1 @c Copyright (C) 1988-2015 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Extensions to the C Language Family
8 @cindex extensions, C language
9 @cindex C language extensions
12 GNU C provides several language features not found in ISO standard C@.
13 (The @option{-pedantic} option directs GCC to print a warning message if
14 any of these features is used.) To test for the availability of these
15 features in conditional compilation, check for a predefined macro
16 @code{__GNUC__}, which is always defined under GCC@.
18 These extensions are available in C and Objective-C@. Most of them are
19 also available in C++. @xref{C++ Extensions,,Extensions to the
20 C++ Language}, for extensions that apply @emph{only} to C++.
22 Some features that are in ISO C99 but not C90 or C++ are also, as
23 extensions, accepted by GCC in C90 mode and in C++.
26 * Statement Exprs:: Putting statements and declarations inside expressions.
27 * Local Labels:: Labels local to a block.
28 * Labels as Values:: Getting pointers to labels, and computed gotos.
29 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
30 * Constructing Calls:: Dispatching a call to another function.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
33 * __int128:: 128-bit integers---@code{__int128}.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Floating Types:: Additional Floating Types.
37 * Half-Precision:: Half-Precision Floating Point.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Named Address Spaces::Named address spaces.
42 * Zero Length:: Zero-length arrays.
43 * Empty Structures:: Structures with no members.
44 * Variable Length:: Arrays whose length is computed at run time.
45 * Variadic Macros:: Macros with a variable number of arguments.
46 * Escaped Newlines:: Slightly looser rules for escaped newlines.
47 * Subscripting:: Any array can be subscripted, even if not an lvalue.
48 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
49 * Pointers to Arrays:: Pointers to arrays with qualifiers work as expected.
50 * Initializers:: Non-constant initializers.
51 * Compound Literals:: Compound literals give structures, unions
53 * Designated Inits:: Labeling elements of initializers.
54 * Case Ranges:: `case 1 ... 9' and such.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Mixed Declarations:: Mixing declarations and code.
57 * Function Attributes:: Declaring that functions have no side effects,
58 or that they can never return.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Label Attributes:: Specifying attributes on labels.
62 * Enumerator Attributes:: Specifying attributes on enumerators.
63 * Attribute Syntax:: Formal syntax for attributes.
64 * Function Prototypes:: Prototype declarations and old-style definitions.
65 * C++ Comments:: C++ comments are recognized.
66 * Dollar Signs:: Dollar sign is allowed in identifiers.
67 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
68 * Alignment:: Inquiring about the alignment of a type or variable.
69 * Inline:: Defining inline functions (as fast as macros).
70 * Volatiles:: What constitutes an access to a volatile object.
71 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
72 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
73 * Incomplete Enums:: @code{enum foo;}, with details to follow.
74 * Function Names:: Printable strings which are the name of the current
76 * Return Address:: Getting the return or frame address of a function.
77 * Vector Extensions:: Using vector instructions through built-in functions.
78 * Offsetof:: Special syntax for implementing @code{offsetof}.
79 * __sync Builtins:: Legacy built-in functions for atomic memory access.
80 * __atomic Builtins:: Atomic built-in functions with memory model.
81 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
82 arithmetic overflow checking.
83 * x86 specific memory model extensions for transactional memory:: x86 memory models.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
87 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
88 * Other Builtins:: Other built-in functions.
89 * Target Builtins:: Built-in functions specific to particular targets.
90 * Target Format Checks:: Format checks specific to particular targets.
91 * Pragmas:: Pragmas accepted by GCC.
92 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
93 * Thread-Local:: Per-thread variables.
94 * Binary constants:: Binary constants using the @samp{0b} prefix.
98 @section Statements and Declarations in Expressions
99 @cindex statements inside expressions
100 @cindex declarations inside expressions
101 @cindex expressions containing statements
102 @cindex macros, statements in expressions
104 @c the above section title wrapped and causes an underfull hbox.. i
105 @c changed it from "within" to "in". --mew 4feb93
106 A compound statement enclosed in parentheses may appear as an expression
107 in GNU C@. This allows you to use loops, switches, and local variables
108 within an expression.
110 Recall that a compound statement is a sequence of statements surrounded
111 by braces; in this construct, parentheses go around the braces. For
115 (@{ int y = foo (); int z;
122 is a valid (though slightly more complex than necessary) expression
123 for the absolute value of @code{foo ()}.
125 The last thing in the compound statement should be an expression
126 followed by a semicolon; the value of this subexpression serves as the
127 value of the entire construct. (If you use some other kind of statement
128 last within the braces, the construct has type @code{void}, and thus
129 effectively no value.)
131 This feature is especially useful in making macro definitions ``safe'' (so
132 that they evaluate each operand exactly once). For example, the
133 ``maximum'' function is commonly defined as a macro in standard C as
137 #define max(a,b) ((a) > (b) ? (a) : (b))
141 @cindex side effects, macro argument
142 But this definition computes either @var{a} or @var{b} twice, with bad
143 results if the operand has side effects. In GNU C, if you know the
144 type of the operands (here taken as @code{int}), you can define
145 the macro safely as follows:
148 #define maxint(a,b) \
149 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
152 Embedded statements are not allowed in constant expressions, such as
153 the value of an enumeration constant, the width of a bit-field, or
154 the initial value of a static variable.
156 If you don't know the type of the operand, you can still do this, but you
157 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
159 In G++, the result value of a statement expression undergoes array and
160 function pointer decay, and is returned by value to the enclosing
161 expression. For instance, if @code{A} is a class, then
170 constructs a temporary @code{A} object to hold the result of the
171 statement expression, and that is used to invoke @code{Foo}.
172 Therefore the @code{this} pointer observed by @code{Foo} is not the
175 In a statement expression, any temporaries created within a statement
176 are destroyed at that statement's end. This makes statement
177 expressions inside macros slightly different from function calls. In
178 the latter case temporaries introduced during argument evaluation are
179 destroyed at the end of the statement that includes the function
180 call. In the statement expression case they are destroyed during
181 the statement expression. For instance,
184 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
185 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
195 has different places where temporaries are destroyed. For the
196 @code{macro} case, the temporary @code{X} is destroyed just after
197 the initialization of @code{b}. In the @code{function} case that
198 temporary is destroyed when the function returns.
200 These considerations mean that it is probably a bad idea to use
201 statement expressions of this form in header files that are designed to
202 work with C++. (Note that some versions of the GNU C Library contained
203 header files using statement expressions that lead to precisely this
206 Jumping into a statement expression with @code{goto} or using a
207 @code{switch} statement outside the statement expression with a
208 @code{case} or @code{default} label inside the statement expression is
209 not permitted. Jumping into a statement expression with a computed
210 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
211 Jumping out of a statement expression is permitted, but if the
212 statement expression is part of a larger expression then it is
213 unspecified which other subexpressions of that expression have been
214 evaluated except where the language definition requires certain
215 subexpressions to be evaluated before or after the statement
216 expression. In any case, as with a function call, the evaluation of a
217 statement expression is not interleaved with the evaluation of other
218 parts of the containing expression. For example,
221 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
225 calls @code{foo} and @code{bar1} and does not call @code{baz} but
226 may or may not call @code{bar2}. If @code{bar2} is called, it is
227 called after @code{foo} and before @code{bar1}.
230 @section Locally Declared Labels
232 @cindex macros, local labels
234 GCC allows you to declare @dfn{local labels} in any nested block
235 scope. A local label is just like an ordinary label, but you can
236 only reference it (with a @code{goto} statement, or by taking its
237 address) within the block in which it is declared.
239 A local label declaration looks like this:
242 __label__ @var{label};
249 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
252 Local label declarations must come at the beginning of the block,
253 before any ordinary declarations or statements.
255 The label declaration defines the label @emph{name}, but does not define
256 the label itself. You must do this in the usual way, with
257 @code{@var{label}:}, within the statements of the statement expression.
259 The local label feature is useful for complex macros. If a macro
260 contains nested loops, a @code{goto} can be useful for breaking out of
261 them. However, an ordinary label whose scope is the whole function
262 cannot be used: if the macro can be expanded several times in one
263 function, the label is multiply defined in that function. A
264 local label avoids this problem. For example:
267 #define SEARCH(value, array, target) \
270 typeof (target) _SEARCH_target = (target); \
271 typeof (*(array)) *_SEARCH_array = (array); \
274 for (i = 0; i < max; i++) \
275 for (j = 0; j < max; j++) \
276 if (_SEARCH_array[i][j] == _SEARCH_target) \
277 @{ (value) = i; goto found; @} \
283 This could also be written using a statement expression:
286 #define SEARCH(array, target) \
289 typeof (target) _SEARCH_target = (target); \
290 typeof (*(array)) *_SEARCH_array = (array); \
293 for (i = 0; i < max; i++) \
294 for (j = 0; j < max; j++) \
295 if (_SEARCH_array[i][j] == _SEARCH_target) \
296 @{ value = i; goto found; @} \
303 Local label declarations also make the labels they declare visible to
304 nested functions, if there are any. @xref{Nested Functions}, for details.
306 @node Labels as Values
307 @section Labels as Values
308 @cindex labels as values
309 @cindex computed gotos
310 @cindex goto with computed label
311 @cindex address of a label
313 You can get the address of a label defined in the current function
314 (or a containing function) with the unary operator @samp{&&}. The
315 value has type @code{void *}. This value is a constant and can be used
316 wherever a constant of that type is valid. For example:
324 To use these values, you need to be able to jump to one. This is done
325 with the computed goto statement@footnote{The analogous feature in
326 Fortran is called an assigned goto, but that name seems inappropriate in
327 C, where one can do more than simply store label addresses in label
328 variables.}, @code{goto *@var{exp};}. For example,
335 Any expression of type @code{void *} is allowed.
337 One way of using these constants is in initializing a static array that
338 serves as a jump table:
341 static void *array[] = @{ &&foo, &&bar, &&hack @};
345 Then you can select a label with indexing, like this:
352 Note that this does not check whether the subscript is in bounds---array
353 indexing in C never does that.
355 Such an array of label values serves a purpose much like that of the
356 @code{switch} statement. The @code{switch} statement is cleaner, so
357 use that rather than an array unless the problem does not fit a
358 @code{switch} statement very well.
360 Another use of label values is in an interpreter for threaded code.
361 The labels within the interpreter function can be stored in the
362 threaded code for super-fast dispatching.
364 You may not use this mechanism to jump to code in a different function.
365 If you do that, totally unpredictable things happen. The best way to
366 avoid this is to store the label address only in automatic variables and
367 never pass it as an argument.
