegin}. This buffer can be
16308 defined using the @code{struct __htm_tdb} C struct defined in
16309 @code{htmintrin.h} and must reside on a double-word boundary. The
16310 second tbegin operand is set to @code{0xff0c}. This enables
16311 save/restore of all GPRs and disables aborts for FPR and AR
16312 manipulations inside the transaction body. The condition code set by
16313 the tbegin instruction is returned as integer value. The tbegin
16314 instruction by definition overwrites the content of all FPRs. The
16315 compiler will generate code which saves and restores the FPRs. For
16316 soft-float code it is recommended to used the @code{*_nofloat}
16317 variant. In order to prevent a TDB from being written it is required
16318 to pass an constant zero value as parameter. Passing the zero value
16319 through a variable is not sufficient. Although modifications of
16320 access registers inside the transaction will not trigger an
16321 transaction abort it is not supported to actually modify them. Access
16322 registers do not get saved when entering a transaction. They will have
16323 undefined state when reaching the abort code.
16326 Macros for the possible return codes of tbegin are defined in the
16327 @code{htmintrin.h} header file:
16330 @item _HTM_TBEGIN_STARTED
16331 @code{tbegin} has been executed as part of normal processing. The
16332 transaction body is supposed to be executed.
16333 @item _HTM_TBEGIN_INDETERMINATE
16334 The transaction was aborted due to an indeterminate condition which
16335 might be persistent.
16336 @item _HTM_TBEGIN_TRANSIENT
16337 The transaction aborted due to a transient failure. The transaction
16338 should be re-executed in that case.
16339 @item _HTM_TBEGIN_PERSISTENT
16340 The transaction aborted due to a persistent failure. Re-execution
16341 under same circumstances will not be productive.
16344 @defmac _HTM_FIRST_USER_ABORT_CODE
16345 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16346 specifies the first abort code which can be used for
16347 @code{__builtin_tabort}. Values below this threshold are reserved for
16351 @deftp {Data type} {struct __htm_tdb}
16352 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16353 the structure of the transaction diagnostic block as specified in the
16354 Principles of Operation manual chapter 5-91.
16357 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16358 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16359 Using this variant in code making use of FPRs will leave the FPRs in
16360 undefined state when entering the transaction abort handler code.
16363 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16364 In addition to @code{__builtin_tbegin} a loop for transient failures
16365 is generated. If tbegin returns a condition code of 2 the transaction
16366 will be retried as often as specified in the second argument. The
16367 perform processor assist instruction is used to tell the CPU about the
16368 number of fails so far.
16371 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16372 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16373 restores. Using this variant in code making use of FPRs will leave
16374 the FPRs in undefined state when entering the transaction abort
16378 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16379 Generates the @code{tbeginc} machine instruction starting a constraint
16380 hardware transaction. The second operand is set to @code{0xff08}.
16383 @deftypefn {Built-in Function} int __builtin_tend (void)
16384 Generates the @code{tend} machine instruction finishing a transaction
16385 and making the changes visible to other threads. The condition code
16386 generated by tend is returned as integer value.
16389 @deftypefn {Built-in Function} void __builtin_tabort (int)
16390 Generates the @code{tabort} machine instruction with the specified
16391 abort code. Abort codes from 0 through 255 are reserved and will
16392 result in an error message.
16395 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16396 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16397 integer parameter is loaded into rX and a value of zero is loaded into
16398 rY. The integer parameter specifies the number of times the
16399 transaction repeatedly aborted.
16402 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16403 Generates the @code{etnd} machine instruction. The current nesting
16404 depth is returned as integer value. For a nesting depth of 0 the code
16405 is not executed as part of an transaction.
16408 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16410 Generates the @code{ntstg} machine instruction. The second argument
16411 is written to the first arguments location. The store operation will
16412 not be rolled-back in case of an transaction abort.
16415 @node SH Built-in Functions
16416 @subsection SH Built-in Functions
16417 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16418 families of processors:
16420 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16421 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16422 used by system code that manages threads and execution contexts. The compiler
16423 normally does not generate code that modifies the contents of @samp{GBR} and
16424 thus the value is preserved across function calls. Changing the @samp{GBR}
16425 value in user code must be done with caution, since the compiler might use
16426 @samp{GBR} in order to access thread local variables.
16430 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16431 Returns the value that is currently set in the @samp{GBR} register.
16432 Memory loads and stores that use the thread pointer as a base address are
16433 turned into @samp{GBR} based displacement loads and stores, if possible.
16441 int get_tcb_value (void)
16443 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16444 return ((my_tcb*)__builtin_thread_pointer ())->c;
16450 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16451 Returns the value that is currently set in the @samp{FPSCR} register.
16454 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16455 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16456 preserving the current values of the FR, SZ and PR bits.
16459 @node SPARC VIS Built-in Functions
16460 @subsection SPARC VIS Built-in Functions
16462 GCC supports SIMD operations on the SPARC using both the generic vector
16463 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16464 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16465 switch, the VIS extension is exposed as the following built-in functions:
16468 typedef int v1si __attribute__ ((vector_size (4)));
16469 typedef int v2si __attribute__ ((vector_size (8)));
16470 typedef short v4hi __attribute__ ((vector_size (8)));
16471 typedef short v2hi __attribute__ ((vector_size (4)));
16472 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16473 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16475 void __builtin_vis_write_gsr (int64_t);
16476 int64_t __builtin_vis_read_gsr (void);
16478 void * __builtin_vis_alignaddr (void *, long);
16479 void * __builtin_vis_alignaddrl (void *, long);
16480 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16481 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16482 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16483 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16485 v4hi __builtin_vis_fexpand (v4qi);
16487 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16488 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16489 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16490 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16491 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16492 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16493 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16495 v4qi __builtin_vis_fpack16 (v4hi);
16496 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16497 v2hi __builtin_vis_fpackfix (v2si);
16498 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16500 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
16502 long __builtin_vis_edge8 (void *, void *);
16503 long __builtin_vis_edge8l (void *, void *);
16504 long __builtin_vis_edge16 (void *, void *);
16505 long __builtin_vis_edge16l (void *, void *);
16506 long __builtin_vis_edge32 (void *, void *);
16507 long __builtin_vis_edge32l (void *, void *);
16509 long __builtin_vis_fcmple16 (v4hi, v4hi);
16510 long __builtin_vis_fcmple32 (v2si, v2si);
16511 long __builtin_vis_fcmpne16 (v4hi, v4hi);
16512 long __builtin_vis_fcmpne32 (v2si, v2si);
16513 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
16514 long __builtin_vis_fcmpgt32 (v2si, v2si);
16515 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
16516 long __builtin_vis_fcmpeq32 (v2si, v2si);
16518 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
16519 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
16520 v2si __builtin_vis_fpadd32 (v2si, v2si);
16521 v1si __builtin_vis_fpadd32s (v1si, v1si);
16522 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
16523 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
16524 v2si __builtin_vis_fpsub32 (v2si, v2si);
16525 v1si __builtin_vis_fpsub32s (v1si, v1si);
16527 long __builtin_vis_array8 (long, long);
16528 long __builtin_vis_array16 (long, long);
16529 long __builtin_vis_array32 (long, long);
16532 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
16533 functions also become available:
16536 long __builtin_vis_bmask (long, long);
16537 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
16538 v2si __builtin_vis_bshufflev2si (v2si, v2si);
16539 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
16540 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
16542 long __builtin_vis_edge8n (void *, void *);
16543 long __builtin_vis_edge8ln (void *, void *);
16544 long __builtin_vis_edge16n (void *, void *);
16545 long __builtin_vis_edge16ln (void *, void *);
16546 long __builtin_vis_edge32n (void *, void *);
16547 long __builtin_vis_edge32ln (void *, void *);
16550 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
16551 functions also become available:
16554 void __builtin_vis_cmask8 (long);
16555 void __builtin_vis_cmask16 (long);
16556 void __builtin_vis_cmask32 (long);
16558 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
16560 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
16561 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
16562 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
16563 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
16564 v2si __builtin_vis_fsll16 (v2si, v2si);
16565 v2si __builtin_vis_fslas16 (v2si, v2si);
16566 v2si __builtin_vis_fsrl16 (v2si, v2si);
16567 v2si __builtin_vis_fsra16 (v2si, v2si);
16569 long __builtin_vis_pdistn (v8qi, v8qi);
16571 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
16573 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
16574 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
16576 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
16577 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
16578 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
16579 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
16580 v2si __builtin_vis_fpadds32 (v2si, v2si);
16581 v1si __builtin_vis_fpadds32s (v1si, v1si);
16582 v2si __builtin_vis_fpsubs32 (v2si, v2si);
16583 v1si __builtin_vis_fpsubs32s (v1si, v1si);
16585 long __builtin_vis_fucmple8 (v8qi, v8qi);
16586 long __builtin_vis_fucmpne8 (v8qi, v8qi);
16587 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
16588 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
16590 float __builtin_vis_fhadds (float, float);
16591 double __builtin_vis_fhaddd (double, double);
16592 float __builtin_vis_fhsubs (float, float);
16593 double __builtin_vis_fhsubd (double, double);
16594 float __builtin_vis_fnhadds (float, float);
16595 double __builtin_vis_fnhaddd (double, double);
16597 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
16598 int64_t __builtin_vis_xmulx (int64_t, int64_t);
16599 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
16602 @node SPU Built-in Functions
16603 @subsection SPU Built-in Functions
16605 GCC provides extensions for the SPU processor as described in the
16606 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
16607 found at @uref{http://cell.scei.co.jp/} or
16608 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
16609 implementation differs in several ways.
16614 The optional extension of specifying vector constants in parentheses is
16618 A vector initializer requires no cast if the vector constant is of the
16619 same type as the variable it is initializing.
16622 If @code{signed} or @code{unsigned} is omitted, the signedness of the
16623 vector type is the default signedness of the base type. The default
16624 varies depending on the operating system, so a portable program should
16625 always specify the signedness.
16628 By default, the keyword @code{__vector} is added. The macro
16629 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
16633 GCC allows using a @code{typedef} name as the type specifier for a
16637 For C, overloaded functions are implemented with macros so the following
16641 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
16645 Since @code{spu_add} is a macro, the vector constant in the example
16646 is treated as four separate arguments. Wrap the entire argument in
16647 parentheses for this to work.
16650 The extended version of @code{__builtin_expect} is not supported.
16654 @emph{Note:} Only the interface described in the aforementioned
16655 specification is supported. Internally, GCC uses built-in functions to
16656 implement the required functionality, but these are not supported and
16657 are subject to change without notice.
16659 @node TI C6X Built-in Functions
16660 @subsection TI C6X Built-in Functions
16662 GCC provides intrinsics to access certain instructions of the TI C6X
16663 processors. These intrinsics, listed below, are available after
16664 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
16665 to C6X instructions.
16669 int _sadd (int, int)
16670 int _ssub (int, int)
16671 int _sadd2 (int, int)
16672 int _ssub2 (int, int)
16673 long long _mpy2 (int, int)
16674 long long _smpy2 (int, int)
16675 int _add4 (int, int)
16676 int _sub4 (int, int)
16677 int _saddu4 (int, int)
16679 int _smpy (int, int)
16680 int _smpyh (int, int)
16681 int _smpyhl (int, int)
16682 int _smpylh (int, int)
16684 int _sshl (int, int)
16685 int _subc (int, int)
16687 int _avg2 (int, int)
16688 int _avgu4 (int, int)
16690 int _clrr (int, int)
16691 int _extr (int, int)
16692 int _extru (int, int)
16698 @node TILE-Gx Built-in Functions
16699 @subsection TILE-Gx Built-in Functions
16701 GCC provides intrinsics to access every instruction of the TILE-Gx
16702 processor. The intrinsics are of the form:
16706 unsigned long long __insn_@var{op} (...)
16710 Where @var{op} is the name of the instruction. Refer to the ISA manual
16711 for the complete list of instructions.
16713 GCC also provides intrinsics to directly access the network registers.
16714 The intrinsics are:
16718 unsigned long long __tile_idn0_receive (void)
16719 unsigned long long __tile_idn1_receive (void)
16720 unsigned long long __tile_udn0_receive (void)
16721 unsigned long long __tile_udn1_receive (void)
16722 unsigned long long __tile_udn2_receive (void)
16723 unsigned long long __tile_udn3_receive (void)
16724 void __tile_idn_send (unsigned long long)
16725 void __tile_udn_send (unsigned long long)
16729 The intrinsic @code{void __tile_network_barrier (void)} is used to
16730 guarantee that no network operations before it are reordered with
16733 @node TILEPro Built-in Functions
16734 @subsection TILEPro Built-in Functions
16736 GCC provides intrinsics to access every instruction of the TILEPro
16737 processor. The intrinsics are of the form:
16741 unsigned __insn_@var{op} (...)
16746 where @var{op} is the name of the instruction. Refer to the ISA manual
16747 for the complete list of instructions.
16749 GCC also provides intrinsics to directly access the network registers.
16750 The intrinsics are:
16754 unsigned __tile_idn0_receive (void)
16755 unsigned __tile_idn1_receive (void)
16756 unsigned __tile_sn_receive (void)
16757 unsigned __tile_udn0_receive (void)
16758 unsigned __tile_udn1_receive (void)
16759 unsigned __tile_udn2_receive (void)
16760 unsigned __tile_udn3_receive (void)
16761 void __tile_idn_send (unsigned)
16762 void __tile_sn_send (unsigned)
16763 void __tile_udn_send (unsigned)
16767 The intrinsic @code{void __tile_network_barrier (void)} is used to
16768 guarantee that no network operations before it are reordered with
16771 @node x86 Built-in Functions
16772 @subsection x86 Built-in Functions
16774 These built-in functions are available for the x86-32 and x86-64 family
16775 of computers, depending on the command-line switches used.
16777 If you specify command-line switches such as @option{-msse},
16778 the compiler could use the extended instruction sets even if the built-ins
16779 are not used explicitly in the program. For this reason, applications
16780 that perform run-time CPU detection must compile separate files for each
16781 supported architecture, using the appropriate flags. In particular,
16782 the file containing the CPU detection code should be compiled without
16785 The following machine modes are available for use with MMX built-in functions
16786 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
16787 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
16788 vector of eight 8-bit integers. Some of the built-in functions operate on
16789 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
16791 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
16792 of two 32-bit floating-point values.
16794 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
16795 floating-point values. Some instructions use a vector of four 32-bit
16796 integers, these use @code{V4SI}. Finally, some instructions operate on an
16797 entire vector register, interpreting it as a 128-bit integer, these use mode
16800 In 64-bit mode, the x86-64 family of processors uses additional built-in
16801 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
16802 floating point and @code{TC} 128-bit complex floating-point values.
16804 The following floating-point built-in functions are available in 64-bit
16805 mode. All of them implement the function that is part of the name.