369 An alternate way to write the above example is
372 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
374 goto *(&&foo + array[i]);
378 This is more friendly to code living in shared libraries, as it reduces
379 the number of dynamic relocations that are needed, and by consequence,
380 allows the data to be read-only.
381 This alternative with label differences is not supported for the AVR target,
382 please use the first approach for AVR programs.
384 The @code{&&foo} expressions for the same label might have different
385 values if the containing function is inlined or cloned. If a program
386 relies on them being always the same,
387 @code{__attribute__((__noinline__,__noclone__))} should be used to
388 prevent inlining and cloning. If @code{&&foo} is used in a static
389 variable initializer, inlining and cloning is forbidden.
391 @node Nested Functions
392 @section Nested Functions
393 @cindex nested functions
394 @cindex downward funargs
397 A @dfn{nested function} is a function defined inside another function.
398 Nested functions are supported as an extension in GNU C, but are not
399 supported by GNU C++.
401 The nested function's name is local to the block where it is defined.
402 For example, here we define a nested function named @code{square}, and
407 foo (double a, double b)
409 double square (double z) @{ return z * z; @}
411 return square (a) + square (b);
416 The nested function can access all the variables of the containing
417 function that are visible at the point of its definition. This is
418 called @dfn{lexical scoping}. For example, here we show a nested
419 function which uses an inherited variable named @code{offset}:
423 bar (int *array, int offset, int size)
425 int access (int *array, int index)
426 @{ return array[index + offset]; @}
429 for (i = 0; i < size; i++)
430 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
435 Nested function definitions are permitted within functions in the places
436 where variable definitions are allowed; that is, in any block, mixed
437 with the other declarations and statements in the block.
439 It is possible to call the nested function from outside the scope of its
440 name by storing its address or passing the address to another function:
443 hack (int *array, int size)
445 void store (int index, int value)
446 @{ array[index] = value; @}
448 intermediate (store, size);
452 Here, the function @code{intermediate} receives the address of
453 @code{store} as an argument. If @code{intermediate} calls @code{store},
454 the arguments given to @code{store} are used to store into @code{array}.
455 But this technique works only so long as the containing function
456 (@code{hack}, in this example) does not exit.
458 If you try to call the nested function through its address after the
459 containing function exits, all hell breaks loose. If you try
460 to call it after a containing scope level exits, and if it refers
461 to some of the variables that are no longer in scope, you may be lucky,
462 but it's not wise to take the risk. If, however, the nested function
463 does not refer to anything that has gone out of scope, you should be
466 GCC implements taking the address of a nested function using a technique
467 called @dfn{trampolines}. This technique was described in
468 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
469 C++ Conference Proceedings, October 17-21, 1988).
471 A nested function can jump to a label inherited from a containing
472 function, provided the label is explicitly declared in the containing
473 function (@pxref{Local Labels}). Such a jump returns instantly to the
474 containing function, exiting the nested function that did the
475 @code{goto} and any intermediate functions as well. Here is an example:
479 bar (int *array, int offset, int size)
482 int access (int *array, int index)
486 return array[index + offset];
490 for (i = 0; i < size; i++)
491 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
495 /* @r{Control comes here from @code{access}
496 if it detects an error.} */
503 A nested function always has no linkage. Declaring one with
504 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
505 before its definition, use @code{auto} (which is otherwise meaningless
506 for function declarations).
509 bar (int *array, int offset, int size)
512 auto int access (int *, int);
514 int access (int *array, int index)
518 return array[index + offset];
524 @node Constructing Calls
525 @section Constructing Function Calls
526 @cindex constructing calls
527 @cindex forwarding calls
529 Using the built-in functions described below, you can record
530 the arguments a function received, and call another function
531 with the same arguments, without knowing the number or types
534 You can also record the return value of that function call,
535 and later return that value, without knowing what data type
536 the function tried to return (as long as your caller expects
539 However, these built-in functions may interact badly with some
540 sophisticated features or other extensions of the language. It
541 is, therefore, not recommended to use them outside very simple
542 functions acting as mere forwarders for their arguments.
544 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
545 This built-in function returns a pointer to data
546 describing how to perform a call with the same arguments as are passed
547 to the current function.
549 The function saves the arg pointer register, structure value address,
550 and all registers that might be used to pass arguments to a function
551 into a block of memory allocated on the stack. Then it returns the
552 address of that block.
555 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
556 This built-in function invokes @var{function}
557 with a copy of the parameters described by @var{arguments}
560 The value of @var{arguments} should be the value returned by
561 @code{__builtin_apply_args}. The argument @var{size} specifies the size
562 of the stack argument data, in bytes.
564 This function returns a pointer to data describing
565 how to return whatever value is returned by @var{function}. The data
566 is saved in a block of memory allocated on the stack.
568 It is not always simple to compute the proper value for @var{size}. The
569 value is used by @code{__builtin_apply} to compute the amount of data
570 that should be pushed on the stack and copied from the incoming argument
574 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
575 This built-in function returns the value described by @var{result} from
576 the containing function. You should specify, for @var{result}, a value
577 returned by @code{__builtin_apply}.
580 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
581 This built-in function represents all anonymous arguments of an inline
582 function. It can be used only in inline functions that are always
583 inlined, never compiled as a separate function, such as those using
584 @code{__attribute__ ((__always_inline__))} or
585 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
586 It must be only passed as last argument to some other function
587 with variable arguments. This is useful for writing small wrapper
588 inlines for variable argument functions, when using preprocessor
589 macros is undesirable. For example:
591 extern int myprintf (FILE *f, const char *format, ...);
592 extern inline __attribute__ ((__gnu_inline__)) int
593 myprintf (FILE *f, const char *format, ...)
595 int r = fprintf (f, "myprintf: ");
598 int s = fprintf (f, format, __builtin_va_arg_pack ());
606 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
607 This built-in function returns the number of anonymous arguments of
608 an inline function. It can be used only in inline functions that
609 are always inlined, never compiled as a separate function, such
610 as those using @code{__attribute__ ((__always_inline__))} or
611 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
612 For example following does link- or run-time checking of open
613 arguments for optimized code:
616 extern inline __attribute__((__gnu_inline__)) int
617 myopen (const char *path, int oflag, ...)
619 if (__builtin_va_arg_pack_len () > 1)
620 warn_open_too_many_arguments ();
622 if (__builtin_constant_p (oflag))
624 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
626 warn_open_missing_mode ();
627 return __open_2 (path, oflag);
629 return open (path, oflag, __builtin_va_arg_pack ());
632 if (__builtin_va_arg_pack_len () < 1)
633 return __open_2 (path, oflag);
635 return open (path, oflag, __builtin_va_arg_pack ());
642 @section Referring to a Type with @code{typeof}
645 @cindex macros, types of arguments
647 Another way to refer to the type of an expression is with @code{typeof}.
648 The syntax of using of this keyword looks like @code{sizeof}, but the
649 construct acts semantically like a type name defined with @code{typedef}.
651 There are two ways of writing the argument to @code{typeof}: with an
652 expression or with a type. Here is an example with an expression:
659 This assumes that @code{x} is an array of pointers to functions;
660 the type described is that of the values of the functions.
662 Here is an example with a typename as the argument:
669 Here the type described is that of pointers to @code{int}.
671 If you are writing a header file that must work when included in ISO C
672 programs, write @code{__typeof__} instead of @code{typeof}.
673 @xref{Alternate Keywords}.
675 A @code{typeof} construct can be used anywhere a typedef name can be
676 used. For example, you can use it in a declaration, in a cast, or inside
677 of @code{sizeof} or @code{typeof}.
679 The operand of @code{typeof} is evaluated for its side effects if and
680 only if it is an expression of variably modified type or the name of
683 @code{typeof} is often useful in conjunction with
684 statement expressions (@pxref{Statement Exprs}).
685 Here is how the two together can
686 be used to define a safe ``maximum'' macro which operates on any
687 arithmetic type and evaluates each of its arguments exactly once:
691 (@{ typeof (a) _a = (a); \
692 typeof (b) _b = (b); \
693 _a > _b ? _a : _b; @})
696 @cindex underscores in variables in macros
697 @cindex @samp{_} in variables in macros
698 @cindex local variables in macros
699 @cindex variables, local, in macros
700 @cindex macros, local variables in
702 The reason for using names that start with underscores for the local
703 variables is to avoid conflicts with variable names that occur within the
704 expressions that are substituted for @code{a} and @code{b}. Eventually we
705 hope to design a new form of declaration syntax that allows you to declare
706 variables whose scopes start only after their initializers; this will be a
707 more reliable way to prevent such conflicts.
710 Some more examples of the use of @code{typeof}:
714 This declares @code{y} with the type of what @code{x} points to.
721 This declares @code{y} as an array of such values.
728 This declares @code{y} as an array of pointers to characters:
731 typeof (typeof (char *)[4]) y;
735 It is equivalent to the following traditional C declaration:
741 To see the meaning of the declaration using @code{typeof}, and why it
742 might be a useful way to write, rewrite it with these macros:
745 #define pointer(T) typeof(T *)
746 #define array(T, N) typeof(T [N])
750 Now the declaration can be rewritten this way:
753 array (pointer (char), 4) y;
757 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
758 pointers to @code{char}.
761 In GNU C, but not GNU C++, you may also declare the type of a variable
762 as @code{__auto_type}. In that case, the declaration must declare
763 only one variable, whose declarator must just be an identifier, the
764 declaration must be initialized, and the type of the variable is
765 determined by the initializer; the name of the variable is not in
766 scope until after the initializer. (In C++, you should use C++11
767 @code{auto} for this purpose.) Using @code{__auto_type}, the
768 ``maximum'' macro above could be written as:
772 (@{ __auto_type _a = (a); \
773 __auto_type _b = (b); \
774 _a > _b ? _a : _b; @})
777 Using @code{__auto_type} instead of @code{typeof} has two advantages:
780 @item Each argument to the macro appears only once in the expansion of
781 the macro. This prevents the size of the macro expansion growing
782 exponentially when calls to such macros are nested inside arguments of
785 @item If the argument to the macro has variably modified type, it is
786 evaluated only once when using @code{__auto_type}, but twice if
787 @code{typeof} is used.
791 @section Conditionals with Omitted Operands
792 @cindex conditional expressions, extensions
793 @cindex omitted middle-operands
794 @cindex middle-operands, omitted
795 @cindex extensions, @code{?:}
796 @cindex @code{?:} extensions
798 The middle operand in a conditional expression may be omitted. Then
799 if the first operand is nonzero, its value is the value of the conditional
802 Therefore, the expression
809 has the value of @code{x} if that is nonzero; otherwise, the value of
812 This example is perfectly equivalent to
818 @cindex side effect in @code{?:}
819 @cindex @code{?:} side effect
821 In this simple case, the ability to omit the middle operand is not
822 especially useful. When it becomes useful is when the first operand does,
823 or may (if it is a macro argument), contain a side effect. Then repeating
824 the operand in the middle would perform the side effect twice. Omitting
825 the middle operand uses the value already computed without the undesirable
826 effects of recomputing it.
829 @section 128-bit Integers
830 @cindex @code{__int128} data types
832 As an extension the integer scalar type @code{__int128} is supported for
833 targets which have an integer mode wide enough to hold 128 bits.
834 Simply write @code{__int128} for a signed 128-bit integer, or
835 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
836 support in GCC for expressing an integer constant of type @code{__int128}
837 for targets with @code{long long} integer less than 128 bits wide.
840 @section Double-Word Integers
841 @cindex @code{long long} data types
842 @cindex double-word arithmetic
843 @cindex multiprecision arithmetic
844 @cindex @code{LL} integer suffix
845 @cindex @code{ULL} integer suffix
847 ISO C99 supports data types for integers that are at least 64 bits wide,
848 and as an extension GCC supports them in C90 mode and in C++.
849 Simply write @code{long long int} for a signed integer, or
850 @code{unsigned long long int} for an unsigned integer. To make an
851 integer constant of type @code{long long int}, add the suffix @samp{LL}
852 to the integer. To make an integer constant of type @code{unsigned long
853 long int}, add the suffix @samp{ULL} to the integer.
855 You can use these types in arithmetic like any other integer types.
856 Addition, subtraction, and bitwise boolean operations on these types
857 are open-coded on all types of machines. Multiplication is open-coded
858 if the machine supports a fullword-to-doubleword widening multiply
859 instruction. Division and shifts are open-coded only on machines that
860 provide special support. The operations that are not open-coded use
861 special library routines that come with GCC@.
863 There may be pitfalls when you use @code{long long} types for function
864 arguments without function prototypes. If a function
865 expects type @code{int} for its argument, and you pass a value of type
866 @code{long long int}, confusion results because the caller and the
867 subroutine disagree about the number of bytes for the argument.
868 Likewise, if the function expects @code{long long int} and you pass
869 @code{int}. The best way to avoid such problems is to use prototypes.
872 @section Complex Numbers
873 @cindex complex numbers
874 @cindex @code{_Complex} keyword
875 @cindex @code{__complex__} keyword
877 ISO C99 supports complex floating data types, and as an extension GCC
878 supports them in C90 mode and in C++. GCC also supports complex integer data
879 types which are not part of ISO C99. You can declare complex types
880 using the keyword @code{_Complex}. As an extension, the older GNU
881 keyword @code{__complex__} is also supported.
883 For example, @samp{_Complex double x;} declares @code{x} as a
884 variable whose real part and imaginary part are both of type
885 @code{double}. @samp{_Complex short int y;} declares @code{y} to
886 have real and imaginary parts of type @code{short int}; this is not
887 likely to be useful, but it shows that the set of complex types is
890 To write a constant with a complex data type, use the suffix @samp{i} or
891 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
892 has type @code{_Complex float} and @code{3i} has type
893 @code{_Complex int}. Such a constant always has a pure imaginary
894 value, but you can form any complex value you like by adding one to a
895 real constant. This is a GNU extension; if you have an ISO C99
896 conforming C library (such as the GNU C Library), and want to construct complex
897 constants of floating type, you should include @code{<complex.h>} and
898 use the macros @code{I} or @code{_Complex_I} instead.
900 @cindex @code{__real__} keyword
901 @cindex @code{__imag__} keyword
902 To extract the real part of a complex-valued expression @var{exp}, write
903 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
904 extract the imaginary part. This is a GNU extension; for values of
905 floating type, you should use the ISO C99 functions @code{crealf},
906 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
907 @code{cimagl}, declared in @code{<complex.h>} and also provided as
908 built-in functions by GCC@.
910 @cindex complex conjugation
911 The operator @samp{~} performs complex conjugation when used on a value
912 with a complex type. This is a GNU extension; for values of
913 floating type, you should use the ISO C99 functions @code{conjf},
914 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
915 provided as built-in functions by GCC@.
917 GCC can allocate complex automatic variables in a noncontiguous
918 fashion; it's even possible for the real part to be in a register while
919 the imaginary part is on the stack (or vice versa). Only the DWARF 2
920 debug info format can represent this, so use of DWARF 2 is recommended.
921 If you are using the stabs debug info format, GCC describes a noncontiguous
922 complex variable as if it were two separate variables of noncomplex type.
923 If the variable's actual name is @code{foo}, the two fictitious
924 variables are named @code{foo$real} and @code{foo$imag}. You can
925 examine and set these two fictitious variables with your debugger.
928 @section Additional Floating Types
929 @cindex additional floating types
930 @cindex @code{__float80} data type
931 @cindex @code{__float128} data type
932 @cindex @code{__ibm128} data type
933 @cindex @code{w} floating point suffix
934 @cindex @code{q} floating point suffix
935 @cindex @code{W} floating point suffix
936 @cindex @code{Q} floating point suffix
938 As an extension, GNU C supports additional floating
939 types, @code{__float80} and @code{__float128} to support 80-bit
940 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
941 Support for additional types includes the arithmetic operators:
942 add, subtract, multiply, divide; unary arithmetic operators;
943 relational operators; equality operators; and conversions to and from
944 integer and other floating types. Use a suffix @samp{w} or @samp{W}
945 in a literal constant of type @code{__float80} or type
946 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
948 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
949 types using the corresponding internal complex type, @code{XCmode} for
950 @code{__float80} type and @code{TCmode} for @code{__float128} type:
953 typedef _Complex float __attribute__((mode(TC))) _Complex128;
954 typedef _Complex float __attribute__((mode(XC))) _Complex80;
957 On PowerPC Linux, Freebsd and Darwin systems, the default for
958 @code{long double} is to use the IBM extended floating point format
959 that uses a pair of @code{double} values to extend the precision.
960 This means that the mode @code{TCmode} was already used by the
961 traditional IBM long double format, and you would need to use the mode
965 typedef _Complex float __attribute__((mode(KC))) _Complex128;
968 Not all targets support additional floating-point types. @code{__float80}
969 and @code{__float128} types are supported on x86 and IA-64 targets.
970 The @code{__float128} type is supported on hppa HP-UX.
971 The @code{__float128} type is supported on PowerPC systems by default
972 if the vector scalar instruction set (VSX) is enabled.
974 On the PowerPC, @code{__ibm128} provides access to the IBM extended
975 double format, and it is intended to be used by the library functions
976 that handle conversions if/when long double is changed to be IEEE
977 128-bit floating point.
980 @section Half-Precision Floating Point
981 @cindex half-precision floating point
982 @cindex @code{__fp16} data type
984 On ARM targets, GCC supports half-precision (16-bit) floating point via
985 the @code{__fp16} type. You must enable this type explicitly
986 with the @option{-mfp16-format} command-line option in order to use it.
988 ARM supports two incompatible representations for half-precision
989 floating-point values. You must choose one of the representations and
990 use it consistently in your program.
992 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
993 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
994 There are 11 bits of significand precision, approximately 3
997 Specifying @option{-mfp16-format=alternative} selects the ARM
998 alternative format. This representation is similar to the IEEE
999 format, but does not support infinities or NaNs. Instead, the range
1000 of exponents is extended, so that this format can represent normalized
1001 values in the range of @math{2^{-14}} to 131008.
1003 The @code{__fp16} type is a storage format only. For purposes
1004 of arithmetic and other operations, @code{__fp16} values in C or C++
1005 expressions are automatically promoted to @code{float}. In addition,
1006 you cannot declare a function with a return value or parameters
1007 of type @code{__fp16}.
1009 Note that conversions from @code{double} to @code{__fp16}
1010 involve an intermediate conversion to @code{float}. Because
1011 of rounding, this can sometimes produce a different result than a
1014 ARM provides hardware support for conversions between
1015 @code{__fp16} and @code{float} values
1016 as an extension to VFP and NEON (Advanced SIMD). GCC generates
1017 code using these hardware instructions if you compile with
1018 options to select an FPU that provides them;
1019 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1020 in addition to the @option{-mfp16-format} option to select
1021 a half-precision format.
1023 Language-level support for the @code{__fp16} data type is
1024 independent of whether GCC generates code using hardware floating-point
1025 instructions. In cases where hardware support is not specified, GCC
1026 implements conversions between @code{__fp16} and @code{float} values
1030 @section Decimal Floating Types
1031 @cindex decimal floating types
1032 @cindex @code{_Decimal32} data type
1033 @cindex @code{_Decimal64} data type
1034 @cindex @code{_Decimal128} data type
1035 @cindex @code{df} integer suffix
1036 @cindex @code{dd} integer suffix
1037 @cindex @code{dl} integer suffix
1038 @cindex @code{DF} integer suffix
1039 @cindex @code{DD} integer suffix
1040 @cindex @code{DL} integer suffix
1042 As an extension, GNU C supports decimal floating types as
1043 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1044 floating types in GCC will evolve as the draft technical report changes.
1045 Calling conventions for any target might also change. Not all targets
1046 support decimal floating types.
1048 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1049 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1050 @code{float}, @code{double}, and @code{long double} whose radix is not
1051 specified by the C standard but is usually two.
1053 Support for decimal floating types includes the arithmetic operators
1054 add, subtract, multiply, divide; unary arithmetic operators;
1055 relational operators; equality operators; and conversions to and from
1056 integer and other floating types. Use a suffix @samp{df} or
1057 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1058 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1061 GCC support of decimal float as specified by the draft technical report
1066 When the value of a decimal floating type cannot be represented in the
1067 integer type to which it is being converted, the result is undefined
1068 rather than the result value specified by the draft technical report.
1071 GCC does not provide the C library functionality associated with
1072 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1073 @file{wchar.h}, which must come from a separate C library implementation.
1074 Because of this the GNU C compiler does not define macro
1075 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1076 the technical report.
1079 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1080 are supported by the DWARF 2 debug information format.
1086 ISO C99 supports floating-point numbers written not only in the usual
1087 decimal notation, such as @code{1.55e1}, but also numbers such as
1088 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1089 supports this in C90 mode (except in some cases when strictly
1090 conforming) and in C++. In that format the
1091 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1092 mandatory. The exponent is a decimal number that indicates the power of
1093 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1100 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1101 is the same as @code{1.55e1}.