16808 __float128 __builtin_fabsq (__float128)
16809 __float128 __builtin_copysignq (__float128, __float128)
16812 The following built-in function is always available.
16815 @item void __builtin_ia32_pause (void)
16816 Generates the @code{pause} machine instruction with a compiler memory
16820 The following floating-point built-in functions are made available in the
16824 @item __float128 __builtin_infq (void)
16825 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
16826 @findex __builtin_infq
16828 @item __float128 __builtin_huge_valq (void)
16829 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
16830 @findex __builtin_huge_valq
16833 The following built-in functions are always available and can be used to
16834 check the target platform type.
16836 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
16837 This function runs the CPU detection code to check the type of CPU and the
16838 features supported. This built-in function needs to be invoked along with the built-in functions
16839 to check CPU type and features, @code{__builtin_cpu_is} and
16840 @code{__builtin_cpu_supports}, only when used in a function that is
16841 executed before any constructors are called. The CPU detection code is
16842 automatically executed in a very high priority constructor.
16844 For example, this function has to be used in @code{ifunc} resolvers that
16845 check for CPU type using the built-in functions @code{__builtin_cpu_is}
16846 and @code{__builtin_cpu_supports}, or in constructors on targets that
16847 don't support constructor priority.
16850 static void (*resolve_memcpy (void)) (void)
16852 // ifunc resolvers fire before constructors, explicitly call the init
16854 __builtin_cpu_init ();
16855 if (__builtin_cpu_supports ("ssse3"))
16856 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
16858 return default_memcpy;
16861 void *memcpy (void *, const void *, size_t)
16862 __attribute__ ((ifunc ("resolve_memcpy")));
16867 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
16868 This function returns a positive integer if the run-time CPU
16869 is of type @var{cpuname}
16870 and returns @code{0} otherwise. The following CPU names can be detected:
16886 Intel Core i7 Nehalem CPU.
16889 Intel Core i7 Westmere CPU.
16892 Intel Core i7 Sandy Bridge CPU.
16898 AMD Family 10h CPU.
16901 AMD Family 10h Barcelona CPU.
16904 AMD Family 10h Shanghai CPU.
16907 AMD Family 10h Istanbul CPU.
16910 AMD Family 14h CPU.
16913 AMD Family 15h CPU.
16916 AMD Family 15h Bulldozer version 1.
16919 AMD Family 15h Bulldozer version 2.
16922 AMD Family 15h Bulldozer version 3.
16925 AMD Family 15h Bulldozer version 4.
16928 AMD Family 16h CPU.
16931 Here is an example:
16933 if (__builtin_cpu_is ("corei7"))
16935 do_corei7 (); // Core i7 specific implementation.
16939 do_generic (); // Generic implementation.
16944 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
16945 This function returns a positive integer if the run-time CPU
16946 supports @var{feature}
16947 and returns @code{0} otherwise. The following features can be detected:
16955 POPCNT instruction.
16963 SSSE3 instructions.
16965 SSE4.1 instructions.
16967 SSE4.2 instructions.
16973 AVX512F instructions.
16976 Here is an example:
16978 if (__builtin_cpu_supports ("popcnt"))
16980 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
16984 count = generic_countbits (n); //generic implementation.
16990 The following built-in functions are made available by @option{-mmmx}.
16991 All of them generate the machine instruction that is part of the name.
16994 v8qi __builtin_ia32_paddb (v8qi, v8qi)
16995 v4hi __builtin_ia32_paddw (v4hi, v4hi)
16996 v2si __builtin_ia32_paddd (v2si, v2si)
16997 v8qi __builtin_ia32_psubb (v8qi, v8qi)
16998 v4hi __builtin_ia32_psubw (v4hi, v4hi)
16999 v2si __builtin_ia32_psubd (v2si, v2si)
17000 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17001 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17002 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17003 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17004 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17005 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17006 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17007 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17008 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17009 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17010 di __builtin_ia32_pand (di, di)
17011 di __builtin_ia32_pandn (di,di)
17012 di __builtin_ia32_por (di, di)
17013 di __builtin_ia32_pxor (di, di)
17014 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17015 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17016 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17017 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17018 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17019 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17020 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17021 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17022 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17023 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17024 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17025 v2si __builtin_ia32_punpckldq (v2si, v2si)
17026 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17027 v4hi __builtin_ia32_packssdw (v2si, v2si)
17028 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17030 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17031 v2si __builtin_ia32_pslld (v2si, v2si)
17032 v1di __builtin_ia32_psllq (v1di, v1di)
17033 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17034 v2si __builtin_ia32_psrld (v2si, v2si)
17035 v1di __builtin_ia32_psrlq (v1di, v1di)
17036 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17037 v2si __builtin_ia32_psrad (v2si, v2si)
17038 v4hi __builtin_ia32_psllwi (v4hi, int)
17039 v2si __builtin_ia32_pslldi (v2si, int)
17040 v1di __builtin_ia32_psllqi (v1di, int)
17041 v4hi __builtin_ia32_psrlwi (v4hi, int)
17042 v2si __builtin_ia32_psrldi (v2si, int)
17043 v1di __builtin_ia32_psrlqi (v1di, int)
17044 v4hi __builtin_ia32_psrawi (v4hi, int)
17045 v2si __builtin_ia32_psradi (v2si, int)
17049 The following built-in functions are made available either with
17050 @option{-msse}, or with a combination of @option{-m3dnow} and
17051 @option{-march=athlon}. All of them generate the machine
17052 instruction that is part of the name.
17055 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17056 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17057 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17058 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17059 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17060 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17061 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17062 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17063 int __builtin_ia32_pmovmskb (v8qi)
17064 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17065 void __builtin_ia32_movntq (di *, di)
17066 void __builtin_ia32_sfence (void)
17069 The following built-in functions are available when @option{-msse} is used.
17070 All of them generate the machine instruction that is part of the name.
17073 int __builtin_ia32_comieq (v4sf, v4sf)
17074 int __builtin_ia32_comineq (v4sf, v4sf)
17075 int __builtin_ia32_comilt (v4sf, v4sf)
17076 int __builtin_ia32_comile (v4sf, v4sf)
17077 int __builtin_ia32_comigt (v4sf, v4sf)
17078 int __builtin_ia32_comige (v4sf, v4sf)
17079 int __builtin_ia32_ucomieq (v4sf, v4sf)
17080 int __builtin_ia32_ucomineq (v4sf, v4sf)
17081 int __builtin_ia32_ucomilt (v4sf, v4sf)
17082 int __builtin_ia32_ucomile (v4sf, v4sf)
17083 int __builtin_ia32_ucomigt (v4sf, v4sf)
17084 int __builtin_ia32_ucomige (v4sf, v4sf)
17085 v4sf __builtin_ia32_addps (v4sf, v4sf)
17086 v4sf __builtin_ia32_subps (v4sf, v4sf)
17087 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17088 v4sf __builtin_ia32_divps (v4sf, v4sf)
17089 v4sf __builtin_ia32_addss (v4sf, v4sf)
17090 v4sf __builtin_ia32_subss (v4sf, v4sf)
17091 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17092 v4sf __builtin_ia32_divss (v4sf, v4sf)
17093 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17094 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17095 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17096 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17097 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17098 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17099 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17100 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17101 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17102 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17103 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17104 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17105 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17106 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17107 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17108 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17109 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17110 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17111 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17112 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17113 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17114 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17115 v4sf __builtin_ia32_minps (v4sf, v4sf)
17116 v4sf __builtin_ia32_minss (v4sf, v4sf)
17117 v4sf __builtin_ia32_andps (v4sf, v4sf)
17118 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17119 v4sf __builtin_ia32_orps (v4sf, v4sf)
17120 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17121 v4sf __builtin_ia32_movss (v4sf, v4sf)
17122 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17123 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17124 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17125 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17126 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17127 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17128 v2si __builtin_ia32_cvtps2pi (v4sf)
17129 int __builtin_ia32_cvtss2si (v4sf)
17130 v2si __builtin_ia32_cvttps2pi (v4sf)
17131 int __builtin_ia32_cvttss2si (v4sf)
17132 v4sf __builtin_ia32_rcpps (v4sf)
17133 v4sf __builtin_ia32_rsqrtps (v4sf)
17134 v4sf __builtin_ia32_sqrtps (v4sf)
17135 v4sf __builtin_ia32_rcpss (v4sf)
17136 v4sf __builtin_ia32_rsqrtss (v4sf)
17137 v4sf __builtin_ia32_sqrtss (v4sf)
17138 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17139 void __builtin_ia32_movntps (float *, v4sf)
17140 int __builtin_ia32_movmskps (v4sf)
17143 The following built-in functions are available when @option{-msse} is used.
17146 @item v4sf __builtin_ia32_loadups (float *)
17147 Generates the @code{movups} machine instruction as a load from memory.
17148 @item void __builtin_ia32_storeups (float *, v4sf)
17149 Generates the @code{movups} machine instruction as a store to memory.
17150 @item v4sf __builtin_ia32_loadss (float *)
17151 Generates the @code{movss} machine instruction as a load from memory.
17152 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17153 Generates the @code{movhps} machine instruction as a load from memory.
17154 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17155 Generates the @code{movlps} machine instruction as a load from memory
17156 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17157 Generates the @code{movhps} machine instruction as a store to memory.
17158 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17159 Generates the @code{movlps} machine instruction as a store to memory.
17162 The following built-in functions are available when @option{-msse2} is used.
17163 All of them generate the machine instruction that is part of the name.
17166 int __builtin_ia32_comisdeq (v2df, v2df)
17167 int __builtin_ia32_comisdlt (v2df, v2df)
17168 int __builtin_ia32_comisdle (v2df, v2df)
17169 int __builtin_ia32_comisdgt (v2df, v2df)
17170 int __builtin_ia32_comisdge (v2df, v2df)
17171 int __builtin_ia32_comisdneq (v2df, v2df)
17172 int __builtin_ia32_ucomisdeq (v2df, v2df)
17173 int __builtin_ia32_ucomisdlt (v2df, v2df)
17174 int __builtin_ia32_ucomisdle (v2df, v2df)
17175 int __builtin_ia32_ucomisdgt (v2df, v2df)
17176 int __builtin_ia32_ucomisdge (v2df, v2df)
17177 int __builtin_ia32_ucomisdneq (v2df, v2df)
17178 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17179 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17180 v2df __builtin_ia32_cmplepd (v2df, v2df)
17181 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17182 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17183 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17184 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17185 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17186 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17187 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17188 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17189 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17190 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17191 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17192 v2df __builtin_ia32_cmplesd (v2df, v2df)
17193 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17194 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17195 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17196 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17197 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17198 v2di __builtin_ia32_paddq (v2di, v2di)
17199 v2di __builtin_ia32_psubq (v2di, v2di)
17200 v2df __builtin_ia32_addpd (v2df, v2df)
17201 v2df __builtin_ia32_subpd (v2df, v2df)
17202 v2df __builtin_ia32_mulpd (v2df, v2df)
17203 v2df __builtin_ia32_divpd (v2df, v2df)
17204 v2df __builtin_ia32_addsd (v2df, v2df)
17205 v2df __builtin_ia32_subsd (v2df, v2df)
17206 v2df __builtin_ia32_mulsd (v2df, v2df)
17207 v2df __builtin_ia32_divsd (v2df, v2df)
17208 v2df __builtin_ia32_minpd (v2df, v2df)
17209 v2df __builtin_ia32_maxpd (v2df, v2df)
17210 v2df __builtin_ia32_minsd (v2df, v2df)
17211 v2df __builtin_ia32_maxsd (v2df, v2df)
17212 v2df __builtin_ia32_andpd (v2df, v2df)
17213 v2df __builtin_ia32_andnpd (v2df, v2df)
17214 v2df __builtin_ia32_orpd (v2df, v2df)
17215 v2df __builtin_ia32_xorpd (v2df, v2df)
17216 v2df __builtin_ia32_movsd (v2df, v2df)
17217 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17218 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17219 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17220 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17221 v4si __builtin_ia32_paddd128 (v4si, v4si)
17222 v2di __builtin_ia32_paddq128 (v2di, v2di)
17223 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17224 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17225 v4si __builtin_ia32_psubd128 (v4si, v4si)
17226 v2di __builtin_ia32_psubq128 (v2di, v2di)
17227 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17228 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17229 v2di __builtin_ia32_pand128 (v2di, v2di)
17230 v2di __builtin_ia32_pandn128 (v2di, v2di)
17231 v2di __builtin_ia32_por128 (v2di, v2di)
17232 v2di __builtin_ia32_pxor128 (v2di, v2di)
17233 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17234 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17235 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17236 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17237 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17238 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17239 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17240 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17241 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17242 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17243 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17244 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17245 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17246 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17247 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17248 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17249 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17250 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17251 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17252 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17253 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17254 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17255 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17256 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17257 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17258 v2df __builtin_ia32_loadupd (double *)
17259 void __builtin_ia32_storeupd (double *, v2df)
17260 v2df __builtin_ia32_loadhpd (v2df, double const *)
17261 v2df __builtin_ia32_loadlpd (v2df, double const *)
17262 int __builtin_ia32_movmskpd (v2df)
17263 int __builtin_ia32_pmovmskb128 (v16qi)
17264 void __builtin_ia32_movnti (int *, int)
17265 void __builtin_ia32_movnti64 (long long int *, long long int)
17266 void __builtin_ia32_movntpd (double *, v2df)
17267 void __builtin_ia32_movntdq (v2df *, v2df)
17268 v4si __builtin_ia32_pshufd (v4si, int)
17269 v8hi __builtin_ia32_pshuflw (v8hi, int)
17270 v8hi __builtin_ia32_pshufhw (v8hi, int)
17271 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17272 v2df __builtin_ia32_sqrtpd (v2df)
17273 v2df __builtin_ia32_sqrtsd (v2df)
17274 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17275 v2df __builtin_ia32_cvtdq2pd (v4si)
17276 v4sf __builtin_ia32_cvtdq2ps (v4si)
17277 v4si __builtin_ia32_cvtpd2dq (v2df)
17278 v2si __builtin_ia32_cvtpd2pi (v2df)
17279 v4sf __builtin_ia32_cvtpd2ps (v2df)
17280 v4si __builtin_ia32_cvttpd2dq (v2df)
17281 v2si __builtin_ia32_cvttpd2pi (v2df)
17282 v2df __builtin_ia32_cvtpi2pd (v2si)
17283 int __builtin_ia32_cvtsd2si (v2df)
17284 int __builtin_ia32_cvttsd2si (v2df)
17285 long long __builtin_ia32_cvtsd2si64 (v2df)
17286 long long __builtin_ia32_cvttsd2si64 (v2df)
17287 v4si __builtin_ia32_cvtps2dq (v4sf)
17288 v2df __builtin_ia32_cvtps2pd (v4sf)
17289 v4si __builtin_ia32_cvttps2dq (v4sf)
17290 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17291 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17292 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17293 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17294 void __builtin_ia32_clflush (const void *)
17295 void __builtin_ia32_lfence (void)
17296 void __builtin_ia32_mfence (void)
17297 v16qi __builtin_ia32_loaddqu (const char *)
17298 void __builtin_ia32_storedqu (char *, v16qi)
17299 v1di __builtin_ia32_pmuludq (v2si, v2si)
17300 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17301 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17302 v4si __builtin_ia32_pslld128 (v4si, v4si)
17303 v2di __builtin_ia32_psllq128 (v2di, v2di)
17304 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17305 v4si __builtin_ia32_psrld128 (v4si, v4si)
17306 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17307 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17308 v4si __builtin_ia32_psrad128 (v4si, v4si)
17309 v2di __builtin_ia32_pslldqi128 (v2di, int)
17310 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17311 v4si __builtin_ia32_pslldi128 (v4si, int)
17312 v2di __builtin_ia32_psllqi128 (v2di, int)
17313 v2di __builtin_ia32_psrldqi128 (v2di, int)
17314 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17315 v4si __builtin_ia32_psrldi128 (v4si, int)
17316 v2di __builtin_ia32_psrlqi128 (v2di, int)
17317 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17318 v4si __builtin_ia32_psradi128 (v4si, int)
17319 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17320 v2di __builtin_ia32_movq128 (v2di)