1103 Unlike for floating-point numbers in the decimal notation the exponent
1104 is always required in the hexadecimal notation. Otherwise the compiler
1105 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1106 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1107 extension for floating-point constants of type @code{float}.
1110 @section Fixed-Point Types
1111 @cindex fixed-point types
1112 @cindex @code{_Fract} data type
1113 @cindex @code{_Accum} data type
1114 @cindex @code{_Sat} data type
1115 @cindex @code{hr} fixed-suffix
1116 @cindex @code{r} fixed-suffix
1117 @cindex @code{lr} fixed-suffix
1118 @cindex @code{llr} fixed-suffix
1119 @cindex @code{uhr} fixed-suffix
1120 @cindex @code{ur} fixed-suffix
1121 @cindex @code{ulr} fixed-suffix
1122 @cindex @code{ullr} fixed-suffix
1123 @cindex @code{hk} fixed-suffix
1124 @cindex @code{k} fixed-suffix
1125 @cindex @code{lk} fixed-suffix
1126 @cindex @code{llk} fixed-suffix
1127 @cindex @code{uhk} fixed-suffix
1128 @cindex @code{uk} fixed-suffix
1129 @cindex @code{ulk} fixed-suffix
1130 @cindex @code{ullk} fixed-suffix
1131 @cindex @code{HR} fixed-suffix
1132 @cindex @code{R} fixed-suffix
1133 @cindex @code{LR} fixed-suffix
1134 @cindex @code{LLR} fixed-suffix
1135 @cindex @code{UHR} fixed-suffix
1136 @cindex @code{UR} fixed-suffix
1137 @cindex @code{ULR} fixed-suffix
1138 @cindex @code{ULLR} fixed-suffix
1139 @cindex @code{HK} fixed-suffix
1140 @cindex @code{K} fixed-suffix
1141 @cindex @code{LK} fixed-suffix
1142 @cindex @code{LLK} fixed-suffix
1143 @cindex @code{UHK} fixed-suffix
1144 @cindex @code{UK} fixed-suffix
1145 @cindex @code{ULK} fixed-suffix
1146 @cindex @code{ULLK} fixed-suffix
1148 As an extension, GNU C supports fixed-point types as
1149 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1150 types in GCC will evolve as the draft technical report changes.
1151 Calling conventions for any target might also change. Not all targets
1152 support fixed-point types.
1154 The fixed-point types are
1155 @code{short _Fract},
1158 @code{long long _Fract},
1159 @code{unsigned short _Fract},
1160 @code{unsigned _Fract},
1161 @code{unsigned long _Fract},
1162 @code{unsigned long long _Fract},
1163 @code{_Sat short _Fract},
1165 @code{_Sat long _Fract},
1166 @code{_Sat long long _Fract},
1167 @code{_Sat unsigned short _Fract},
1168 @code{_Sat unsigned _Fract},
1169 @code{_Sat unsigned long _Fract},
1170 @code{_Sat unsigned long long _Fract},
1171 @code{short _Accum},
1174 @code{long long _Accum},
1175 @code{unsigned short _Accum},
1176 @code{unsigned _Accum},
1177 @code{unsigned long _Accum},
1178 @code{unsigned long long _Accum},
1179 @code{_Sat short _Accum},
1181 @code{_Sat long _Accum},
1182 @code{_Sat long long _Accum},
1183 @code{_Sat unsigned short _Accum},
1184 @code{_Sat unsigned _Accum},
1185 @code{_Sat unsigned long _Accum},
1186 @code{_Sat unsigned long long _Accum}.
1188 Fixed-point data values contain fractional and optional integral parts.
1189 The format of fixed-point data varies and depends on the target machine.
1191 Support for fixed-point types includes:
1194 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1196 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1198 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1200 binary shift operators (@code{<<}, @code{>>})
1202 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1204 equality operators (@code{==}, @code{!=})
1206 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1207 @code{<<=}, @code{>>=})
1209 conversions to and from integer, floating-point, or fixed-point types
1212 Use a suffix in a fixed-point literal constant:
1214 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1215 @code{_Sat short _Fract}
1216 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1217 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1218 @code{_Sat long _Fract}
1219 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1220 @code{_Sat long long _Fract}
1221 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1222 @code{_Sat unsigned short _Fract}
1223 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1224 @code{_Sat unsigned _Fract}
1225 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1226 @code{_Sat unsigned long _Fract}
1227 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1228 and @code{_Sat unsigned long long _Fract}
1229 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1230 @code{_Sat short _Accum}
1231 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1232 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1233 @code{_Sat long _Accum}
1234 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1235 @code{_Sat long long _Accum}
1236 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1237 @code{_Sat unsigned short _Accum}
1238 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1239 @code{_Sat unsigned _Accum}
1240 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1241 @code{_Sat unsigned long _Accum}
1242 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1243 and @code{_Sat unsigned long long _Accum}
1246 GCC support of fixed-point types as specified by the draft technical report
1251 Pragmas to control overflow and rounding behaviors are not implemented.
1254 Fixed-point types are supported by the DWARF 2 debug information format.
1256 @node Named Address Spaces
1257 @section Named Address Spaces
1258 @cindex Named Address Spaces
1260 As an extension, GNU C supports named address spaces as
1261 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1262 address spaces in GCC will evolve as the draft technical report
1263 changes. Calling conventions for any target might also change. At
1264 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1265 address spaces other than the generic address space.
1267 Address space identifiers may be used exactly like any other C type
1268 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1269 document for more details.
1271 @anchor{AVR Named Address Spaces}
1272 @subsection AVR Named Address Spaces
1274 On the AVR target, there are several address spaces that can be used
1275 in order to put read-only data into the flash memory and access that
1276 data by means of the special instructions @code{LPM} or @code{ELPM}
1277 needed to read from flash.
1279 Per default, any data including read-only data is located in RAM
1280 (the generic address space) so that non-generic address spaces are
1281 needed to locate read-only data in flash memory
1282 @emph{and} to generate the right instructions to access this data
1283 without using (inline) assembler code.
1287 @cindex @code{__flash} AVR Named Address Spaces
1288 The @code{__flash} qualifier locates data in the
1289 @code{.progmem.data} section. Data is read using the @code{LPM}
1290 instruction. Pointers to this address space are 16 bits wide.
1297 @cindex @code{__flash1} AVR Named Address Spaces
1298 @cindex @code{__flash2} AVR Named Address Spaces
1299 @cindex @code{__flash3} AVR Named Address Spaces
1300 @cindex @code{__flash4} AVR Named Address Spaces
1301 @cindex @code{__flash5} AVR Named Address Spaces
1302 These are 16-bit address spaces locating data in section
1303 @code{.progmem@var{N}.data} where @var{N} refers to
1304 address space @code{__flash@var{N}}.
1305 The compiler sets the @code{RAMPZ} segment register appropriately
1306 before reading data by means of the @code{ELPM} instruction.
1309 @cindex @code{__memx} AVR Named Address Spaces
1310 This is a 24-bit address space that linearizes flash and RAM:
1311 If the high bit of the address is set, data is read from
1312 RAM using the lower two bytes as RAM address.
1313 If the high bit of the address is clear, data is read from flash
1314 with @code{RAMPZ} set according to the high byte of the address.
1315 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1317 Objects in this address space are located in @code{.progmemx.data}.
1323 char my_read (const __flash char ** p)
1325 /* p is a pointer to RAM that points to a pointer to flash.
1326 The first indirection of p reads that flash pointer
1327 from RAM and the second indirection reads a char from this
1333 /* Locate array[] in flash memory */
1334 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1340 /* Return 17 by reading from flash memory */
1341 return array[array[i]];
1346 For each named address space supported by avr-gcc there is an equally
1347 named but uppercase built-in macro defined.
1348 The purpose is to facilitate testing if respective address space
1349 support is available or not:
1353 const __flash int var = 1;
1360 #include <avr/pgmspace.h> /* From AVR-LibC */
1362 const int var PROGMEM = 1;
1366 return (int) pgm_read_word (&var);
1368 #endif /* __FLASH */
1372 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1373 locates data in flash but
1374 accesses to these data read from generic address space, i.e.@:
1376 so that you need special accessors like @code{pgm_read_byte}
1377 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1378 together with attribute @code{progmem}.
1381 @b{Limitations and caveats}
1385 Reading across the 64@tie{}KiB section boundary of
1386 the @code{__flash} or @code{__flash@var{N}} address spaces
1387 shows undefined behavior. The only address space that
1388 supports reading across the 64@tie{}KiB flash segment boundaries is
1392 If you use one of the @code{__flash@var{N}} address spaces
1393 you must arrange your linker script to locate the
1394 @code{.progmem@var{N}.data} sections according to your needs.
1397 Any data or pointers to the non-generic address spaces must
1398 be qualified as @code{const}, i.e.@: as read-only data.
1399 This still applies if the data in one of these address
1400 spaces like software version number or calibration lookup table are intended to
1401 be changed after load time by, say, a boot loader. In this case
1402 the right qualification is @code{const} @code{volatile} so that the compiler
1403 must not optimize away known values or insert them
1404 as immediates into operands of instructions.
1407 The following code initializes a variable @code{pfoo}
1408 located in static storage with a 24-bit address:
1410 extern const __memx char foo;
1411 const __memx void *pfoo = &foo;
1415 Such code requires at least binutils 2.23, see
1416 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1420 @subsection M32C Named Address Spaces
1421 @cindex @code{__far} M32C Named Address Spaces
1423 On the M32C target, with the R8C and M16C CPU variants, variables
1424 qualified with @code{__far} are accessed using 32-bit addresses in
1425 order to access memory beyond the first 64@tie{}Ki bytes. If
1426 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1429 @subsection RL78 Named Address Spaces
1430 @cindex @code{__far} RL78 Named Address Spaces
1432 On the RL78 target, variables qualified with @code{__far} are accessed
1433 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1434 addresses. Non-far variables are assumed to appear in the topmost
1435 64@tie{}KiB of the address space.
1437 @subsection SPU Named Address Spaces
1438 @cindex @code{__ea} SPU Named Address Spaces
1440 On the SPU target variables may be declared as
1441 belonging to another address space by qualifying the type with the
1442 @code{__ea} address space identifier:
1449 The compiler generates special code to access the variable @code{i}.
1450 It may use runtime library
1451 support, or generate special machine instructions to access that address
1454 @subsection x86 Named Address Spaces
1455 @cindex x86 named address spaces
1457 On the x86 target, variables may be declared as being relative
1458 to the @code{%fs} or @code{%gs} segments.