17323 The following built-in functions are available when @option{-msse3} is used.
17324 All of them generate the machine instruction that is part of the name.
17327 v2df __builtin_ia32_addsubpd (v2df, v2df)
17328 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17329 v2df __builtin_ia32_haddpd (v2df, v2df)
17330 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17331 v2df __builtin_ia32_hsubpd (v2df, v2df)
17332 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17333 v16qi __builtin_ia32_lddqu (char const *)
17334 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17335 v4sf __builtin_ia32_movshdup (v4sf)
17336 v4sf __builtin_ia32_movsldup (v4sf)
17337 void __builtin_ia32_mwait (unsigned int, unsigned int)
17340 The following built-in functions are available when @option{-mssse3} is used.
17341 All of them generate the machine instruction that is part of the name.
17344 v2si __builtin_ia32_phaddd (v2si, v2si)
17345 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17346 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17347 v2si __builtin_ia32_phsubd (v2si, v2si)
17348 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17349 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17350 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17351 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17352 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17353 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17354 v2si __builtin_ia32_psignd (v2si, v2si)
17355 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17356 v1di __builtin_ia32_palignr (v1di, v1di, int)
17357 v8qi __builtin_ia32_pabsb (v8qi)
17358 v2si __builtin_ia32_pabsd (v2si)
17359 v4hi __builtin_ia32_pabsw (v4hi)
17362 The following built-in functions are available when @option{-mssse3} is used.
17363 All of them generate the machine instruction that is part of the name.
17366 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17367 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17368 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17369 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17370 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17371 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17372 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17373 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17374 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17375 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17376 v4si __builtin_ia32_psignd128 (v4si, v4si)
17377 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17378 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17379 v16qi __builtin_ia32_pabsb128 (v16qi)
17380 v4si __builtin_ia32_pabsd128 (v4si)
17381 v8hi __builtin_ia32_pabsw128 (v8hi)
17384 The following built-in functions are available when @option{-msse4.1} is
17385 used. All of them generate the machine instruction that is part of the
17389 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17390 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17391 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17392 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17393 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17394 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17395 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17396 v2di __builtin_ia32_movntdqa (v2di *);
17397 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17398 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17399 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17400 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17401 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17402 v8hi __builtin_ia32_phminposuw128 (v8hi)
17403 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17404 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17405 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17406 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17407 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17408 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17409 v4si __builtin_ia32_pminud128 (v4si, v4si)
17410 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17411 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17412 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17413 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17414 v2di __builtin_ia32_pmovsxdq128 (v4si)
17415 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17416 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17417 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17418 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17419 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17420 v2di __builtin_ia32_pmovzxdq128 (v4si)
17421 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17422 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17423 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17424 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17425 int __builtin_ia32_ptestc128 (v2di, v2di)
17426 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17427 int __builtin_ia32_ptestz128 (v2di, v2di)
17428 v2df __builtin_ia32_roundpd (v2df, const int)
17429 v4sf __builtin_ia32_roundps (v4sf, const int)
17430 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17431 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17434 The following built-in functions are available when @option{-msse4.1} is
17438 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17439 Generates the @code{insertps} machine instruction.
17440 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17441 Generates the @code{pextrb} machine instruction.
17442 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17443 Generates the @code{pinsrb} machine instruction.
17444 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17445 Generates the @code{pinsrd} machine instruction.
17446 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17447 Generates the @code{pinsrq} machine instruction in 64bit mode.
17450 The following built-in functions are changed to generate new SSE4.1
17451 instructions when @option{-msse4.1} is used.
17454 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17455 Generates the @code{extractps} machine instruction.
17456 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17457 Generates the @code{pextrd} machine instruction.
17458 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17459 Generates the @code{pextrq} machine instruction in 64bit mode.
17462 The following built-in functions are available when @option{-msse4.2} is
17463 used. All of them generate the machine instruction that is part of the
17467 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17468 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17469 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17470 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17471 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17472 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17473 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17474 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17475 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17476 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17477 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17478 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17479 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17480 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17481 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17484 The following built-in functions are available when @option{-msse4.2} is
17488 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17489 Generates the @code{crc32b} machine instruction.
17490 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17491 Generates the @code{crc32w} machine instruction.
17492 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17493 Generates the @code{crc32l} machine instruction.
17494 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17495 Generates the @code{crc32q} machine instruction.
17498 The following built-in functions are changed to generate new SSE4.2
17499 instructions when @option{-msse4.2} is used.
17502 @item int __builtin_popcount (unsigned int)
17503 Generates the @code{popcntl} machine instruction.
17504 @item int __builtin_popcountl (unsigned long)
17505 Generates the @code{popcntl} or @code{popcntq} machine instruction,
17506 depending on the size of @code{unsigned long}.
17507 @item int __builtin_popcountll (unsigned long long)
17508 Generates the @code{popcntq} machine instruction.
17511 The following built-in functions are available when @option{-mavx} is
17512 used. All of them generate the machine instruction that is part of the
17516 v4df __builtin_ia32_addpd256 (v4df,v4df)
17517 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
17518 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
17519 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
17520 v4df __builtin_ia32_andnpd256 (v4df,v4df)
17521 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
17522 v4df __builtin_ia32_andpd256 (v4df,v4df)
17523 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
17524 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
17525 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
17526 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
17527 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
17528 v2df __builtin_ia32_cmppd (v2df,v2df,int)
17529 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
17530 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
17531 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
17532 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
17533 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
17534 v4df __builtin_ia32_cvtdq2pd256 (v4si)
17535 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
17536 v4si __builtin_ia32_cvtpd2dq256 (v4df)
17537 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
17538 v8si __builtin_ia32_cvtps2dq256 (v8sf)
17539 v4df __builtin_ia32_cvtps2pd256 (v4sf)
17540 v4si __builtin_ia32_cvttpd2dq256 (v4df)
17541 v8si __builtin_ia32_cvttps2dq256 (v8sf)
17542 v4df __builtin_ia32_divpd256 (v4df,v4df)
17543 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
17544 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
17545 v4df __builtin_ia32_haddpd256 (v4df,v4df)
17546 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
17547 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
17548 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
17549 v32qi __builtin_ia32_lddqu256 (pcchar)
17550 v32qi __builtin_ia32_loaddqu256 (pcchar)
17551 v4df __builtin_ia32_loadupd256 (pcdouble)
17552 v8sf __builtin_ia32_loadups256 (pcfloat)
17553 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
17554 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
17555 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
17556 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
17557 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
17558 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
17559 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
17560 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
17561 v4df __builtin_ia32_maxpd256 (v4df,v4df)
17562 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
17563 v4df __builtin_ia32_minpd256 (v4df,v4df)
17564 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
17565 v4df __builtin_ia32_movddup256 (v4df)
17566 int __builtin_ia32_movmskpd256 (v4df)
17567 int __builtin_ia32_movmskps256 (v8sf)
17568 v8sf __builtin_ia32_movshdup256 (v8sf)
17569 v8sf __builtin_ia32_movsldup256 (v8sf)
17570 v4df __builtin_ia32_mulpd256 (v4df,v4df)
17571 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
17572 v4df __builtin_ia32_orpd256 (v4df,v4df)
17573 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
17574 v2df __builtin_ia32_pd_pd256 (v4df)
17575 v4df __builtin_ia32_pd256_pd (v2df)
17576 v4sf __builtin_ia32_ps_ps256 (v8sf)
17577 v8sf __builtin_ia32_ps256_ps (v4sf)
17578 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
17579 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
17580 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
17581 v8sf __builtin_ia32_rcpps256 (v8sf)
17582 v4df __builtin_ia32_roundpd256 (v4df,int)
17583 v8sf __builtin_ia32_roundps256 (v8sf,int)
17584 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
17585 v8sf __builtin_ia32_rsqrtps256 (v8sf)
17586 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
17587 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
17588 v4si __builtin_ia32_si_si256 (v8si)
17589 v8si __builtin_ia32_si256_si (v4si)
17590 v4df __builtin_ia32_sqrtpd256 (v4df)
17591 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
17592 v8sf __builtin_ia32_sqrtps256 (v8sf)
17593 void __builtin_ia32_storedqu256 (pchar,v32qi)
17594 void __builtin_ia32_storeupd256 (pdouble,v4df)
17595 void __builtin_ia32_storeups256 (pfloat,v8sf)
17596 v4df __builtin_ia32_subpd256 (v4df,v4df)
17597 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
17598 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
17599 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
17600 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
17601 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
17602 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
17603 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
17604 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
17605 v4sf __builtin_ia32_vbroadcastss (pcfloat)
17606 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
17607 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
17608 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
17609 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
17610 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
17611 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
17612 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
17613 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
17614 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
17615 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
17616 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
17617 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
17618 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
17619 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
17620 v2df __builtin_ia32_vpermilpd (v2df,int)
17621 v4df __builtin_ia32_vpermilpd256 (v4df,int)
17622 v4sf __builtin_ia32_vpermilps (v4sf,int)
17623 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
17624 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
17625 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
17626 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
17627 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
17628 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
17629 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
17630 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
17631 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
17632 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
17633 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
17634 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
17635 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
17636 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
17637 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
17638 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
17639 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
17640 void __builtin_ia32_vzeroall (void)
17641 void __builtin_ia32_vzeroupper (void)
17642 v4df __builtin_ia32_xorpd256 (v4df,v4df)
17643 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
17646 The following built-in functions are available when @option{-mavx2} is
17647 used. All of them generate the machine instruction that is part of the
17651 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
17652 v32qi __builtin_ia32_pabsb256 (v32qi)
17653 v16hi __builtin_ia32_pabsw256 (v16hi)
17654 v8si __builtin_ia32_pabsd256 (v8si)
17655 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
17656 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
17657 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
17658 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
17659 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
17660 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
17661 v8si __builtin_ia32_paddd256 (v8si,v8si)
17662 v4di __builtin_ia32_paddq256 (v4di,v4di)
17663 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
17664 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
17665 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
17666 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
17667 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
17668 v4di __builtin_ia32_andsi256 (v4di,v4di)
17669 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
17670 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
17671 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
17672 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
17673 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
17674 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
17675 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
17676 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
17677 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
17678 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
17679 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
17680 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
17681 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
17682 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
17683 v8si __builtin_ia32_phaddd256 (v8si,v8si)
17684 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
17685 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
17686 v8si __builtin_ia32_phsubd256 (v8si,v8si)
17687 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
17688 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
17689 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
17690 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
17691 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
17692 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
17693 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
17694 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
17695 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
17696 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
17697 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
17698 v8si __builtin_ia32_pminsd256 (v8si,v8si)
17699 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
17700 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
17701 v8si __builtin_ia32_pminud256 (v8si,v8si)
17702 int __builtin_ia32_pmovmskb256 (v32qi)
17703 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
17704 v8si __builtin_ia32_pmovsxbd256 (v16qi)
17705 v4di __builtin_ia32_pmovsxbq256 (v16qi)
17706 v8si __builtin_ia32_pmovsxwd256 (v8hi)
17707 v4di __builtin_ia32_pmovsxwq256 (v8hi)
17708 v4di __builtin_ia32_pmovsxdq256 (v4si)
17709 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
17710 v8si __builtin_ia32_pmovzxbd256 (v16qi)
17711 v4di __builtin_ia32_pmovzxbq256 (v16qi)
17712 v8si __builtin_ia32_pmovzxwd256 (v8hi)
17713 v4di __builtin_ia32_pmovzxwq256 (v8hi)
17714 v4di __builtin_ia32_pmovzxdq256 (v4si)
17715 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
17716 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
17717 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
17718 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
17719 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
17720 v8si __builtin_ia32_pmulld256 (v8si,v8si)
17721 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
17722 v4di __builtin_ia32_por256 (v4di,v4di)
17723 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
17724 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
17725 v8si __builtin_ia32_pshufd256 (v8si,int)
17726 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
17727 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
17728 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
17729 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
17730 v8si __builtin_ia32_psignd256 (v8si,v8si)
17731 v4di __builtin_ia32_pslldqi256 (v4di,int)
17732 v16hi __builtin_ia32_psllwi256 (16hi,int)
17733 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
17734 v8si __builtin_ia32_pslldi256 (v8si,int)
17735 v8si __builtin_ia32_pslld256(v8si,v4si)
17736 v4di __builtin_ia32_psllqi256 (v4di,int)
17737 v4di __builtin_ia32_psllq256(v4di,v2di)
17738 v16hi __builtin_ia32_psrawi256 (v16hi,int)
17739 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
17740 v8si __builtin_ia32_psradi256 (v8si,int)
17741 v8si __builtin_ia32_psrad256 (v8si,v4si)
17742 v4di __builtin_ia32_psrldqi256 (v4di, int)
17743 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
17744 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
17745 v8si __builtin_ia32_psrldi256 (v8si,int)
17746 v8si __builtin_ia32_psrld256 (v8si,v4si)
17747 v4di __builtin_ia32_psrlqi256 (v4di,int)
17748 v4di __builtin_ia32_psrlq256(v4di,v2di)
17749 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
17750 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
17751 v8si __builtin_ia32_psubd256 (v8si,v8si)
17752 v4di __builtin_ia32_psubq256 (v4di,v4di)
17753 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
17754 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
17755 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
17756 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
17757 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
17758 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
17759 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
17760 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
17761 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
17762 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
17763 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
17764 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
17765 v4di __builtin_ia32_pxor256 (v4di,v4di)
17766 v4di __builtin_ia32_movntdqa256 (pv4di)
17767 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
17768 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
17769 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
17770 v4di __builtin_ia32_vbroadcastsi256 (v2di)
17771 v4si __builtin_ia32_pblendd128 (v4si,v4si)
17772 v8si __builtin_ia32_pblendd256 (v8si,v8si)
17773 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
17774 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
17775 v8si __builtin_ia32_pbroadcastd256 (v4si)
17776 v4di __builtin_ia32_pbroadcastq256 (v2di)
17777 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
17778 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
17779 v4si __builtin_ia32_pbroadcastd128 (v4si)
17780 v2di __builtin_ia32_pbroadcastq128 (v2di)
17781 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
17782 v4df __builtin_ia32_permdf256 (v4df,int)
17783 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
17784 v4di __builtin_ia32_permdi256 (v4di,int)
17785 v4di __builtin_ia32_permti256 (v4di,v4di,int)
17786 v4di __builtin_ia32_extract128i256 (v4di,int)
17787 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
17788 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
17789 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
17790 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
17791 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
17792 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
17793 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
17794 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
17795 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
17796 v8si __builtin_ia32_psllv8si (v8si,v8si)
17797 v4si __builtin_ia32_psllv4si (v4si,v4si)
17798 v4di __builtin_ia32_psllv4di (v4di,v4di)
17799 v2di __builtin_ia32_psllv2di (v2di,v2di)
17800 v8si __builtin_ia32_psrav8si (v8si,v8si)
17801 v4si __builtin_ia32_psrav4si (v4si,v4si)
17802 v8si __builtin_ia32_psrlv8si (v8si,v8si)
17803 v4si __builtin_ia32_psrlv4si (v4si,v4si)
17804 v4di __builtin_ia32_psrlv4di (v4di,v4di)
17805 v2di __builtin_ia32_psrlv2di (v2di,v2di)
17806 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
17807 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
17808 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
17809 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
17810 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
17811 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
17812 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
17813 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
17814 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
17815 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
17816 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
17817 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
17818 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
17819 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
17820 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
17821 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
17824 The following built-in functions are available when @option{-maes} is
17825 used. All of them generate the machine instruction that is part of the
17829 v2di __builtin_ia32_aesenc128 (v2di, v2di)
17830 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
17831 v2di __builtin_ia32_aesdec128 (v2di, v2di)
17832 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
17833 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
17834 v2di __builtin_ia32_aesimc128 (v2di)
17837 The following built-in function is available when @option{-mpclmul} is
17841 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
17842 Generates the @code{pclmulqdq} machine instruction.
17845 The following built-in function is available when @option{-mfsgsbase} is
17846 used. All of them generate the machine instruction that is part of the
17850 unsigned int __builtin_ia32_rdfsbase32 (void)
17851 unsigned long long __builtin_ia32_rdfsbase64 (void)
17852 unsigned int __builtin_ia32_rdgsbase32 (void)
17853 unsigned long long __builtin_ia32_rdgsbase64 (void)
17854 void _writefsbase_u32 (unsigned int)
17855 void _writefsbase_u64 (unsigned long long)
17856 void _writegsbase_u32 (unsigned int)
17857 void _writegsbase_u64 (unsigned long long)
17860 The following built-in function is available when @option{-mrdrnd} is
17861 used. All of them generate the machine instruction that is part of the
17865 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
17866 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
17867 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
17870 The following built-in functions are available when @option{-msse4a} is used.
17871 All of them generate the machine instruction that is part of the name.
17874 void __builtin_ia32_movntsd (double *, v2df)
17875 void __builtin_ia32_movntss (float *, v4sf)
17876 v2di __builtin_ia32_extrq (v2di, v16qi)
17877 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
17878 v2di __builtin_ia32_insertq (v2di, v2di)
17879 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
17882 The following built-in functions are available when @option{-mxop} is used.
17884 v2df __builtin_ia32_vfrczpd (v2df)
17885 v4sf __builtin_ia32_vfrczps (v4sf)
17886 v2df __builtin_ia32_vfrczsd (v2df)
17887 v4sf __builtin_ia32_vfrczss (v4sf)
17888 v4df __builtin_ia32_vfrczpd256 (v4df)
17889 v8sf __builtin_ia32_vfrczps256 (v8sf)
17890 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
17891 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
17892 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
17893 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
17894 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
17895 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
17896 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
17897 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
17898 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
17899 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
17900 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
17901 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
17902 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
17903 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
17904 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
17905 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
17906 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
17907 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
17908 v4si __builtin_ia32_vpcomequd (v4si, v4si)
17909 v2di __builtin_ia32_vpcomequq (v2di, v2di)
17910 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
17911 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
17912 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
17913 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
17914 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
17915 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
17916 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
17917 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
17918 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
17919 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
17920 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
17921 v4si __builtin_ia32_vpcomged (v4si, v4si)
17922 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
17923 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
17924 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
17925 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
17926 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
17927 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
17928 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
17929 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
17930 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
17931 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
17932 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
17933 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
17934 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
17935 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
17936 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
17937 v4si __builtin_ia32_vpcomled (v4si, v4si)
17938 v2di __builtin_ia32_vpcomleq (v2di, v2di)
17939 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
17940 v4si __builtin_ia32_vpcomleud (v4si, v4si)
17941 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
17942 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
17943 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
17944 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
17945 v4si __builtin_ia32_vpcomltd (v4si, v4si)
17946 v2di __builtin_ia32_vpcomltq (v2di, v2di)
17947 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
17948 v4si __builtin_ia32_vpcomltud (v4si, v4si)
17949 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
17950 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
17951 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
17952 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
17953 v4si __builtin_ia32_vpcomned (v4si, v4si)
17954 v2di __builtin_ia32_vpcomneq (v2di, v2di)
17955 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
17956 v4si __builtin_ia32_vpcomneud (v4si, v4si)
17957 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
17958 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
17959 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
17960 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
17961 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
17962 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
17963 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
17964 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
17965 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
17966 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
17967 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
17968 v4si __builtin_ia32_vphaddbd (v16qi)
17969 v2di __builtin_ia32_vphaddbq (v16qi)
17970 v8hi __builtin_ia32_vphaddbw (v16qi)
17971 v2di __builtin_ia32_vphadddq (v4si)
17972 v4si __builtin_ia32_vphaddubd (v16qi)
17973 v2di __builtin_ia32_vphaddubq (v16qi)
17974 v8hi __builtin_ia32_vphaddubw (v16qi)
17975 v2di __builtin_ia32_vphaddudq (v4si)
17976 v4si __builtin_ia32_vphadduwd (v8hi)
17977 v2di __builtin_ia32_vphadduwq (v8hi)
17978 v4si __builtin_ia32_vphaddwd (v8hi)
17979 v2di __builtin_ia32_vphaddwq (v8hi)
17980 v8hi __builtin_ia32_vphsubbw (v16qi)
17981 v2di __builtin_ia32_vphsubdq (v4si)
17982 v4si __builtin_ia32_vphsubwd (v8hi)
17983 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
17984 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
17985 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
17986 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
17987 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
17988 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
17989 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
17990 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
17991 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
17992 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
17993 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
17994 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
17995 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
17996 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
17997 v4si __builtin_ia32_vprotd (v4si, v4si)
17998 v2di __builtin_ia32_vprotq (v2di, v2di)
17999 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18000 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18001 v4si __builtin_ia32_vpshad (v4si, v4si)
18002 v2di __builtin_ia32_vpshaq (v2di, v2di)
18003 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18004 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18005 v4si __builtin_ia32_vpshld (v4si, v4si)
18006 v2di __builtin_ia32_vpshlq (v2di, v2di)
18007 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18010 The following built-in functions are available when @option{-mfma4} is used.
18011 All of them generate the machine instruction that is part of the name.
18014 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18015 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18016 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18017 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18018 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18019 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18020 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18021 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18022 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18023 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18024 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18025 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18026 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18027 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18028 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18029 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18030 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18031 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18032 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18033 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18034 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18035 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18036 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18037 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18038 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18039 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18040 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18041 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18042 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18043 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18044 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18045 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18049 The following built-in functions are available when @option{-mlwp} is used.
18052 void __builtin_ia32_llwpcb16 (void *);
18053 void __builtin_ia32_llwpcb32 (void *);
18054 void __builtin_ia32_llwpcb64 (void *);
18055 void * __builtin_ia32_llwpcb16 (void);
18056 void * __builtin_ia32_llwpcb32 (void);
18057 void * __builtin_ia32_llwpcb64 (void);
18058 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18059 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18060 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18061 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18062 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18063 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18066 The following built-in functions are available when @option{-mbmi} is used.
18067 All of them generate the machine instruction that is part of the name.
18069 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18070 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18073 The following built-in functions are available when @option{-mbmi2} is used.
18074 All of them generate the machine instruction that is part of the name.
18076 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18077 unsigned int _pdep_u32 (unsigned int, unsigned int)
18078 unsigned int _pext_u32 (unsigned int, unsigned int)
18079 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18080 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18081 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18084 The following built-in functions are available when @option{-mlzcnt} is used.
18085 All of them generate the machine instruction that is part of the name.
18087 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18088 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18089 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18092 The following built-in functions are available when @option{-mfxsr} is used.
18093 All of them generate the machine instruction that is part of the name.
18095 void __builtin_ia32_fxsave (void *)
18096 void __builtin_ia32_fxrstor (void *)
18097 void __builtin_ia32_fxsave64 (void *)
18098 void __builtin_ia32_fxrstor64 (void *)
18101 The following built-in functions are available when @option{-mxsave} is used.
18102 All of them generate the machine instruction that is part of the name.
18104 void __builtin_ia32_xsave (void *, long long)
18105 void __builtin_ia32_xrstor (void *, long long)
18106 void __builtin_ia32_xsave64 (void *, long long)
18107 void __builtin_ia32_xrstor64 (void *, long long)
18110 The following built-in functions are available when @option{-mxsaveopt} is used.
18111 All of them generate the machine instruction that is part of the name.
18113 void __builtin_ia32_xsaveopt (void *, long long)
18114 void __builtin_ia32_xsaveopt64 (void *, long long)
18117 The following built-in functions are available when @option{-mtbm} is used.
18118 Both of them generate the immediate form of the bextr machine instruction.
18120 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18121 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18125 The following built-in functions are available when @option{-m3dnow} is used.
18126 All of them generate the machine instruction that is part of the name.
18129 void __builtin_ia32_femms (void)
18130 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18131 v2si __builtin_ia32_pf2id (v2sf)
18132 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18133 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18134 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18135 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18136 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18137 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18138 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18139 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18140 v2sf __builtin_ia32_pfrcp (v2sf)
18141 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18142 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18143 v2sf __builtin_ia32_pfrsqrt (v2sf)
18144 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18145 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18146 v2sf __builtin_ia32_pi2fd (v2si)
18147 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18150 The following built-in functions are available when both @option{-m3dnow}
18151 and @option{-march=athlon} are used. All of them generate the machine
18152 instruction that is part of the name.
18155 v2si __builtin_ia32_pf2iw (v2sf)
18156 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18157 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18158 v2sf __builtin_ia32_pi2fw (v2si)
18159 v2sf __builtin_ia32_pswapdsf (v2sf)
18160 v2si __builtin_ia32_pswapdsi (v2si)
18163 The following built-in functions are available when @option{-mrtm} is used
18164 They are used for restricted transactional memory. These are the internal
18165 low level functions. Normally the functions in
18166 @ref{x86 transactional memory intrinsics} should be used instead.
18169 int __builtin_ia32_xbegin ()
18170 void __builtin_ia32_xend ()
18171 void __builtin_ia32_xabort (status)
18172 int __builtin_ia32_xtest ()
18175 The following built-in functions are available when @option{-mmwaitx} is used.
18176 All of them generate the machine instruction that is part of the name.
18178 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18179 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18182 @node x86 transactional memory intrinsics
18183 @subsection x86 Transactional Memory Intrinsics
18185 These hardware transactional memory intrinsics for x86 allow you to use
18186 memory transactions with RTM (Restricted Transactional Memory).
18187 This support is enabled with the @option{-mrtm} option.
18188 For using HLE (Hardware Lock Elision) see
18189 @ref{x86 specific memory model extensions for transactional memory} instead.
18191 A memory transaction commits all changes to memory in an atomic way,
18192 as visible to other threads. If the transaction fails it is rolled back
18193 and all side effects discarded.
18195 Generally there is no guarantee that a memory transaction ever succeeds
18196 and suitable fallback code always needs to be supplied.
18198 @deftypefn {RTM Function} {unsigned} _xbegin ()
18199 Start a RTM (Restricted Transactional Memory) transaction.
18200 Returns @code{_XBEGIN_STARTED} when the transaction
18201 started successfully (note this is not 0, so the constant has to be
18202 explicitly tested).
18204 If the transaction aborts, all side-effects
18205 are undone and an abort code encoded as a bit mask is returned.
18206 The following macros are defined:
18209 @item _XABORT_EXPLICIT
18210 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18211 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18212 @item _XABORT_RETRY
18213 Transaction retry is possible.
18214 @item _XABORT_CONFLICT
18215 Transaction abort due to a memory conflict with another thread.
18216 @item _XABORT_CAPACITY
18217 Transaction abort due to the transaction using too much memory.
18218 @item _XABORT_DEBUG
18219 Transaction abort due to a debug trap.
18220 @item _XABORT_NESTED
18221 Transaction abort in an inner nested transaction.
18224 There is no guarantee
18225 any transaction ever succeeds, so there always needs to be a valid
18229 @deftypefn {RTM Function} {void} _xend ()
18230 Commit the current transaction. When no transaction is active this faults.
18231 All memory side-effects of the transaction become visible
18232 to other threads in an atomic manner.
18235 @deftypefn {RTM Function} {int} _xtest ()
18236 Return a nonzero value if a transaction is currently active, otherwise 0.
18239 @deftypefn {RTM Function} {void} _xabort (status)
18240 Abort the current transaction. When no transaction is active this is a no-op.
18241 The @var{status} is an 8-bit constant; its value is encoded in the return
18242 value from @code{_xbegin}.
18245 Here is an example showing handling for @code{_XABORT_RETRY}
18246 and a fallback path for other failures:
18249 #include <immintrin.h>
18251 int n_tries, max_tries;
18252 unsigned status = _XABORT_EXPLICIT;
18255 for (n_tries = 0; n_tries < max_tries; n_tries++)
18257 status = _xbegin ();
18258 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18261 if (status == _XBEGIN_STARTED)
18263 ... transaction code...
18268 ... non-transactional fallback path...
18273 Note that, in most cases, the transactional and non-transactional code
18274 must synchronize together to ensure consistency.
18276 @node Target Format Checks
18277 @section Format Checks Specific to Particular Target Machines
18279 For some target machines, GCC supports additional options to the
18281 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18284 * Solaris Format Checks::
18285 * Darwin Format Checks::
18288 @node Solaris Format Checks
18289 @subsection Solaris Format Checks
18291 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18292 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18293 conversions, and the two-argument @code{%b} conversion for displaying
18294 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18296 @node Darwin Format Checks
18297 @subsection Darwin Format Checks
18299 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18300 attribute context. Declarations made with such attribution are parsed for correct syntax
18301 and format argument types. However, parsing of the format string itself is currently undefined
18302 and is not carried out by this version of the compiler.
18304 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18305 also be used as format arguments. Note that the relevant headers are only likely to be
18306 available on Darwin (OSX) installations. On such installations, the XCode and system
18307 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18308 associated functions.
18311 @section Pragmas Accepted by GCC
18313 @cindex @code{#pragma}
18315 GCC supports several types of pragmas, primarily in order to compile
18316 code originally written for other compilers. Note that in general
18317 we do not recommend the use of pragmas; @xref{Function Attributes},
18318 for further explanation.