1463 @cindex @code{__seg_fs} x86 named address space
1464 @cindex @code{__seg_gs} x86 named address space
1465 The object is accessed with the respective segment override prefix.
1467 The respective segment base must be set via some method specific to
1468 the operating system. Rather than require an expensive system call
1469 to retrieve the segment base, these address spaces are not considered
1470 to be subspaces of the generic (flat) address space. This means that
1471 explicit casts are required to convert pointers between these address
1472 spaces and the generic address space. In practice the application
1473 should cast to @code{uintptr_t} and apply the segment base offset
1474 that it installed previously.
1476 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1477 defined when these address spaces are supported.
1480 @cindex @code{__seg_tls} x86 named address space
1481 Some operating systems define either the @code{%fs} or @code{%gs}
1482 segment as the thread-local storage base for each thread. Objects
1483 within this address space are accessed with the appropriate
1484 segment override prefix.
1486 The pointer located at address 0 within the segment contains the
1487 offset of the segment within the generic address space. Thus this
1488 address space is considered a subspace of the generic address space,
1489 and the known segment offset is applied when converting addresses
1490 to and from the generic address space.
1492 The preprocessor symbol @code{__SEG_TLS} is defined when this
1493 address space is supported.
1498 @section Arrays of Length Zero
1499 @cindex arrays of length zero
1500 @cindex zero-length arrays
1501 @cindex length-zero arrays
1502 @cindex flexible array members
1504 Zero-length arrays are allowed in GNU C@. They are very useful as the
1505 last element of a structure that is really a header for a variable-length
1514 struct line *thisline = (struct line *)
1515 malloc (sizeof (struct line) + this_length);
1516 thisline->length = this_length;
1519 In ISO C90, you would have to give @code{contents} a length of 1, which
1520 means either you waste space or complicate the argument to @code{malloc}.
1522 In ISO C99, you would use a @dfn{flexible array member}, which is
1523 slightly different in syntax and semantics:
1527 Flexible array members are written as @code{contents[]} without
1531 Flexible array members have incomplete type, and so the @code{sizeof}
1532 operator may not be applied. As a quirk of the original implementation
1533 of zero-length arrays, @code{sizeof} evaluates to zero.
1536 Flexible array members may only appear as the last member of a
1537 @code{struct} that is otherwise non-empty.
1540 A structure containing a flexible array member, or a union containing
1541 such a structure (possibly recursively), may not be a member of a
1542 structure or an element of an array. (However, these uses are
1543 permitted by GCC as extensions.)
1546 Non-empty initialization of zero-length
1547 arrays is treated like any case where there are more initializer
1548 elements than the array holds, in that a suitable warning about ``excess
1549 elements in array'' is given, and the excess elements (all of them, in
1550 this case) are ignored.
1552 GCC allows static initialization of flexible array members.
1553 This is equivalent to defining a new structure containing the original
1554 structure followed by an array of sufficient size to contain the data.
1555 E.g.@: in the following, @code{f1} is constructed as if it were declared
1561 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1564 struct f1 f1; int data[3];
1565 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1569 The convenience of this extension is that @code{f1} has the desired
1570 type, eliminating the need to consistently refer to @code{f2.f1}.
1572 This has symmetry with normal static arrays, in that an array of
1573 unknown size is also written with @code{[]}.
1575 Of course, this extension only makes sense if the extra data comes at
1576 the end of a top-level object, as otherwise we would be overwriting
1577 data at subsequent offsets. To avoid undue complication and confusion
1578 with initialization of deeply nested arrays, we simply disallow any
1579 non-empty initialization except when the structure is the top-level
1580 object. For example:
1583 struct foo @{ int x; int y[]; @};
1584 struct bar @{ struct foo z; @};
1586 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1587 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1588 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1589 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1592 @node Empty Structures
1593 @section Structures with No Members
1594 @cindex empty structures
1595 @cindex zero-size structures
1597 GCC permits a C structure to have no members:
1604 The structure has size zero. In C++, empty structures are part
1605 of the language. G++ treats empty structures as if they had a single
1606 member of type @code{char}.
1608 @node Variable Length
1609 @section Arrays of Variable Length
1610 @cindex variable-length arrays
1611 @cindex arrays of variable length
1614 Variable-length automatic arrays are allowed in ISO C99, and as an
1615 extension GCC accepts them in C90 mode and in C++. These arrays are
1616 declared like any other automatic arrays, but with a length that is not
1617 a constant expression. The storage is allocated at the point of
1618 declaration and deallocated when the block scope containing the declaration
1624 concat_fopen (char *s1, char *s2, char *mode)
1626 char str[strlen (s1) + strlen (s2) + 1];
1629 return fopen (str, mode);
1633 @cindex scope of a variable length array
1634 @cindex variable-length array scope
1635 @cindex deallocating variable length arrays
1636 Jumping or breaking out of the scope of the array name deallocates the
1637 storage. Jumping into the scope is not allowed; you get an error
1640 @cindex variable-length array in a structure
1641 As an extension, GCC accepts variable-length arrays as a member of
1642 a structure or a union. For example:
1648 struct S @{ int x[n]; @};
1652 @cindex @code{alloca} vs variable-length arrays
1653 You can use the function @code{alloca} to get an effect much like
1654 variable-length arrays. The function @code{alloca} is available in
1655 many other C implementations (but not in all). On the other hand,
1656 variable-length arrays are more elegant.
1658 There are other differences between these two methods. Space allocated
1659 with @code{alloca} exists until the containing @emph{function} returns.
1660 The space for a variable-length array is deallocated as soon as the array
1661 name's scope ends. (If you use both variable-length arrays and
1662 @code{alloca} in the same function, deallocation of a variable-length array
1663 also deallocates anything more recently allocated with @code{alloca}.)
1665 You can also use variable-length arrays as arguments to functions:
1669 tester (int len, char data[len][len])
1675 The length of an array is computed once when the storage is allocated
1676 and is remembered for the scope of the array in case you access it with
1679 If you want to pass the array first and the length afterward, you can
1680 use a forward declaration in the parameter list---another GNU extension.
1684 tester (int len; char data[len][len], int len)
1690 @cindex parameter forward declaration
1691 The @samp{int len} before the semicolon is a @dfn{parameter forward
1692 declaration}, and it serves the purpose of making the name @code{len}
1693 known when the declaration of @code{data} is parsed.
1695 You can write any number of such parameter forward declarations in the
1696 parameter list. They can be separated by commas or semicolons, but the
1697 last one must end with a semicolon, which is followed by the ``real''
1698 parameter declarations. Each forward declaration must match a ``real''
1699 declaration in parameter name and data type. ISO C99 does not support
1700 parameter forward declarations.
1702 @node Variadic Macros
1703 @section Macros with a Variable Number of Arguments.
1704 @cindex variable number of arguments
1705 @cindex macro with variable arguments
1706 @cindex rest argument (in macro)
1707 @cindex variadic macros
1709 In the ISO C standard of 1999, a macro can be declared to accept a
1710 variable number of arguments much as a function can. The syntax for
1711 defining the macro is similar to that of a function. Here is an
1715 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1719 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1720 such a macro, it represents the zero or more tokens until the closing
1721 parenthesis that ends the invocation, including any commas. This set of
1722 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1723 wherever it appears. See the CPP manual for more information.
1725 GCC has long supported variadic macros, and used a different syntax that
1726 allowed you to give a name to the variable arguments just like any other
1727 argument. Here is an example:
1730 #define debug(format, args...) fprintf (stderr, format, args)
1734 This is in all ways equivalent to the ISO C example above, but arguably
1735 more readable and descriptive.
1737 GNU CPP has two further variadic macro extensions, and permits them to
1738 be used with either of the above forms of macro definition.
1740 In standard C, you are not allowed to leave the variable argument out
1741 entirely; but you are allowed to pass an empty argument. For example,
1742 this invocation is invalid in ISO C, because there is no comma after
1749 GNU CPP permits you to completely omit the variable arguments in this
1750 way. In the above examples, the compiler would complain, though since
1751 the expansion of the macro still has the extra comma after the format
1754 To help solve this problem, CPP behaves specially for variable arguments
1755 used with the token paste operator, @samp{##}. If instead you write
1758 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1762 and if the variable arguments are omitted or empty, the @samp{##}
1763 operator causes the preprocessor to remove the comma before it. If you
1764 do provide some variable arguments in your macro invocation, GNU CPP
1765 does not complain about the paste operation and instead places the
1766 variable arguments after the comma. Just like any other pasted macro
1767 argument, these arguments are not macro expanded.
1769 @node Escaped Newlines
1770 @section Slightly Looser Rules for Escaped Newlines
1771 @cindex escaped newlines
1772 @cindex newlines (escaped)
1774 The preprocessor treatment of escaped newlines is more relaxed
1775 than that specified by the C90 standard, which requires the newline
1776 to immediately follow a backslash.
1777 GCC's implementation allows whitespace in the form
1778 of spaces, horizontal and vertical tabs, and form feeds between the
1779 backslash and the subsequent newline. The preprocessor issues a
1780 warning, but treats it as a valid escaped newline and combines the two
1781 lines to form a single logical line. This works within comments and
1782 tokens, as well as between tokens. Comments are @emph{not} treated as
1783 whitespace for the purposes of this relaxation, since they have not
1784 yet been replaced with spaces.
1787 @section Non-Lvalue Arrays May Have Subscripts
1788 @cindex subscripting
1789 @cindex arrays, non-lvalue
1791 @cindex subscripting and function values
1792 In ISO C99, arrays that are not lvalues still decay to pointers, and
1793 may be subscripted, although they may not be modified or used after
1794 the next sequence point and the unary @samp{&} operator may not be
1795 applied to them. As an extension, GNU C allows such arrays to be
1796 subscripted in C90 mode, though otherwise they do not decay to
1797 pointers outside C99 mode. For example,
1798 this is valid in GNU C though not valid in C90:
1802 struct foo @{int a[4];@};
1808 return f().a[index];
1814 @section Arithmetic on @code{void}- and Function-Pointers
1815 @cindex void pointers, arithmetic
1816 @cindex void, size of pointer to
1817 @cindex function pointers, arithmetic
1818 @cindex function, size of pointer to
1820 In GNU C, addition and subtraction operations are supported on pointers to
1821 @code{void} and on pointers to functions. This is done by treating the
1822 size of a @code{void} or of a function as 1.