18321 * AArch64 Pragmas::
18325 * RS/6000 and PowerPC Pragmas::
18327 * Solaris Pragmas::
18328 * Symbol-Renaming Pragmas::
18329 * Structure-Packing Pragmas::
18331 * Diagnostic Pragmas::
18332 * Visibility Pragmas::
18333 * Push/Pop Macro Pragmas::
18334 * Function Specific Option Pragmas::
18335 * Loop-Specific Pragmas::
18338 @node AArch64 Pragmas
18339 @subsection AArch64 Pragmas
18341 The pragmas defined by the AArch64 target correspond to the AArch64
18342 target function attributes. They can be specified as below:
18344 #pragma GCC target("string")
18347 where @code{@var{string}} can be any string accepted as an AArch64 target
18348 attribute. @xref{AArch64 Function Attributes}, for more details
18349 on the permissible values of @code{string}.
18352 @subsection ARM Pragmas
18354 The ARM target defines pragmas for controlling the default addition of
18355 @code{long_call} and @code{short_call} attributes to functions.
18356 @xref{Function Attributes}, for information about the effects of these
18361 @cindex pragma, long_calls
18362 Set all subsequent functions to have the @code{long_call} attribute.
18364 @item no_long_calls
18365 @cindex pragma, no_long_calls
18366 Set all subsequent functions to have the @code{short_call} attribute.
18368 @item long_calls_off
18369 @cindex pragma, long_calls_off
18370 Do not affect the @code{long_call} or @code{short_call} attributes of
18371 subsequent functions.
18375 @subsection M32C Pragmas
18378 @item GCC memregs @var{number}
18379 @cindex pragma, memregs
18380 Overrides the command-line option @code{-memregs=} for the current
18381 file. Use with care! This pragma must be before any function in the
18382 file, and mixing different memregs values in different objects may
18383 make them incompatible. This pragma is useful when a
18384 performance-critical function uses a memreg for temporary values,
18385 as it may allow you to reduce the number of memregs used.
18387 @item ADDRESS @var{name} @var{address}
18388 @cindex pragma, address
18389 For any declared symbols matching @var{name}, this does three things
18390 to that symbol: it forces the symbol to be located at the given
18391 address (a number), it forces the symbol to be volatile, and it
18392 changes the symbol's scope to be static. This pragma exists for
18393 compatibility with other compilers, but note that the common
18394 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18398 #pragma ADDRESS port3 0x103
18405 @subsection MeP Pragmas
18409 @item custom io_volatile (on|off)
18410 @cindex pragma, custom io_volatile
18411 Overrides the command-line option @code{-mio-volatile} for the current
18412 file. Note that for compatibility with future GCC releases, this
18413 option should only be used once before any @code{io} variables in each
18416 @item GCC coprocessor available @var{registers}
18417 @cindex pragma, coprocessor available
18418 Specifies which coprocessor registers are available to the register
18419 allocator. @var{registers} may be a single register, register range
18420 separated by ellipses, or comma-separated list of those. Example:
18423 #pragma GCC coprocessor available $c0...$c10, $c28
18426 @item GCC coprocessor call_saved @var{registers}
18427 @cindex pragma, coprocessor call_saved
18428 Specifies which coprocessor registers are to be saved and restored by
18429 any function using them. @var{registers} may be a single register,
18430 register range separated by ellipses, or comma-separated list of
18434 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18437 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18438 @cindex pragma, coprocessor subclass
18439 Creates and defines a register class. These register classes can be
18440 used by inline @code{asm} constructs. @var{registers} may be a single
18441 register, register range separated by ellipses, or comma-separated
18442 list of those. Example:
18445 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18447 asm ("cpfoo %0" : "=B" (x));
18450 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18451 @cindex pragma, disinterrupt
18452 For the named functions, the compiler adds code to disable interrupts
18453 for the duration of those functions. If any functions so named
18454 are not encountered in the source, a warning is emitted that the pragma is
18455 not used. Examples:
18458 #pragma disinterrupt foo
18459 #pragma disinterrupt bar, grill
18460 int foo () @{ @dots{} @}
18463 @item GCC call @var{name} , @var{name} @dots{}
18464 @cindex pragma, call
18465 For the named functions, the compiler always uses a register-indirect
18466 call model when calling the named functions. Examples:
18475 @node RS/6000 and PowerPC Pragmas
18476 @subsection RS/6000 and PowerPC Pragmas
18478 The RS/6000 and PowerPC targets define one pragma for controlling
18479 whether or not the @code{longcall} attribute is added to function
18480 declarations by default. This pragma overrides the @option{-mlongcall}
18481 option, but not the @code{longcall} and @code{shortcall} attributes.
18482 @xref{RS/6000 and PowerPC Options}, for more information about when long
18483 calls are and are not necessary.
18487 @cindex pragma, longcall
18488 Apply the @code{longcall} attribute to all subsequent function
18492 Do not apply the @code{longcall} attribute to subsequent function
18496 @c Describe h8300 pragmas here.
18497 @c Describe sh pragmas here.
18498 @c Describe v850 pragmas here.
18500 @node Darwin Pragmas
18501 @subsection Darwin Pragmas
18503 The following pragmas are available for all architectures running the
18504 Darwin operating system. These are useful for compatibility with other
18508 @item mark @var{tokens}@dots{}
18509 @cindex pragma, mark
18510 This pragma is accepted, but has no effect.
18512 @item options align=@var{alignment}
18513 @cindex pragma, options align
18514 This pragma sets the alignment of fields in structures. The values of
18515 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
18516 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
18517 properly; to restore the previous setting, use @code{reset} for the
18520 @item segment @var{tokens}@dots{}
18521 @cindex pragma, segment
18522 This pragma is accepted, but has no effect.
18524 @item unused (@var{var} [, @var{var}]@dots{})
18525 @cindex pragma, unused
18526 This pragma declares variables to be possibly unused. GCC does not
18527 produce warnings for the listed variables. The effect is similar to
18528 that of the @code{unused} attribute, except that this pragma may appear
18529 anywhere within the variables' scopes.
18532 @node Solaris Pragmas
18533 @subsection Solaris Pragmas
18535 The Solaris target supports @code{#pragma redefine_extname}
18536 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
18537 @code{#pragma} directives for compatibility with the system compiler.
18540 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
18541 @cindex pragma, align
18543 Increase the minimum alignment of each @var{variable} to @var{alignment}.
18544 This is the same as GCC's @code{aligned} attribute @pxref{Variable
18545 Attributes}). Macro expansion occurs on the arguments to this pragma
18546 when compiling C and Objective-C@. It does not currently occur when
18547 compiling C++, but this is a bug which may be fixed in a future
18550 @item fini (@var{function} [, @var{function}]...)
18551 @cindex pragma, fini
18553 This pragma causes each listed @var{function} to be called after
18554 main, or during shared module unloading, by adding a call to the
18555 @code{.fini} section.
18557 @item init (@var{function} [, @var{function}]...)
18558 @cindex pragma, init
18560 This pragma causes each listed @var{function} to be called during
18561 initialization (before @code{main}) or during shared module loading, by
18562 adding a call to the @code{.init} section.
18566 @node Symbol-Renaming Pragmas
18567 @subsection Symbol-Renaming Pragmas
18569 GCC supports a @code{#pragma} directive that changes the name used in
18570 assembly for a given declaration. While this pragma is supported on all
18571 platforms, it is intended primarily to provide compatibility with the
18572 Solaris system headers. This effect can also be achieved using the asm
18573 labels extension (@pxref{Asm Labels}).
18576 @item redefine_extname @var{oldname} @var{newname}
18577 @cindex pragma, redefine_extname
18579 This pragma gives the C function @var{oldname} the assembly symbol
18580 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
18581 is defined if this pragma is available (currently on all platforms).
18584 This pragma and the asm labels extension interact in a complicated
18585 manner. Here are some corner cases you may want to be aware of:
18588 @item This pragma silently applies only to declarations with external
18589 linkage. Asm labels do not have this restriction.
18591 @item In C++, this pragma silently applies only to declarations with
18592 ``C'' linkage. Again, asm labels do not have this restriction.
18594 @item If either of the ways of changing the assembly name of a
18595 declaration are applied to a declaration whose assembly name has
18596 already been determined (either by a previous use of one of these
18597 features, or because the compiler needed the assembly name in order to
18598 generate code), and the new name is different, a warning issues and
18599 the name does not change.
18601 @item The @var{oldname} used by @code{#pragma redefine_extname} is
18602 always the C-language name.
18605 @node Structure-Packing Pragmas
18606 @subsection Structure-Packing Pragmas
18608 For compatibility with Microsoft Windows compilers, GCC supports a
18609 set of @code{#pragma} directives that change the maximum alignment of
18610 members of structures (other than zero-width bit-fields), unions, and
18611 classes subsequently defined. The @var{n} value below always is required
18612 to be a small power of two and specifies the new alignment in bytes.
18615 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
18616 @item @code{#pragma pack()} sets the alignment to the one that was in
18617 effect when compilation started (see also command-line option
18618 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
18619 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
18620 setting on an internal stack and then optionally sets the new alignment.
18621 @item @code{#pragma pack(pop)} restores the alignment setting to the one
18622 saved at the top of the internal stack (and removes that stack entry).
18623 Note that @code{#pragma pack([@var{n}])} does not influence this internal
18624 stack; thus it is possible to have @code{#pragma pack(push)} followed by
18625 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
18626 @code{#pragma pack(pop)}.
18629 Some targets, e.g.@: x86 and PowerPC, support the @code{ms_struct}
18630 @code{#pragma} which lays out a structure as the documented
18631 @code{__attribute__ ((ms_struct))}.
18633 @item @code{#pragma ms_struct on} turns on the layout for structures
18635 @item @code{#pragma ms_struct off} turns off the layout for structures
18637 @item @code{#pragma ms_struct reset} goes back to the default layout.
18641 @subsection Weak Pragmas
18643 For compatibility with SVR4, GCC supports a set of @code{#pragma}
18644 directives for declaring symbols to be weak, and defining weak
18648 @item #pragma weak @var{symbol}
18649 @cindex pragma, weak
18650 This pragma declares @var{symbol} to be weak, as if the declaration
18651 had the attribute of the same name. The pragma may appear before
18652 or after the declaration of @var{symbol}. It is not an error for
18653 @var{symbol} to never be defined at all.
18655 @item #pragma weak @var{symbol1} = @var{symbol2}
18656 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
18657 It is an error if @var{symbol2} is not defined in the current
18661 @node Diagnostic Pragmas
18662 @subsection Diagnostic Pragmas
18664 GCC allows the user to selectively enable or disable certain types of
18665 diagnostics, and change the kind of the diagnostic. For example, a
18666 project's policy might require that all sources compile with
18667 @option{-Werror} but certain files might have exceptions allowing
18668 specific types of warnings. Or, a project might selectively enable
18669 diagnostics and treat them as errors depending on which preprocessor
18670 macros are defined.
18673 @item #pragma GCC diagnostic @var{kind} @var{option}
18674 @cindex pragma, diagnostic
18676 Modifies the disposition of a diagnostic. Note that not all
18677 diagnostics are modifiable; at the moment only warnings (normally
18678 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
18679 Use @option{-fdiagnostics-show-option} to determine which diagnostics
18680 are controllable and which option controls them.
18682 @var{kind} is @samp{error} to treat this diagnostic as an error,
18683 @samp{warning} to treat it like a warning (even if @option{-Werror} is
18684 in effect), or @samp{ignored} if the diagnostic is to be ignored.
18685 @var{option} is a double quoted string that matches the command-line
18689 #pragma GCC diagnostic warning "-Wformat"
18690 #pragma GCC diagnostic error "-Wformat"
18691 #pragma GCC diagnostic ignored "-Wformat"
18694 Note that these pragmas override any command-line options. GCC keeps
18695 track of the location of each pragma, and issues diagnostics according
18696 to the state as of that point in the source file. Thus, pragmas occurring
18697 after a line do not affect diagnostics caused by that line.
18699 @item #pragma GCC diagnostic push
18700 @itemx #pragma GCC diagnostic pop
18702 Causes GCC to remember the state of the diagnostics as of each
18703 @code{push}, and restore to that point at each @code{pop}. If a
18704 @code{pop} has no matching @code{push}, the command-line options are
18708 #pragma GCC diagnostic error "-Wuninitialized"
18709 foo(a); /* error is given for this one */
18710 #pragma GCC diagnostic push
18711 #pragma GCC diagnostic ignored "-Wuninitialized"
18712 foo(b); /* no diagnostic for this one */
18713 #pragma GCC diagnostic pop
18714 foo(c); /* error is given for this one */
18715 #pragma GCC diagnostic pop
18716 foo(d); /* depends on command-line options */
18721 GCC also offers a simple mechanism for printing messages during
18725 @item #pragma message @var{string}
18726 @cindex pragma, diagnostic
18728 Prints @var{string} as a compiler message on compilation. The message
18729 is informational only, and is neither a compilation warning nor an error.
18732 #pragma message "Compiling " __FILE__ "..."
18735 @var{string} may be parenthesized, and is printed with location
18736 information. For example,
18739 #define DO_PRAGMA(x) _Pragma (#x)
18740 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
18742 TODO(Remember to fix this)
18746 prints @samp{/tmp/file.c:4: note: #pragma message:
18747 TODO - Remember to fix this}.
18751 @node Visibility Pragmas
18752 @subsection Visibility Pragmas
18755 @item #pragma GCC visibility push(@var{visibility})
18756 @itemx #pragma GCC visibility pop
18757 @cindex pragma, visibility
18759 This pragma allows the user to set the visibility for multiple
18760 declarations without having to give each a visibility attribute
18761 (@pxref{Function Attributes}).
18763 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
18764 declarations. Class members and template specializations are not
18765 affected; if you want to override the visibility for a particular
18766 member or instantiation, you must use an attribute.
18771 @node Push/Pop Macro Pragmas
18772 @subsection Push/Pop Macro Pragmas
18774 For compatibility with Microsoft Windows compilers, GCC supports
18775 @samp{#pragma push_macro(@var{"macro_name"})}
18776 and @samp{#pragma pop_macro(@var{"macro_name"})}.
18779 @item #pragma push_macro(@var{"macro_name"})
18780 @cindex pragma, push_macro
18781 This pragma saves the value of the macro named as @var{macro_name} to
18782 the top of the stack for this macro.
18784 @item #pragma pop_macro(@var{"macro_name"})
18785 @cindex pragma, pop_macro
18786 This pragma sets the value of the macro named as @var{macro_name} to
18787 the value on top of the stack for this macro. If the stack for
18788 @var{macro_name} is empty, the value of the macro remains unchanged.
18795 #pragma push_macro("X")
18798 #pragma pop_macro("X")
18803 In this example, the definition of X as 1 is saved by @code{#pragma
18804 push_macro} and restored by @code{#pragma pop_macro}.
18806 @node Function Specific Option Pragmas
18807 @subsection Function Specific Option Pragmas
18810 @item #pragma GCC target (@var{"string"}...)
18811 @cindex pragma GCC target
18813 This pragma allows you to set target specific options for functions
18814 defined later in the source file. One or more strings can be
18815 specified. Each function that is defined after this point is as
18816 if @code{attribute((target("STRING")))} was specified for that
18817 function. The parenthesis around the options is optional.
18818 @xref{Function Attributes}, for more information about the
18819 @code{target} attribute and the attribute syntax.
18821 The @code{#pragma GCC target} pragma is presently implemented for
18822 x86, PowerPC, and Nios II targets only.
18826 @item #pragma GCC optimize (@var{"string"}...)
18827 @cindex pragma GCC optimize
18829 This pragma allows you to set global optimization options for functions
18830 defined later in the source file. One or more strings can be
18831 specified. Each function that is defined after this point is as
18832 if @code{attribute((optimize("STRING")))} was specified for that
18833 function. The parenthesis around the options is optional.
18834 @xref{Function Attributes}, for more information about the
18835 @code{optimize} attribute and the attribute syntax.
18839 @item #pragma GCC push_options
18840 @itemx #pragma GCC pop_options
18841 @cindex pragma GCC push_options
18842 @cindex pragma GCC pop_options
18844 These pragmas maintain a stack of the current target and optimization
18845 options. It is intended for include files where you temporarily want
18846 to switch to using a different @samp{#pragma GCC target} or
18847 @samp{#pragma GCC optimize} and then to pop back to the previous
18852 @item #pragma GCC reset_options
18853 @cindex pragma GCC reset_options
18855 This pragma clears the current @code{#pragma GCC target} and
18856 @code{#pragma GCC optimize} to use the default switches as specified
18857 on the command line.
18860 @node Loop-Specific Pragmas
18861 @subsection Loop-Specific Pragmas
18864 @item #pragma GCC ivdep
18865 @cindex pragma GCC ivdep
18868 With this pragma, the programmer asserts that there are no loop-carried
18869 dependencies which would prevent consecutive iterations of
18870 the following loop from executing concurrently with SIMD
18871 (single instruction multiple data) instructions.
18873 For example, the compiler can only unconditionally vectorize the following
18874 loop with the pragma:
18877 void foo (int n, int *a, int *b, int *c)
18881 for (i = 0; i < n; ++i)
18882 a[i] = b[i] + c[i];
18887 In this example, using the @code{restrict} qualifier had the same
18888 effect. In the following example, that would not be possible. Assume
18889 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
18890 that it can unconditionally vectorize the following loop:
18893 void ignore_vec_dep (int *a, int k, int c, int m)
18896 for (int i = 0; i < m; i++)
18897 a[i] = a[i + k] * c;
18902 @node Unnamed Fields
18903 @section Unnamed Structure and Union Fields
18904 @cindex @code{struct}
18905 @cindex @code{union}
18907 As permitted by ISO C11 and for compatibility with other compilers,
18908 GCC allows you to define
18909 a structure or union that contains, as fields, structures and unions
18910 without names. For example:
18924 In this example, you are able to access members of the unnamed
18925 union with code like @samp{foo.b}. Note that only unnamed structs and
18926 unions are allowed, you may not have, for example, an unnamed
18929 You must never create such structures that cause ambiguous field definitions.
18930 For example, in this structure:
18942 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
18943 The compiler gives errors for such constructs.
18945 @opindex fms-extensions
18946 Unless @option{-fms-extensions} is used, the unnamed field must be a
18947 structure or union definition without a tag (for example, @samp{struct
18948 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
18949 also be a definition with a tag such as @samp{struct foo @{ int a;
18950 @};}, a reference to a previously defined structure or union such as
18951 @samp{struct foo;}, or a reference to a @code{typedef} name for a
18952 previously defined structure or union type.
18954 @opindex fplan9-extensions
18955 The option @option{-fplan9-extensions} enables
18956 @option{-fms-extensions} as well as two other extensions. First, a
18957 pointer to a structure is automatically converted to a pointer to an
18958 anonymous field for assignments and function calls. For example:
18961 struct s1 @{ int a; @};
18962 struct s2 @{ struct s1; @};
18963 extern void f1 (struct s1 *);
18964 void f2 (struct s2 *p) @{ f1 (p); @}
18968 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
18969 converted into a pointer to the anonymous field.
18971 Second, when the type of an anonymous field is a @code{typedef} for a
18972 @code{struct} or @code{union}, code may refer to the field using the
18973 name of the @code{typedef}.
18976 typedef struct @{ int a; @} s1;
18977 struct s2 @{ s1; @};
18978 s1 f1 (struct s2 *p) @{ return p->s1; @}
18981 These usages are only permitted when they are not ambiguous.
18984 @section Thread-Local Storage
18985 @cindex Thread-Local Storage
18986 @cindex @acronym{TLS}
18987 @cindex @code{__thread}
18989 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
18990 are allocated such that there is one instance of the variable per extant
18991 thread. The runtime model GCC uses to implement this originates
18992 in the IA-64 processor-specific ABI, but has since been migrated
18993 to other processors as well. It requires significant support from
18994 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
18995 system libraries (@file{libc.so} and @file{libpthread.so}), so it
18996 is not available everywhere.
18998 At the user level, the extension is visible with a new storage
18999 class keyword: @code{__thread}. For example:
19003 extern __thread struct state s;
19004 static __thread char *p;
19007 The @code{__thread} specifier may be used alone, with the @code{extern}
19008 or @code{static} specifiers, but with no other storage class specifier.
19009 When used with @code{extern} or @code{static}, @code{__thread} must appear
19010 immediately after the other storage class specifier.
19012 The @code{__thread} specifier may be applied to any global, file-scoped
19013 static, function-scoped static, or static data member of a class. It may
19014 not be applied to block-scoped automatic or non-static data member.
19016 When the address-of operator is applied to a thread-local variable, it is
19017 evaluated at run time and returns the address of the current thread's
19018 instance of that variable. An address so obtained may be used by any
19019 thread. When a thread terminates, any pointers to thread-local variables
19020 in that thread become invalid.
19022 No static initialization may refer to the address of a thread-local variable.
19024 In C++, if an initializer is present for a thread-local variable, it must
19025 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19028 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19029 ELF Handling For Thread-Local Storage} for a detailed explanation of
19030 the four thread-local storage addressing models, and how the runtime
19031 is expected to function.
19034 * C99 Thread-Local Edits::
19035 * C++98 Thread-Local Edits::
19038 @node C99 Thread-Local Edits
19039 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19041 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19042 that document the exact semantics of the language extension.
19046 @cite{5.1.2 Execution environments}
19048 Add new text after paragraph 1
19051 Within either execution environment, a @dfn{thread} is a flow of
19052 control within a program. It is implementation defined whether
19053 or not there may be more than one thread associated with a program.
19054 It is implementation defined how threads beyond the first are
19055 created, the name and type of the function called at thread
19056 startup, and how threads may be terminated. However, objects
19057 with thread storage duration shall be initialized before thread
19062 @cite{6.2.4 Storage durations of objects}
19064 Add new text before paragraph 3
19067 An object whose identifier is declared with the storage-class
19068 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19069 Its lifetime is the entire execution of the thread, and its
19070 stored value is initialized only once, prior to thread startup.
19074 @cite{6.4.1 Keywords}
19076 Add @code{__thread}.
19079 @cite{6.7.1 Storage-class specifiers}
19081 Add @code{__thread} to the list of storage class specifiers in
19084 Change paragraph 2 to
19087 With the exception of @code{__thread}, at most one storage-class
19088 specifier may be given [@dots{}]. The @code{__thread} specifier may
19089 be used alone, or immediately following @code{extern} or
19093 Add new text after paragraph 6
19096 The declaration of an identifier for a variable that has
19097 block scope that specifies @code{__thread} shall also
19098 specify either @code{extern} or @code{static}.
19100 The @code{__thread} specifier shall be used only with
19105 @node C++98 Thread-Local Edits
19106 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19108 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19109 that document the exact semantics of the language extension.
19113 @b{[intro.execution]}
19115 New text after paragraph 4
19118 A @dfn{thread} is a flow of control within the abstract machine.
19119 It is implementation defined whether or not there may be more than
19123 New text after paragraph 7
19126 It is unspecified whether additional action must be taken to
19127 ensure when and whether side effects are visible to other threads.
19133 Add @code{__thread}.
19136 @b{[basic.start.main]}
19138 Add after paragraph 5
19141 The thread that begins execution at the @code{main} function is called
19142 the @dfn{main thread}. It is implementation defined how functions
19143 beginning threads other than the main thread are designated or typed.
19144 A function so designated, as well as the @code{main} function, is called
19145 a @dfn{thread startup function}. It is implementation defined what
19146 happens if a thread startup function returns. It is implementation
19147 defined what happens to other threads when any thread calls @code{exit}.
19151 @b{[basic.start.init]}
19153 Add after paragraph 4
19156 The storage for an object of thread storage duration shall be
19157 statically initialized before the first statement of the thread startup
19158 function. An object of thread storage duration shall not require
19159 dynamic initialization.
19163 @b{[basic.start.term]}
19165 Add after paragraph 3
19168 The type of an object with thread storage duration shall not have a
19169 non-trivial destructor, nor shall it be an array type whose elements
19170 (directly or indirectly) have non-trivial destructors.
19176 Add ``thread storage duration'' to the list in paragraph 1.
19181 Thread, static, and automatic storage durations are associated with
19182 objects introduced by declarations [@dots{}].
19185 Add @code{__thread} to the list of specifiers in paragraph 3.
19188 @b{[basic.stc.thread]}
19190 New section before @b{[basic.stc.static]}
19193 The keyword @code{__thread} applied to a non-local object gives the
19194 object thread storage duration.
19196 A local variable or class data member declared both @code{static}
19197 and @code{__thread} gives the variable or member thread storage
19202 @b{[basic.stc.static]}
19207 All objects that have neither thread storage duration, dynamic
19208 storage duration nor are local [@dots{}].
19214 Add @code{__thread} to the list in paragraph 1.
19219 With the exception of @code{__thread}, at most one
19220 @var{storage-class-specifier} shall appear in a given
19221 @var{decl-specifier-seq}. The @code{__thread} specifier may
19222 be used alone, or immediately following the @code{extern} or
19223 @code{static} specifiers. [@dots{}]
19226 Add after paragraph 5
19229 The @code{__thread} specifier can be applied only to the names of objects
19230 and to anonymous unions.
19236 Add after paragraph 6
19239 Non-@code{static} members shall not be @code{__thread}.
19243 @node Binary constants
19244 @section Binary Constants using the @samp{0b} Prefix
19245 @cindex Binary constants using the @samp{0b} prefix
19247 Integer constants can be written as binary constants, consisting of a
19248 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19249 @samp{0B}. This is particularly useful in environments that operate a
19250 lot on the bit level (like microcontrollers).
19252 The following statements are identical:
19261 The type of these constants follows the same rules as for octal or
19262 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19265 @node C++ Extensions
19266 @chapter Extensions to the C++ Language
19267 @cindex extensions, C++ language
19268 @cindex C++ language extensions
19270 The GNU compiler provides these extensions to the C++ language (and you
19271 can also use most of the C language extensions in your C++ programs). If you
19272 want to write code that checks whether these features are available, you can
19273 test for the GNU compiler the same way as for C programs: check for a
19274 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19275 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19276 Predefined Macros,cpp,The GNU C Preprocessor}).
19279 * C++ Volatiles:: What constitutes an access to a volatile object.
19280 * Restricted Pointers:: C99 restricted pointers and references.
19281 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19282 * C++ Interface:: You can use a single C++ header file for both
19283 declarations and definitions.
19284 * Template Instantiation:: Methods for ensuring that exactly one copy of
19285 each needed template instantiation is emitted.
19286 * Bound member functions:: You can extract a function pointer to the
19287 method denoted by a @samp{->*} or @samp{.*} expression.
19288 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19289 * Function Multiversioning:: Declaring multiple function versions.
19290 * Namespace Association:: Strong using-directives for namespace association.
19291 * Type Traits:: Compiler support for type traits.
19292 * C++ Concepts:: Improved support for generic programming.
19293 * Java Exceptions:: Tweaking exception handling to work with Java.
19294 * Deprecated Features:: Things will disappear from G++.
19295 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19298 @node C++ Volatiles
19299 @section When is a Volatile C++ Object Accessed?
19300 @cindex accessing volatiles
19301 @cindex volatile read
19302 @cindex volatile write
19303 @cindex volatile access
19305 The C++ standard differs from the C standard in its treatment of
19306 volatile objects. It fails to specify what constitutes a volatile
19307 access, except to say that C++ should behave in a similar manner to C
19308 with respect to volatiles, where possible. However, the different
19309 lvalueness of expressions between C and C++ complicate the behavior.
19310 G++ behaves the same as GCC for volatile access, @xref{C
19311 Extensions,,Volatiles}, for a description of GCC's behavior.
19313 The C and C++ language specifications differ when an object is
19314 accessed in a void context:
19317 volatile int *src = @var{somevalue};
19321 The C++ standard specifies that such expressions do not undergo lvalue
19322 to rvalue conversion, and that the type of the dereferenced object may
19323 be incomplete. The C++ standard does not specify explicitly that it
19324 is lvalue to rvalue conversion that is responsible for causing an
19325 access. There is reason to believe that it is, because otherwise
19326 certain simple expressions become undefined. However, because it
19327 would surprise most programmers, G++ treats dereferencing a pointer to
19328 volatile object of complete type as GCC would do for an equivalent
19329 type in C@. When the object has incomplete type, G++ issues a
19330 warning; if you wish to force an error, you must force a conversion to
19331 rvalue with, for instance, a static cast.
19333 When using a reference to volatile, G++ does not treat equivalent
19334 expressions as accesses to volatiles, but instead issues a warning that
19335 no volatile is accessed. The rationale for this is that otherwise it
19336 becomes difficult to determine where volatile access occur, and not
19337 possible to ignore the return value from functions returning volatile
19338 references. Again, if you wish to force a read, cast the reference to
19341 G++ implements the same behavior as GCC does when assigning to a
19342 volatile object---there is no reread of the assigned-to object, the
19343 assigned rvalue is reused. Note that in C++ assignment expressions
19344 are lvalues, and if used as an lvalue, the volatile object is
19345 referred to. For instance, @var{vref} refers to @var{vobj}, as
19346 expected, in the following example:
19350 volatile int &vref = vobj = @var{something};
19353 @node Restricted Pointers
19354 @section Restricting Pointer Aliasing
19355 @cindex restricted pointers
19356 @cindex restricted references
19357 @cindex restricted this pointer
19359 As with the C front end, G++ understands the C99 feature of restricted pointers,
19360 specified with the @code{__restrict__}, or @code{__restrict} type
19361 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19362 language flag, @code{restrict} is not a keyword in C++.
19364 In addition to allowing restricted pointers, you can specify restricted
19365 references, which indicate that the reference is not aliased in the local
19369 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19376 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19377 @var{rref} refers to a (different) unaliased integer.
19379 You may also specify whether a member function's @var{this} pointer is
19380 unaliased by using @code{__restrict__} as a member function qualifier.