1824 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1825 and on function types, and returns 1.
1827 @opindex Wpointer-arith
1828 The option @option{-Wpointer-arith} requests a warning if these extensions
1831 @node Pointers to Arrays
1832 @section Pointers to Arrays with Qualifiers Work as Expected
1833 @cindex pointers to arrays
1834 @cindex const qualifier
1836 In GNU C, pointers to arrays with qualifiers work similar to pointers
1837 to other qualified types. For example, a value of type @code{int (*)[5]}
1838 can be used to initialize a variable of type @code{const int (*)[5]}.
1839 These types are incompatible in ISO C because the @code{const} qualifier
1840 is formally attached to the element type of the array and not the
1845 transpose (int N, int M, double out[M][N], const double in[N][M]);
1849 transpose(3, 2, y, x);
1853 @section Non-Constant Initializers
1854 @cindex initializers, non-constant
1855 @cindex non-constant initializers
1857 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1858 automatic variable are not required to be constant expressions in GNU C@.
1859 Here is an example of an initializer with run-time varying elements:
1862 foo (float f, float g)
1864 float beat_freqs[2] = @{ f-g, f+g @};
1869 @node Compound Literals
1870 @section Compound Literals
1871 @cindex constructor expressions
1872 @cindex initializations in expressions
1873 @cindex structures, constructor expression
1874 @cindex expressions, constructor
1875 @cindex compound literals
1876 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1878 ISO C99 supports compound literals. A compound literal looks like
1879 a cast containing an initializer. Its value is an object of the
1880 type specified in the cast, containing the elements specified in
1881 the initializer; it is an lvalue. As an extension, GCC supports
1882 compound literals in C90 mode and in C++, though the semantics are
1883 somewhat different in C++.
1885 Usually, the specified type is a structure. Assume that
1886 @code{struct foo} and @code{structure} are declared as shown:
1889 struct foo @{int a; char b[2];@} structure;
1893 Here is an example of constructing a @code{struct foo} with a compound literal:
1896 structure = ((struct foo) @{x + y, 'a', 0@});
1900 This is equivalent to writing the following:
1904 struct foo temp = @{x + y, 'a', 0@};
1909 You can also construct an array, though this is dangerous in C++, as
1910 explained below. If all the elements of the compound literal are
1911 (made up of) simple constant expressions, suitable for use in
1912 initializers of objects of static storage duration, then the compound
1913 literal can be coerced to a pointer to its first element and used in
1914 such an initializer, as shown here:
1917 char **foo = (char *[]) @{ "x", "y", "z" @};
1920 Compound literals for scalar types and union types are
1921 also allowed, but then the compound literal is equivalent
1924 As a GNU extension, GCC allows initialization of objects with static storage
1925 duration by compound literals (which is not possible in ISO C99, because
1926 the initializer is not a constant).
1927 It is handled as if the object is initialized only with the bracket
1928 enclosed list if the types of the compound literal and the object match.
1929 The initializer list of the compound literal must be constant.
1930 If the object being initialized has array type of unknown size, the size is
1931 determined by compound literal size.
1934 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1935 static int y[] = (int []) @{1, 2, 3@};
1936 static int z[] = (int [3]) @{1@};
1940 The above lines are equivalent to the following:
1942 static struct foo x = @{1, 'a', 'b'@};
1943 static int y[] = @{1, 2, 3@};
1944 static int z[] = @{1, 0, 0@};
1947 In C, a compound literal designates an unnamed object with static or
1948 automatic storage duration. In C++, a compound literal designates a
1949 temporary object, which only lives until the end of its
1950 full-expression. As a result, well-defined C code that takes the
1951 address of a subobject of a compound literal can be undefined in C++,
1952 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1953 For instance, if the array compound literal example above appeared
1954 inside a function, any subsequent use of @samp{foo} in C++ has
1955 undefined behavior because the lifetime of the array ends after the
1956 declaration of @samp{foo}.
1958 As an optimization, the C++ compiler sometimes gives array compound
1959 literals longer lifetimes: when the array either appears outside a
1960 function or has const-qualified type. If @samp{foo} and its
1961 initializer had elements of @samp{char *const} type rather than
1962 @samp{char *}, or if @samp{foo} were a global variable, the array
1963 would have static storage duration. But it is probably safest just to
1964 avoid the use of array compound literals in code compiled as C++.
1966 @node Designated Inits
1967 @section Designated Initializers
1968 @cindex initializers with labeled elements
1969 @cindex labeled elements in initializers
1970 @cindex case labels in initializers
1971 @cindex designated initializers
1973 Standard C90 requires the elements of an initializer to appear in a fixed
1974 order, the same as the order of the elements in the array or structure
1977 In ISO C99 you can give the elements in any order, specifying the array
1978 indices or structure field names they apply to, and GNU C allows this as
1979 an extension in C90 mode as well. This extension is not
1980 implemented in GNU C++.
1982 To specify an array index, write
1983 @samp{[@var{index}] =} before the element value. For example,
1986 int a[6] = @{ [4] = 29, [2] = 15 @};
1993 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1997 The index values must be constant expressions, even if the array being
1998 initialized is automatic.
2000 An alternative syntax for this that has been obsolete since GCC 2.5 but
2001 GCC still accepts is to write @samp{[@var{index}]} before the element
2002 value, with no @samp{=}.
2004 To initialize a range of elements to the same value, write
2005 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2006 extension. For example,
2009 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2013 If the value in it has side-effects, the side-effects happen only once,
2014 not for each initialized field by the range initializer.
2017 Note that the length of the array is the highest value specified
2020 In a structure initializer, specify the name of a field to initialize
2021 with @samp{.@var{fieldname} =} before the element value. For example,
2022 given the following structure,
2025 struct point @{ int x, y; @};
2029 the following initialization
2032 struct point p = @{ .y = yvalue, .x = xvalue @};
2039 struct point p = @{ xvalue, yvalue @};
2042 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2043 @samp{@var{fieldname}:}, as shown here:
2046 struct point p = @{ y: yvalue, x: xvalue @};
2049 Omitted field members are implicitly initialized the same as objects
2050 that have static storage duration.
2053 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2054 @dfn{designator}. You can also use a designator (or the obsolete colon
2055 syntax) when initializing a union, to specify which element of the union
2056 should be used. For example,
2059 union foo @{ int i; double d; @};
2061 union foo f = @{ .d = 4 @};
2065 converts 4 to a @code{double} to store it in the union using
2066 the second element. By contrast, casting 4 to type @code{union foo}
2067 stores it into the union as the integer @code{i}, since it is
2068 an integer. (@xref{Cast to Union}.)
2070 You can combine this technique of naming elements with ordinary C
2071 initialization of successive elements. Each initializer element that
2072 does not have a designator applies to the next consecutive element of the
2073 array or structure. For example,
2076 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2083 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2086 Labeling the elements of an array initializer is especially useful
2087 when the indices are characters or belong to an @code{enum} type.
2092 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2093 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2096 @cindex designator lists
2097 You can also write a series of @samp{.@var{fieldname}} and
2098 @samp{[@var{index}]} designators before an @samp{=} to specify a
2099 nested subobject to initialize; the list is taken relative to the
2100 subobject corresponding to the closest surrounding brace pair. For
2101 example, with the @samp{struct point} declaration above:
2104 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2108 If the same field is initialized multiple times, it has the value from
2109 the last initialization. If any such overridden initialization has
2110 side-effect, it is unspecified whether the side-effect happens or not.
2111 Currently, GCC discards them and issues a warning.
2114 @section Case Ranges
2116 @cindex ranges in case statements
2118 You can specify a range of consecutive values in a single @code{case} label,
2122 case @var{low} ... @var{high}:
2126 This has the same effect as the proper number of individual @code{case}
2127 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2129 This feature is especially useful for ranges of ASCII character codes:
2135 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2136 it may be parsed wrong when you use it with integer values. For example,
2151 @section Cast to a Union Type
2152 @cindex cast to a union
2153 @cindex union, casting to a
2155 A cast to union type is similar to other casts, except that the type
2156 specified is a union type. You can specify the type either with
2157 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2158 a constructor, not a cast, and hence does not yield an lvalue like
2159 normal casts. (@xref{Compound Literals}.)
2161 The types that may be cast to the union type are those of the members
2162 of the union. Thus, given the following union and variables:
2165 union foo @{ int i; double d; @};
2171 both @code{x} and @code{y} can be cast to type @code{union foo}.
2173 Using the cast as the right-hand side of an assignment to a variable of
2174 union type is equivalent to storing in a member of the union:
2179 u = (union foo) x @equiv{} u.i = x
2180 u = (union foo) y @equiv{} u.d = y
2183 You can also use the union cast as a function argument:
2186 void hack (union foo);
2188 hack ((union foo) x);
2191 @node Mixed Declarations
2192 @section Mixed Declarations and Code
2193 @cindex mixed declarations and code
2194 @cindex declarations, mixed with code
2195 @cindex code, mixed with declarations
2197 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2198 within compound statements. As an extension, GNU C also allows this in
2199 C90 mode. For example, you could do:
2208 Each identifier is visible from where it is declared until the end of
2209 the enclosing block.
2211 @node Function Attributes
2212 @section Declaring Attributes of Functions
2213 @cindex function attributes
2214 @cindex declaring attributes of functions
2215 @cindex @code{volatile} applied to function
2216 @cindex @code{const} applied to function
2218 In GNU C, you can use function attributes to declare certain things
2219 about functions called in your program which help the compiler
2220 optimize calls and check your code more carefully. For example, you
2221 can use attributes to declare that a function never returns
2222 (@code{noreturn}), returns a value depending only on its arguments
2223 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2225 You can also use attributes to control memory placement, code
2226 generation options or call/return conventions within the function
2227 being annotated. Many of these attributes are target-specific. For
2228 example, many targets support attributes for defining interrupt
2229 handler functions, which typically must follow special register usage
2230 and return conventions.
2232 Function attributes are introduced by the @code{__attribute__} keyword
2233 on a declaration, followed by an attribute specification inside double
2234 parentheses. You can specify multiple attributes in a declaration by
2235 separating them by commas within the double parentheses or by
2236 immediately following an attribute declaration with another attribute
2237 declaration. @xref{Attribute Syntax}, for the exact rules on
2238 attribute syntax and placement.
2240 GCC also supports attributes on
2241 variable declarations (@pxref{Variable Attributes}),
2242 labels (@pxref{Label Attributes}),
2243 enumerators (@pxref{Enumerator Attributes}),
2244 and types (@pxref{Type Attributes}).