19383 void T::fn () __restrict__
19390 Within the body of @code{T::fn}, @var{this} has the effective
19391 definition @code{T *__restrict__ const this}. Notice that the
19392 interpretation of a @code{__restrict__} member function qualifier is
19393 different to that of @code{const} or @code{volatile} qualifier, in that it
19394 is applied to the pointer rather than the object. This is consistent with
19395 other compilers that implement restricted pointers.
19397 As with all outermost parameter qualifiers, @code{__restrict__} is
19398 ignored in function definition matching. This means you only need to
19399 specify @code{__restrict__} in a function definition, rather than
19400 in a function prototype as well.
19402 @node Vague Linkage
19403 @section Vague Linkage
19404 @cindex vague linkage
19406 There are several constructs in C++ that require space in the object
19407 file but are not clearly tied to a single translation unit. We say that
19408 these constructs have ``vague linkage''. Typically such constructs are
19409 emitted wherever they are needed, though sometimes we can be more
19413 @item Inline Functions
19414 Inline functions are typically defined in a header file which can be
19415 included in many different compilations. Hopefully they can usually be
19416 inlined, but sometimes an out-of-line copy is necessary, if the address
19417 of the function is taken or if inlining fails. In general, we emit an
19418 out-of-line copy in all translation units where one is needed. As an
19419 exception, we only emit inline virtual functions with the vtable, since
19420 it always requires a copy.
19422 Local static variables and string constants used in an inline function
19423 are also considered to have vague linkage, since they must be shared
19424 between all inlined and out-of-line instances of the function.
19428 C++ virtual functions are implemented in most compilers using a lookup
19429 table, known as a vtable. The vtable contains pointers to the virtual
19430 functions provided by a class, and each object of the class contains a
19431 pointer to its vtable (or vtables, in some multiple-inheritance
19432 situations). If the class declares any non-inline, non-pure virtual
19433 functions, the first one is chosen as the ``key method'' for the class,
19434 and the vtable is only emitted in the translation unit where the key
19437 @emph{Note:} If the chosen key method is later defined as inline, the
19438 vtable is still emitted in every translation unit that defines it.
19439 Make sure that any inline virtuals are declared inline in the class
19440 body, even if they are not defined there.
19442 @item @code{type_info} objects
19443 @cindex @code{type_info}
19445 C++ requires information about types to be written out in order to
19446 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19447 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19448 object is written out along with the vtable so that @samp{dynamic_cast}
19449 can determine the dynamic type of a class object at run time. For all
19450 other types, we write out the @samp{type_info} object when it is used: when
19451 applying @samp{typeid} to an expression, throwing an object, or
19452 referring to a type in a catch clause or exception specification.
19454 @item Template Instantiations
19455 Most everything in this section also applies to template instantiations,
19456 but there are other options as well.
19457 @xref{Template Instantiation,,Where's the Template?}.
19461 When used with GNU ld version 2.8 or later on an ELF system such as
19462 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
19463 these constructs will be discarded at link time. This is known as
19466 On targets that don't support COMDAT, but do support weak symbols, GCC
19467 uses them. This way one copy overrides all the others, but
19468 the unused copies still take up space in the executable.
19470 For targets that do not support either COMDAT or weak symbols,
19471 most entities with vague linkage are emitted as local symbols to
19472 avoid duplicate definition errors from the linker. This does not happen
19473 for local statics in inlines, however, as having multiple copies
19474 almost certainly breaks things.
19476 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
19477 another way to control placement of these constructs.
19479 @node C++ Interface
19480 @section C++ Interface and Implementation Pragmas
19482 @cindex interface and implementation headers, C++
19483 @cindex C++ interface and implementation headers
19484 @cindex pragmas, interface and implementation
19486 @code{#pragma interface} and @code{#pragma implementation} provide the
19487 user with a way of explicitly directing the compiler to emit entities
19488 with vague linkage (and debugging information) in a particular
19491 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
19492 by COMDAT support and the ``key method'' heuristic
19493 mentioned in @ref{Vague Linkage}. Using them can actually cause your
19494 program to grow due to unnecessary out-of-line copies of inline
19498 @item #pragma interface
19499 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
19500 @kindex #pragma interface
19501 Use this directive in @emph{header files} that define object classes, to save
19502 space in most of the object files that use those classes. Normally,
19503 local copies of certain information (backup copies of inline member
19504 functions, debugging information, and the internal tables that implement
19505 virtual functions) must be kept in each object file that includes class
19506 definitions. You can use this pragma to avoid such duplication. When a
19507 header file containing @samp{#pragma interface} is included in a
19508 compilation, this auxiliary information is not generated (unless
19509 the main input source file itself uses @samp{#pragma implementation}).
19510 Instead, the object files contain references to be resolved at link
19513 The second form of this directive is useful for the case where you have
19514 multiple headers with the same name in different directories. If you
19515 use this form, you must specify the same string to @samp{#pragma
19518 @item #pragma implementation
19519 @itemx #pragma implementation "@var{objects}.h"
19520 @kindex #pragma implementation
19521 Use this pragma in a @emph{main input file}, when you want full output from
19522 included header files to be generated (and made globally visible). The
19523 included header file, in turn, should use @samp{#pragma interface}.
19524 Backup copies of inline member functions, debugging information, and the
19525 internal tables used to implement virtual functions are all generated in
19526 implementation files.
19528 @cindex implied @code{#pragma implementation}
19529 @cindex @code{#pragma implementation}, implied
19530 @cindex naming convention, implementation headers
19531 If you use @samp{#pragma implementation} with no argument, it applies to
19532 an include file with the same basename@footnote{A file's @dfn{basename}
19533 is the name stripped of all leading path information and of trailing
19534 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
19535 file. For example, in @file{allclass.cc}, giving just
19536 @samp{#pragma implementation}
19537 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
19539 Use the string argument if you want a single implementation file to
19540 include code from multiple header files. (You must also use
19541 @samp{#include} to include the header file; @samp{#pragma
19542 implementation} only specifies how to use the file---it doesn't actually
19545 There is no way to split up the contents of a single header file into
19546 multiple implementation files.
19549 @cindex inlining and C++ pragmas
19550 @cindex C++ pragmas, effect on inlining
19551 @cindex pragmas in C++, effect on inlining
19552 @samp{#pragma implementation} and @samp{#pragma interface} also have an
19553 effect on function inlining.
19555 If you define a class in a header file marked with @samp{#pragma
19556 interface}, the effect on an inline function defined in that class is
19557 similar to an explicit @code{extern} declaration---the compiler emits
19558 no code at all to define an independent version of the function. Its
19559 definition is used only for inlining with its callers.
19561 @opindex fno-implement-inlines
19562 Conversely, when you include the same header file in a main source file
19563 that declares it as @samp{#pragma implementation}, the compiler emits
19564 code for the function itself; this defines a version of the function
19565 that can be found via pointers (or by callers compiled without
19566 inlining). If all calls to the function can be inlined, you can avoid
19567 emitting the function by compiling with @option{-fno-implement-inlines}.
19568 If any calls are not inlined, you will get linker errors.
19570 @node Template Instantiation
19571 @section Where's the Template?
19572 @cindex template instantiation
19574 C++ templates are the first language feature to require more
19575 intelligence from the environment than one usually finds on a UNIX
19576 system. Somehow the compiler and linker have to make sure that each
19577 template instance occurs exactly once in the executable if it is needed,
19578 and not at all otherwise. There are two basic approaches to this
19579 problem, which are referred to as the Borland model and the Cfront model.
19582 @item Borland model
19583 Borland C++ solved the template instantiation problem by adding the code
19584 equivalent of common blocks to their linker; the compiler emits template
19585 instances in each translation unit that uses them, and the linker
19586 collapses them together. The advantage of this model is that the linker
19587 only has to consider the object files themselves; there is no external
19588 complexity to worry about. This disadvantage is that compilation time
19589 is increased because the template code is being compiled repeatedly.
19590 Code written for this model tends to include definitions of all
19591 templates in the header file, since they must be seen to be
19595 The AT&T C++ translator, Cfront, solved the template instantiation
19596 problem by creating the notion of a template repository, an
19597 automatically maintained place where template instances are stored. A
19598 more modern version of the repository works as follows: As individual
19599 object files are built, the compiler places any template definitions and
19600 instantiations encountered in the repository. At link time, the link
19601 wrapper adds in the objects in the repository and compiles any needed
19602 instances that were not previously emitted. The advantages of this
19603 model are more optimal compilation speed and the ability to use the
19604 system linker; to implement the Borland model a compiler vendor also
19605 needs to replace the linker. The disadvantages are vastly increased
19606 complexity, and thus potential for error; for some code this can be
19607 just as transparent, but in practice it can been very difficult to build
19608 multiple programs in one directory and one program in multiple
19609 directories. Code written for this model tends to separate definitions
19610 of non-inline member templates into a separate file, which should be
19611 compiled separately.
19614 When used with GNU ld version 2.8 or later on an ELF system such as
19615 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
19616 Borland model. On other systems, G++ implements neither automatic
19619 You have the following options for dealing with template instantiations:
19624 Compile your template-using code with @option{-frepo}. The compiler
19625 generates files with the extension @samp{.rpo} listing all of the
19626 template instantiations used in the corresponding object files that
19627 could be instantiated there; the link wrapper, @samp{collect2},
19628 then updates the @samp{.rpo} files to tell the compiler where to place
19629 those instantiations and rebuild any affected object files. The
19630 link-time overhead is negligible after the first pass, as the compiler
19631 continues to place the instantiations in the same files.
19633 This is your best option for application code written for the Borland
19634 model, as it just works. Code written for the Cfront model
19635 needs to be modified so that the template definitions are available at
19636 one or more points of instantiation; usually this is as simple as adding
19637 @code{#include <tmethods.cc>} to the end of each template header.
19639 For library code, if you want the library to provide all of the template
19640 instantiations it needs, just try to link all of its object files
19641 together; the link will fail, but cause the instantiations to be
19642 generated as a side effect. Be warned, however, that this may cause
19643 conflicts if multiple libraries try to provide the same instantiations.
19644 For greater control, use explicit instantiation as described in the next
19648 @opindex fno-implicit-templates
19649 Compile your code with @option{-fno-implicit-templates} to disable the
19650 implicit generation of template instances, and explicitly instantiate
19651 all the ones you use. This approach requires more knowledge of exactly
19652 which instances you need than do the others, but it's less
19653 mysterious and allows greater control. You can scatter the explicit
19654 instantiations throughout your program, perhaps putting them in the
19655 translation units where the instances are used or the translation units
19656 that define the templates themselves; you can put all of the explicit
19657 instantiations you need into one big file; or you can create small files
19664 template class Foo<int>;
19665 template ostream& operator <<
19666 (ostream&, const Foo<int>&);
19670 for each of the instances you need, and create a template instantiation
19671 library from those.
19673 If you are using Cfront-model code, you can probably get away with not
19674 using @option{-fno-implicit-templates} when compiling files that don't
19675 @samp{#include} the member template definitions.
19677 If you use one big file to do the instantiations, you may want to
19678 compile it without @option{-fno-implicit-templates} so you get all of the
19679 instances required by your explicit instantiations (but not by any
19680 other files) without having to specify them as well.
19682 The ISO C++ 2011 standard allows forward declaration of explicit
19683 instantiations (with @code{extern}). G++ supports explicit instantiation
19684 declarations in C++98 mode and has extended the template instantiation
19685 syntax to support instantiation of the compiler support data for a
19686 template class (i.e.@: the vtable) without instantiating any of its
19687 members (with @code{inline}), and instantiation of only the static data
19688 members of a template class, without the support data or member
19689 functions (with @code{static}):
19692 extern template int max (int, int);
19693 inline template class Foo<int>;
19694 static template class Foo<int>;
19698 Do nothing. Pretend G++ does implement automatic instantiation
19699 management. Code written for the Borland model works fine, but
19700 each translation unit contains instances of each of the templates it
19701 uses. In a large program, this can lead to an unacceptable amount of code
19705 @node Bound member functions
19706 @section Extracting the Function Pointer from a Bound Pointer to Member Function
19708 @cindex pointer to member function
19709 @cindex bound pointer to member function
19711 In C++, pointer to member functions (PMFs) are implemented using a wide
19712 pointer of sorts to handle all the possible call mechanisms; the PMF
19713 needs to store information about how to adjust the @samp{this} pointer,
19714 and if the function pointed to is virtual, where to find the vtable, and
19715 where in the vtable to look for the member function. If you are using
19716 PMFs in an inner loop, you should really reconsider that decision. If
19717 that is not an option, you can extract the pointer to the function that
19718 would be called for a given object/PMF pair and call it directly inside
19719 the inner loop, to save a bit of time.
19721 Note that you still pay the penalty for the call through a
19722 function pointer; on most modern architectures, such a call defeats the
19723 branch prediction features of the CPU@. This is also true of normal
19724 virtual function calls.
19726 The syntax for this extension is
19730 extern int (A::*fp)();
19731 typedef int (*fptr)(A *);
19733 fptr p = (fptr)(a.*fp);
19736 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
19737 no object is needed to obtain the address of the function. They can be
19738 converted to function pointers directly:
19741 fptr p1 = (fptr)(&A::foo);
19744 @opindex Wno-pmf-conversions
19745 You must specify @option{-Wno-pmf-conversions} to use this extension.
19747 @node C++ Attributes
19748 @section C++-Specific Variable, Function, and Type Attributes
19750 Some attributes only make sense for C++ programs.
19753 @item abi_tag ("@var{tag}", ...)
19754 @cindex @code{abi_tag} function attribute
19755 @cindex @code{abi_tag} variable attribute
19756 @cindex @code{abi_tag} type attribute
19757 The @code{abi_tag} attribute can be applied to a function, variable, or class
19758 declaration. It modifies the mangled name of the entity to
19759 incorporate the tag name, in order to distinguish the function or
19760 class from an earlier version with a different ABI; perhaps the class
19761 has changed size, or the function has a different return type that is
19762 not encoded in the mangled name.
19764 The attribute can also be applied to an inline namespace, but does not
19765 affect the mangled name of the namespace; in this case it is only used
19766 for @option{-Wabi-tag} warnings and automatic tagging of functions and
19767 variables. Tagging inline namespaces is generally preferable to
19768 tagging individual declarations, but the latter is sometimes
19769 necessary, such as when only certain members of a class need to be
19772 The argument can be a list of strings of arbitrary length. The
19773 strings are sorted on output, so the order of the list is
19776 A redeclaration of an entity must not add new ABI tags,
19777 since doing so would change the mangled name.
19779 The ABI tags apply to a name, so all instantiations and
19780 specializations of a template have the same tags. The attribute will
19781 be ignored if applied to an explicit specialization or instantiation.
19783 The @option{-Wabi-tag} flag enables a warning about a class which does
19784 not have all the ABI tags used by its subobjects and virtual functions; for users with code
19785 that needs to coexist with an earlier ABI, using this option can help
19786 to find all affected types that need to be tagged.
19788 When a type involving an ABI tag is used as the type of a variable or
19789 return type of a function where that tag is not already present in the
19790 signature of the function, the tag is automatically applied to the
19791 variable or function. @option{-Wabi-tag} also warns about this
19792 situation; this warning can be avoided by explicitly tagging the
19793 variable or function or moving it into a tagged inline namespace.