2246 There is some overlap between the purposes of attributes and pragmas
2247 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2248 found convenient to use @code{__attribute__} to achieve a natural
2249 attachment of attributes to their corresponding declarations, whereas
2250 @code{#pragma} is of use for compatibility with other compilers
2251 or constructs that do not naturally form part of the grammar.
2253 In addition to the attributes documented here,
2254 GCC plugins may provide their own attributes.
2257 * Common Function Attributes::
2258 * AArch64 Function Attributes::
2259 * ARC Function Attributes::
2260 * ARM Function Attributes::
2261 * AVR Function Attributes::
2262 * Blackfin Function Attributes::
2263 * CR16 Function Attributes::
2264 * Epiphany Function Attributes::
2265 * H8/300 Function Attributes::
2266 * IA-64 Function Attributes::
2267 * M32C Function Attributes::
2268 * M32R/D Function Attributes::
2269 * m68k Function Attributes::
2270 * MCORE Function Attributes::
2271 * MeP Function Attributes::
2272 * MicroBlaze Function Attributes::
2273 * Microsoft Windows Function Attributes::
2274 * MIPS Function Attributes::
2275 * MSP430 Function Attributes::
2276 * NDS32 Function Attributes::
2277 * Nios II Function Attributes::
2278 * PowerPC Function Attributes::
2279 * RL78 Function Attributes::
2280 * RX Function Attributes::
2281 * S/390 Function Attributes::
2282 * SH Function Attributes::
2283 * SPU Function Attributes::
2284 * Symbian OS Function Attributes::
2285 * Visium Function Attributes::
2286 * x86 Function Attributes::
2287 * Xstormy16 Function Attributes::
2290 @node Common Function Attributes
2291 @subsection Common Function Attributes
2293 The following attributes are supported on most targets.
2296 @c Keep this table alphabetized by attribute name. Treat _ as space.
2298 @item alias ("@var{target}")
2299 @cindex @code{alias} function attribute
2300 The @code{alias} attribute causes the declaration to be emitted as an
2301 alias for another symbol, which must be specified. For instance,
2304 void __f () @{ /* @r{Do something.} */; @}
2305 void f () __attribute__ ((weak, alias ("__f")));
2309 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2310 mangled name for the target must be used. It is an error if @samp{__f}
2311 is not defined in the same translation unit.
2313 This attribute requires assembler and object file support,
2314 and may not be available on all targets.
2316 @item aligned (@var{alignment})
2317 @cindex @code{aligned} function attribute
2318 This attribute specifies a minimum alignment for the function,
2321 You cannot use this attribute to decrease the alignment of a function,
2322 only to increase it. However, when you explicitly specify a function
2323 alignment this overrides the effect of the
2324 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2327 Note that the effectiveness of @code{aligned} attributes may be
2328 limited by inherent limitations in your linker. On many systems, the
2329 linker is only able to arrange for functions to be aligned up to a
2330 certain maximum alignment. (For some linkers, the maximum supported
2331 alignment may be very very small.) See your linker documentation for
2332 further information.
2334 The @code{aligned} attribute can also be used for variables and fields
2335 (@pxref{Variable Attributes}.)
2338 @cindex @code{alloc_align} function attribute
2339 The @code{alloc_align} 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 one of the functions parameters. GCC uses this
2342 information to improve pointer alignment analysis.
2344 The function parameter denoting the allocated alignment is specified by
2345 one integer argument, whose number is the argument of the attribute.
2346 Argument numbering starts at one.
2351 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2355 declares that @code{my_memalign} returns memory with minimum alignment
2356 given by parameter 1.
2359 @cindex @code{alloc_size} function attribute
2360 The @code{alloc_size} attribute is used to tell the compiler that the
2361 function return value points to memory, where the size is given by
2362 one or two of the functions parameters. GCC uses this
2363 information to improve the correctness of @code{__builtin_object_size}.
2365 The function parameter(s) denoting the allocated size are specified by
2366 one or two integer arguments supplied to the attribute. The allocated size
2367 is either the value of the single function argument specified or the product
2368 of the two function arguments specified. Argument numbering starts at
2374 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2375 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2379 declares that @code{my_calloc} returns memory of the size given by
2380 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2381 of the size given by parameter 2.
2384 @cindex @code{always_inline} function attribute
2385 Generally, functions are not inlined unless optimization is specified.
2386 For functions declared inline, this attribute inlines the function
2387 independent of any restrictions that otherwise apply to inlining.
2388 Failure to inline such a function is diagnosed as an error.
2389 Note that if such a function is called indirectly the compiler may
2390 or may not inline it depending on optimization level and a failure
2391 to inline an indirect call may or may not be diagnosed.
2394 @cindex @code{artificial} function attribute
2395 This attribute is useful for small inline wrappers that if possible
2396 should appear during debugging as a unit. Depending on the debug
2397 info format it either means marking the function as artificial
2398 or using the caller location for all instructions within the inlined
2401 @item assume_aligned
2402 @cindex @code{assume_aligned} function attribute
2403 The @code{assume_aligned} attribute is used to tell the compiler that the
2404 function return value points to memory, where the returned pointer minimum
2405 alignment is given by the first argument.
2406 If the attribute has two arguments, the second argument is misalignment offset.
2411 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2412 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2416 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2417 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2420 @item bnd_instrument
2421 @cindex @code{bnd_instrument} function attribute
2422 The @code{bnd_instrument} attribute on functions is used to inform the
2423 compiler that the function should be instrumented when compiled
2424 with the @option{-fchkp-instrument-marked-only} option.
2427 @cindex @code{bnd_legacy} function attribute
2428 @cindex Pointer Bounds Checker attributes
2429 The @code{bnd_legacy} attribute on functions is used to inform the
2430 compiler that the function should not be instrumented when compiled
2431 with the @option{-fcheck-pointer-bounds} option.
2434 @cindex @code{cold} function attribute
2435 The @code{cold} attribute on functions is used to inform the compiler that
2436 the function is unlikely to be executed. The function is optimized for
2437 size rather than speed and on many targets it is placed into a special
2438 subsection of the text section so all cold functions appear close together,
2439 improving code locality of non-cold parts of program. The paths leading
2440 to calls of cold functions within code are marked as unlikely by the branch
2441 prediction mechanism. It is thus useful to mark functions used to handle
2442 unlikely conditions, such as @code{perror}, as cold to improve optimization
2443 of hot functions that do call marked functions in rare occasions.
2445 When profile feedback is available, via @option{-fprofile-use}, cold functions
2446 are automatically detected and this attribute is ignored.
2449 @cindex @code{const} function attribute
2450 @cindex functions that have no side effects
2451 Many functions do not examine any values except their arguments, and
2452 have no effects except the return value. Basically this is just slightly
2453 more strict class than the @code{pure} attribute below, since function is not
2454 allowed to read global memory.
2456 @cindex pointer arguments
2457 Note that a function that has pointer arguments and examines the data
2458 pointed to must @emph{not} be declared @code{const}. Likewise, a
2459 function that calls a non-@code{const} function usually must not be
2460 @code{const}. It does not make sense for a @code{const} function to
2465 @itemx constructor (@var{priority})
2466 @itemx destructor (@var{priority})
2467 @cindex @code{constructor} function attribute
2468 @cindex @code{destructor} function attribute
2469 The @code{constructor} attribute causes the function to be called
2470 automatically before execution enters @code{main ()}. Similarly, the
2471 @code{destructor} attribute causes the function to be called
2472 automatically after @code{main ()} completes or @code{exit ()} is
2473 called. Functions with these attributes are useful for
2474 initializing data that is used implicitly during the execution of
2477 You may provide an optional integer priority to control the order in
2478 which constructor and destructor functions are run. A constructor
2479 with a smaller priority number runs before a constructor with a larger
2480 priority number; the opposite relationship holds for destructors. So,
2481 if you have a constructor that allocates a resource and a destructor
2482 that deallocates the same resource, both functions typically have the
2483 same priority. The priorities for constructor and destructor
2484 functions are the same as those specified for namespace-scope C++
2485 objects (@pxref{C++ Attributes}).
2487 These attributes are not currently implemented for Objective-C@.
2490 @itemx deprecated (@var{msg})
2491 @cindex @code{deprecated} function attribute
2492 The @code{deprecated} attribute results in a warning if the function
2493 is used anywhere in the source file. This is useful when identifying
2494 functions that are expected to be removed in a future version of a
2495 program. The warning also includes the location of the declaration
2496 of the deprecated function, to enable users to easily find further
2497 information about why the function is deprecated, or what they should
2498 do instead. Note that the warnings only occurs for uses:
2501 int old_fn () __attribute__ ((deprecated));
2503 int (*fn_ptr)() = old_fn;
2507 results in a warning on line 3 but not line 2. The optional @var{msg}
2508 argument, which must be a string, is printed in the warning if
2511 The @code{deprecated} attribute can also be used for variables and
2512 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2514 @item error ("@var{message}")
2515 @itemx warning ("@var{message}")
2516 @cindex @code{error} function attribute
2517 @cindex @code{warning} function attribute
2518 If the @code{error} or @code{warning} attribute
2519 is used on a function declaration and a call to such a function
2520 is not eliminated through dead code elimination or other optimizations,
2521 an error or warning (respectively) that includes @var{message} is diagnosed.
2523 for compile-time checking, especially together with @code{__builtin_constant_p}
2524 and inline functions where checking the inline function arguments is not
2525 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2527 While it is possible to leave the function undefined and thus invoke
2528 a link failure (to define the function with
2529 a message in @code{.gnu.warning*} section),
2530 when using these attributes the problem is diagnosed
2531 earlier and with exact location of the call even in presence of inline
2532 functions or when not emitting debugging information.
2534 @item externally_visible
2535 @cindex @code{externally_visible} function attribute
2536 This attribute, attached to a global variable or function, nullifies
2537 the effect of the @option{-fwhole-program} command-line option, so the
2538 object remains visible outside the current compilation unit.
2540 If @option{-fwhole-program} is used together with @option{-flto} and
2541 @command{gold} is used as the linker plugin,
2542 @code{externally_visible} attributes are automatically added to functions
2543 (not variable yet due to a current @command{gold} issue)
2544 that are accessed outside of LTO objects according to resolution file
2545 produced by @command{gold}.
2546 For other linkers that cannot generate resolution file,
2547 explicit @code{externally_visible} attributes are still necessary.
2550 @cindex @code{flatten} function attribute
2551 Generally, inlining into a function is limited. For a function marked with
2552 this attribute, every call inside this function is inlined, if possible.
2553 Whether the function itself is considered for inlining depends on its size and
2554 the current inlining parameters.
2556 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2557 @cindex @code{format} function attribute
2558 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2560 The @code{format} attribute specifies that a function takes @code{printf},
2561 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2562 should be type-checked against a format string. For example, the
2567 my_printf (void *my_object, const char *my_format, ...)
2568 __attribute__ ((format (printf, 2, 3)));
2572 causes the compiler to check the arguments in calls to @code{my_printf}
2573 for consistency with the @code{printf} style format string argument
2576 The parameter @var{archetype} determines how the format string is
2577 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2578 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2579 @code{strfmon}. (You can also use @code{__printf__},
2580 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2581 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2582 @code{ms_strftime} are also present.
2583 @var{archetype} values such as @code{printf} refer to the formats accepted
2584 by the system's C runtime library,
2585 while values prefixed with @samp{gnu_} always refer
2586 to the formats accepted by the GNU C Library. On Microsoft Windows
2587 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2588 @file{msvcrt.dll} library.
2589 The parameter @var{string-index}
2590 specifies which argument is the format string argument (starting
2591 from 1), while @var{first-to-check} is the number of the first
2592 argument to check against the format string. For functions
2593 where the arguments are not available to be checked (such as
2594 @code{vprintf}), specify the third parameter as zero. In this case the
2595 compiler only checks the format string for consistency. For
2596 @code{strftime} formats, the third parameter is required to be zero.
2597 Since non-static C++ methods have an implicit @code{this} argument, the
2598 arguments of such methods should be counted from two, not one, when
2599 giving values for @var{string-index} and @var{first-to-check}.
2601 In the example above, the format string (@code{my_format}) is the second
2602 argument of the function @code{my_print}, and the arguments to check
2603 start with the third argument, so the correct parameters for the format
2604 attribute are 2 and 3.
2606 @opindex ffreestanding
2607 @opindex fno-builtin
2608 The @code{format} attribute allows you to identify your own functions
2609 that take format strings as arguments, so that GCC can check the
2610 calls to these functions for errors. The compiler always (unless
2611 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2612 for the standard library functions @code{printf}, @code{fprintf},
2613 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2614 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2615 warnings are requested (using @option{-Wformat}), so there is no need to
2616 modify the header file @file{stdio.h}. In C99 mode, the functions
2617 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2618 @code{vsscanf} are also checked. Except in strictly conforming C
2619 standard modes, the X/Open function @code{strfmon} is also checked as
2620 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2621 @xref{C Dialect Options,,Options Controlling C Dialect}.
2623 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2624 recognized in the same context. Declarations including these format attributes
2625 are parsed for correct syntax, however the result of checking of such format
2626 strings is not yet defined, and is not carried out by this version of the
2629 The target may also provide additional types of format checks.
2630 @xref{Target Format Checks,,Format Checks Specific to Particular
2633 @item format_arg (@var{string-index})
2634 @cindex @code{format_arg} function attribute
2635 @opindex Wformat-nonliteral
2636 The @code{format_arg} attribute specifies that a function takes a format
2637 string for a @code{printf}, @code{scanf}, @code{strftime} or
2638 @code{strfmon} style function and modifies it (for example, to translate
2639 it into another language), so the result can be passed to a
2640 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2641 function (with the remaining arguments to the format function the same
2642 as they would have been for the unmodified string). For example, the
2647 my_dgettext (char *my_domain, const char *my_format)
2648 __attribute__ ((format_arg (2)));
2652 causes the compiler to check the arguments in calls to a @code{printf},
2653 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2654 format string argument is a call to the @code{my_dgettext} function, for
2655 consistency with the format string argument @code{my_format}. If the
2656 @code{format_arg} attribute had not been specified, all the compiler
2657 could tell in such calls to format functions would be that the format
2658 string argument is not constant; this would generate a warning when
2659 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2660 without the attribute.
2662 The parameter @var{string-index} specifies which argument is the format
2663 string argument (starting from one). Since non-static C++ methods have
2664 an implicit @code{this} argument, the arguments of such methods should
2665 be counted from two.
2667 The @code{format_arg} attribute allows you to identify your own
2668 functions that modify format strings, so that GCC can check the
2669 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2670 type function whose operands are a call to one of your own function.
2671 The compiler always treats @code{gettext}, @code{dgettext}, and
2672 @code{dcgettext} in this manner except when strict ISO C support is
2673 requested by @option{-ansi} or an appropriate @option{-std} option, or
2674 @option{-ffreestanding} or @option{-fno-builtin}
2675 is used. @xref{C Dialect Options,,Options
2676 Controlling C Dialect}.
2678 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2679 @code{NSString} reference for compatibility with the @code{format} attribute
2682 The target may also allow additional types in @code{format-arg} attributes.
2683 @xref{Target Format Checks,,Format Checks Specific to Particular
2687 @cindex @code{gnu_inline} function attribute
2688 This attribute should be used with a function that is also declared
2689 with the @code{inline} keyword. It directs GCC to treat the function
2690 as if it were defined in gnu90 mode even when compiling in C99 or
2693 If the function is declared @code{extern}, then this definition of the
2694 function is used only for inlining. In no case is the function
2695 compiled as a standalone function, not even if you take its address
2696 explicitly. Such an address becomes an external reference, as if you
2697 had only declared the function, and had not defined it. This has
2698 almost the effect of a macro. The way to use this is to put a
2699 function definition in a header file with this attribute, and put
2700 another copy of the function, without @code{extern}, in a library
2701 file. The definition in the header file causes most calls to the
2702 function to be inlined. If any uses of the function remain, they
2703 refer to the single copy in the library. Note that the two
2704 definitions of the functions need not be precisely the same, although
2705 if they do not have the same effect your program may behave oddly.
2707 In C, if the function is neither @code{extern} nor @code{static}, then
2708 the function is compiled as a standalone function, as well as being
2709 inlined where possible.
2711 This is how GCC traditionally handled functions declared
2712 @code{inline}. Since ISO C99 specifies a different semantics for
2713 @code{inline}, this function attribute is provided as a transition
2714 measure and as a useful feature in its own right. This attribute is
2715 available in GCC 4.1.3 and later. It is available if either of the
2716 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2717 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2718 Function is As Fast As a Macro}.
2720 In C++, this attribute does not depend on @code{extern} in any way,
2721 but it still requires the @code{inline} keyword to enable its special
2725 @cindex @code{hot} function attribute
2726 The @code{hot} attribute on a function is used to inform the compiler that
2727 the function is a hot spot of the compiled program. The function is
2728 optimized more aggressively and on many targets it is placed into a special
2729 subsection of the text section so all hot functions appear close together,
2732 When profile feedback is available, via @option{-fprofile-use}, hot functions
2733 are automatically detected and this attribute is ignored.
2735 @item ifunc ("@var{resolver}")
2736 @cindex @code{ifunc} function attribute
2737 @cindex indirect functions
2738 @cindex functions that are dynamically resolved
2739 The @code{ifunc} attribute is used to mark a function as an indirect
2740 function using the STT_GNU_IFUNC symbol type extension to the ELF
2741 standard. This allows the resolution of the symbol value to be
2742 determined dynamically at load time, and an optimized version of the
2743 routine can be selected for the particular processor or other system
2744 characteristics determined then. To use this attribute, first define
2745 the implementation functions available, and a resolver function that
2746 returns a pointer to the selected implementation function. The
2747 implementation functions' declarations must match the API of the
2748 function being implemented, the resolver's declaration is be a
2749 function returning pointer to void function returning void:
2752 void *my_memcpy (void *dst, const void *src, size_t len)
2757 static void (*resolve_memcpy (void)) (void)
2759 return my_memcpy; // we'll just always select this routine
2764 The exported header file declaring the function the user calls would
2768 extern void *memcpy (void *, const void *, size_t);
2772 allowing the user to call this as a regular function, unaware of the
2773 implementation. Finally, the indirect function needs to be defined in
2774 the same translation unit as the resolver function:
2777 void *memcpy (void *, const void *, size_t)
2778 __attribute__ ((ifunc ("resolve_memcpy")));
2781 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2782 and GNU C Library version 2.11.1 are required to use this feature.
2785 @itemx interrupt_handler
2786 Many GCC back ends support attributes to indicate that a function is
2787 an interrupt handler, which tells the compiler to generate function
2788 entry and exit sequences that differ from those from regular
2789 functions. The exact syntax and behavior are target-specific;
2790 refer to the following subsections for details.
2793 @cindex @code{leaf} function attribute
2794 Calls to external functions with this attribute must return to the current
2795 compilation unit only by return or by exception handling. In particular, leaf
2796 functions are not allowed to call callback function passed to it from the current
2797 compilation unit or directly call functions exported by the unit or longjmp
2798 into the unit. Leaf function might still call functions from other compilation
2799 units and thus they are not necessarily leaf in the sense that they contain no
2800 function calls at all.
2802 The attribute is intended for library functions to improve dataflow analysis.
2803 The compiler takes the hint that any data not escaping the current compilation unit can
2804 not be used or modified by the leaf function. For example, the @code{sin} function
2805 is a leaf function, but @code{qsort} is not.
2807 Note that leaf functions might invoke signals and signal handlers might be
2808 defined in the current compilation unit and use static variables. The only
2809 compliant way to write such a signal handler is to declare such variables
2812 The attribute has no effect on functions defined within the current compilation
2813 unit. This is to allow easy merging of multiple compilation units into one,
2814 for example, by using the link-time optimization. For this reason the
2815 attribute is not allowed on types to annotate indirect calls.
2819 @cindex @code{malloc} function attribute
2820 @cindex functions that behave like malloc
2821 This tells the compiler that a function is @code{malloc}-like, i.e.,
2822 that the pointer @var{P} returned by the function cannot alias any
2823 other pointer valid when the function returns, and moreover no
2824 pointers to valid objects occur in any storage addressed by @var{P}.
2826 Using this attribute can improve optimization. Functions like
2827 @code{malloc} and @code{calloc} have this property because they return
2828 a pointer to uninitialized or zeroed-out storage. However, functions
2829 like @code{realloc} do not have this property, as they can return a
2830 pointer to storage containing pointers.
2833 @cindex @code{no_icf} function attribute
2834 This function attribute prevents a functions from being merged with another
2835 semantically equivalent function.
2837 @item no_instrument_function
2838 @cindex @code{no_instrument_function} function attribute
2839 @opindex finstrument-functions
2840