19795 @item init_priority (@var{priority})
19796 @cindex @code{init_priority} variable attribute
19798 In Standard C++, objects defined at namespace scope are guaranteed to be
19799 initialized in an order in strict accordance with that of their definitions
19800 @emph{in a given translation unit}. No guarantee is made for initializations
19801 across translation units. However, GNU C++ allows users to control the
19802 order of initialization of objects defined at namespace scope with the
19803 @code{init_priority} attribute by specifying a relative @var{priority},
19804 a constant integral expression currently bounded between 101 and 65535
19805 inclusive. Lower numbers indicate a higher priority.
19807 In the following example, @code{A} would normally be created before
19808 @code{B}, but the @code{init_priority} attribute reverses that order:
19811 Some_Class A __attribute__ ((init_priority (2000)));
19812 Some_Class B __attribute__ ((init_priority (543)));
19816 Note that the particular values of @var{priority} do not matter; only their
19819 @item java_interface
19820 @cindex @code{java_interface} type attribute
19822 This type attribute informs C++ that the class is a Java interface. It may
19823 only be applied to classes declared within an @code{extern "Java"} block.
19824 Calls to methods declared in this interface are dispatched using GCJ's
19825 interface table mechanism, instead of regular virtual table dispatch.
19828 @cindex @code{warn_unused} type attribute
19830 For C++ types with non-trivial constructors and/or destructors it is
19831 impossible for the compiler to determine whether a variable of this
19832 type is truly unused if it is not referenced. This type attribute
19833 informs the compiler that variables of this type should be warned
19834 about if they appear to be unused, just like variables of fundamental
19837 This attribute is appropriate for types which just represent a value,
19838 such as @code{std::string}; it is not appropriate for types which
19839 control a resource, such as @code{std::mutex}.
19841 This attribute is also accepted in C, but it is unnecessary because C
19842 does not have constructors or destructors.
19846 See also @ref{Namespace Association}.
19848 @node Function Multiversioning
19849 @section Function Multiversioning
19850 @cindex function versions
19852 With the GNU C++ front end, for x86 targets, you may specify multiple
19853 versions of a function, where each function is specialized for a
19854 specific target feature. At runtime, the appropriate version of the
19855 function is automatically executed depending on the characteristics of
19856 the execution platform. Here is an example.
19859 __attribute__ ((target ("default")))
19862 // The default version of foo.
19866 __attribute__ ((target ("sse4.2")))
19869 // foo version for SSE4.2
19873 __attribute__ ((target ("arch=atom")))
19876 // foo version for the Intel ATOM processor
19880 __attribute__ ((target ("arch=amdfam10")))
19883 // foo version for the AMD Family 0x10 processors.
19890 assert ((*p) () == foo ());
19895 In the above example, four versions of function foo are created. The
19896 first version of foo with the target attribute "default" is the default
19897 version. This version gets executed when no other target specific
19898 version qualifies for execution on a particular platform. A new version
19899 of foo is created by using the same function signature but with a
19900 different target string. Function foo is called or a pointer to it is
19901 taken just like a regular function. GCC takes care of doing the
19902 dispatching to call the right version at runtime. Refer to the
19903 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
19904 Function Multiversioning} for more details.
19906 @node Namespace Association
19907 @section Namespace Association
19909 @strong{Caution:} The semantics of this extension are equivalent
19910 to C++ 2011 inline namespaces. Users should use inline namespaces
19911 instead as this extension will be removed in future versions of G++.
19913 A using-directive with @code{__attribute ((strong))} is stronger
19914 than a normal using-directive in two ways:
19918 Templates from the used namespace can be specialized and explicitly
19919 instantiated as though they were members of the using namespace.
19922 The using namespace is considered an associated namespace of all
19923 templates in the used namespace for purposes of argument-dependent
19927 The used namespace must be nested within the using namespace so that
19928 normal unqualified lookup works properly.
19930 This is useful for composing a namespace transparently from
19931 implementation namespaces. For example:
19936 template <class T> struct A @{ @};
19938 using namespace debug __attribute ((__strong__));
19939 template <> struct A<int> @{ @}; // @r{OK to specialize}
19941 template <class T> void f (A<T>);
19946 f (std::A<float>()); // @r{lookup finds} std::f
19952 @section Type Traits
19954 The C++ front end implements syntactic extensions that allow
19955 compile-time determination of
19956 various characteristics of a type (or of a
19960 @item __has_nothrow_assign (type)
19961 If @code{type} is const qualified or is a reference type then the trait is
19962 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
19963 is true, else if @code{type} is a cv class or union type with copy assignment
19964 operators that are known not to throw an exception then the trait is true,
19965 else it is false. Requires: @code{type} shall be a complete type,
19966 (possibly cv-qualified) @code{void}, or an array of unknown bound.
19968 @item __has_nothrow_copy (type)
19969 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
19970 @code{type} is a cv class or union type with copy constructors that
19971 are known not to throw an exception then the trait is true, else it is false.
19972 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
19973 @code{void}, or an array of unknown bound.
19975 @item __has_nothrow_constructor (type)
19976 If @code{__has_trivial_constructor (type)} is true then the trait is
19977 true, else if @code{type} is a cv class or union type (or array
19978 thereof) with a default constructor that is known not to throw an
19979 exception then the trait is true, else it is false. Requires:
19980 @code{type} shall be a complete type, (possibly cv-qualified)
19981 @code{void}, or an array of unknown bound.
19983 @item __has_trivial_assign (type)
19984 If @code{type} is const qualified or is a reference type then the trait is
19985 false. Otherwise if @code{__is_pod (type)} is true then the trait is
19986 true, else if @code{type} is a cv class or union type with a trivial
19987 copy assignment ([class.copy]) then the trait is true, else it is
19988 false. Requires: @code{type} shall be a complete type, (possibly
19989 cv-qualified) @code{void}, or an array of unknown bound.
19991 @item __has_trivial_copy (type)
19992 If @code{__is_pod (type)} is true or @code{type} is a reference type
19993 then the trait is true, else if @code{type} is a cv class or union type
19994 with a trivial copy constructor ([class.copy]) then the trait
19995 is true, else it is false. Requires: @code{type} shall be a complete
19996 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
19998 @item __has_trivial_constructor (type)
19999 If @code{__is_pod (type)} is true then the trait is true, else if
20000 @code{type} is a cv class or union type (or array thereof) with a
20001 trivial default constructor ([class.ctor]) then the trait is true,
20002 else it is false. Requires: @code{type} shall be a complete
20003 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20005 @item __has_trivial_destructor (type)
20006 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20007 the trait is true, else if @code{type} is a cv class or union type (or
20008 array thereof) with a trivial destructor ([class.dtor]) then the trait
20009 is true, else it is false. Requires: @code{type} shall be a complete
20010 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20012 @item __has_virtual_destructor (type)
20013 If @code{type} is a class type with a virtual destructor
20014 ([class.dtor]) then the trait is true, else it is false. Requires:
20015 @code{type} shall be a complete type, (possibly cv-qualified)
20016 @code{void}, or an array of unknown bound.
20018 @item __is_abstract (type)
20019 If @code{type} is an abstract class ([class.abstract]) then the trait
20020 is true, else it is false. Requires: @code{type} shall be a complete
20021 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20023 @item __is_base_of (base_type, derived_type)
20024 If @code{base_type} is a base class of @code{derived_type}
20025 ([class.derived]) then the trait is true, otherwise it is false.
20026 Top-level cv qualifications of @code{base_type} and
20027 @code{derived_type} are ignored. For the purposes of this trait, a
20028 class type is considered is own base. Requires: if @code{__is_class
20029 (base_type)} and @code{__is_class (derived_type)} are true and
20030 @code{base_type} and @code{derived_type} are not the same type
20031 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20032 type. Diagnostic is produced if this requirement is not met.
20034 @item __is_class (type)
20035 If @code{type} is a cv class type, and not a union type
20036 ([basic.compound]) the trait is true, else it is false.
20038 @item __is_empty (type)
20039 If @code{__is_class (type)} is false then the trait is false.
20040 Otherwise @code{type} is considered empty if and only if: @code{type}
20041 has no non-static data members, or all non-static data members, if
20042 any, are bit-fields of length 0, and @code{type} has no virtual
20043 members, and @code{type} has no virtual base classes, and @code{type}
20044 has no base classes @code{base_type} for which
20045 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20046 be a complete type, (possibly cv-qualified) @code{void}, or an array
20049 @item __is_enum (type)
20050 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20051 true, else it is false.
20053 @item __is_literal_type (type)
20054 If @code{type} is a literal type ([basic.types]) the trait is
20055 true, else it is false. Requires: @code{type} shall be a complete type,
20056 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20058 @item __is_pod (type)
20059 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20060 else it is false. Requires: @code{type} shall be a complete type,
20061 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20063 @item __is_polymorphic (type)
20064 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20065 is true, else it is false. Requires: @code{type} shall be a complete
20066 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20068 @item __is_standard_layout (type)
20069 If @code{type} is a standard-layout type ([basic.types]) the trait is
20070 true, else it is false. Requires: @code{type} shall be a complete
20071 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20073 @item __is_trivial (type)
20074 If @code{type} is a trivial type ([basic.types]) the trait is
20075 true, else it is false. Requires: @code{type} shall be a complete
20076 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20078 @item __is_union (type)
20079 If @code{type} is a cv union type ([basic.compound]) the trait is
20080 true, else it is false.
20082 @item __underlying_type (type)
20083 The underlying type of @code{type}. Requires: @code{type} shall be
20084 an enumeration type ([dcl.enum]).
20090 @section C++ Concepts
20092 C++ concepts provide much-improved support for generic programming. In
20093 particular, they allow the specification of constraints on template arguments.
20094 The constraints are used to extend the usual overloading and partial
20095 specialization capabilities of the language, allowing generic data structures
20096 and algorithms to be ``refined'' based on their properties rather than their
20099 The following keywords are reserved for concepts.
20103 States an expression as an assumption, and if possible, verifies that the
20104 assumption is valid. For example, @code{assume(n > 0)}.
20107 Introduces an axiom definition. Axioms introduce requirements on values.
20110 Introduces a universally quantified object in an axiom. For example,
20111 @code{forall (int n) n + 0 == n}).
20114 Introduces a concept definition. Concepts are sets of syntactic and semantic
20115 requirements on types and their values.
20118 Introduces constraints on template arguments or requirements for a member
20119 function of a class template.
20123 The front end also exposes a number of internal mechanism that can be used
20124 to simplify the writing of type traits. Note that some of these traits are
20125 likely to be removed in the future.
20128 @item __is_same (type1, type2)
20129 A binary type trait: true whenever the type arguments are the same.
20134 @node Java Exceptions
20135 @section Java Exceptions
20137 The Java language uses a slightly different exception handling model
20138 from C++. Normally, GNU C++ automatically detects when you are
20139 writing C++ code that uses Java exceptions, and handle them
20140 appropriately. However, if C++ code only needs to execute destructors
20141 when Java exceptions are thrown through it, GCC guesses incorrectly.
20142 Sample problematic code is:
20145 struct S @{ ~S(); @};
20146 extern void bar(); // @r{is written in Java, and may throw exceptions}
20155 The usual effect of an incorrect guess is a link failure, complaining of
20156 a missing routine called @samp{__gxx_personality_v0}.
20158 You can inform the compiler that Java exceptions are to be used in a
20159 translation unit, irrespective of what it might think, by writing
20160 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20161 @samp{#pragma} must appear before any functions that throw or catch
20162 exceptions, or run destructors when exceptions are thrown through them.
20164 You cannot mix Java and C++ exceptions in the same translation unit. It
20165 is believed to be safe to throw a C++ exception from one file through
20166 another file compiled for the Java exception model, or vice versa, but
20167 there may be bugs in this area.
20169 @node Deprecated Features
20170 @section Deprecated Features
20172 In the past, the GNU C++ compiler was extended to experiment with new
20173 features, at a time when the C++ language was still evolving. Now that
20174 the C++ standard is complete, some of those features are superseded by
20175 superior alternatives. Using the old features might cause a warning in
20176 some cases that the feature will be dropped in the future. In other
20177 cases, the feature might be gone already.
20179 While the list below is not exhaustive, it documents some of the options
20180 that are now deprecated:
20183 @item -fexternal-templates
20184 @itemx -falt-external-templates
20185 These are two of the many ways for G++ to implement template
20186 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20187 defines how template definitions have to be organized across
20188 implementation units. G++ has an implicit instantiation mechanism that
20189 should work just fine for standard-conforming code.
20191 @item -fstrict-prototype
20192 @itemx -fno-strict-prototype
20193 Previously it was possible to use an empty prototype parameter list to
20194 indicate an unspecified number of parameters (like C), rather than no
20195 parameters, as C++ demands. This feature has been removed, except where
20196 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20199 G++ allows a virtual function returning @samp{void *} to be overridden
20200 by one returning a different pointer type. This extension to the
20201 covariant return type rules is now deprecated and will be removed from a
20204 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20205 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20206 and are now removed from G++. Code using these operators should be
20207 modified to use @code{std::min} and @code{std::max} instead.
20209 The named return value extension has been deprecated, and is now
20212 The use of initializer lists with new expressions has been deprecated,
20213 and is now removed from G++.
20215 Floating and complex non-type template parameters have been deprecated,
20216 and are now removed from G++.
20218 The implicit typename extension has been deprecated and is now
20221 The use of default arguments in function pointers, function typedefs
20222 and other places where they are not permitted by the standard is
20223 deprecated and will be removed from a future version of G++.
20225 G++ allows floating-point literals to appear in integral constant expressions,
20226 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20227 This extension is deprecated and will be removed from a future version.
20229 G++ allows static data members of const floating-point type to be declared
20230 with an initializer in a class definition. The standard only allows
20231 initializers for static members of const integral types and const
20232 enumeration types so this extension has been deprecated and will be removed
20233 from a future version.
20235 @node Backwards Compatibility
20236 @section Backwards Compatibility
20237 @cindex Backwards Compatibility
20238 @cindex ARM [Annotated C++ Reference Manual]
20240 Now that there is a definitive ISO standard C++, G++ has a specification
20241 to adhere to. The C++ language evolved over time, and features that
20242 used to be acceptable in previous drafts of the standard, such as the ARM
20243 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20244 compilation of C++ written to such drafts, G++ contains some backwards
20245 compatibilities. @emph{All such backwards compatibility features are
20246 liable to disappear in future versions of G++.} They should be considered
20247 deprecated. @xref{Deprecated Features}.
20251 If a variable is declared at for scope, it used to remain in scope until
20252 the end of the scope that contained the for statement (rather than just
20253 within the for scope). G++ retains this, but issues a warning, if such a
20254 variable is accessed outside the for scope.
20256 @item Implicit C language
20257 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20258 scope to set the language. On such systems, all header files are
20259 implicitly scoped inside a C language scope. Also, an empty prototype
20260 @code{()} is treated as an unspecified number of arguments, rather
20261 than no arguments, as C++ demands.
20264 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20265 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr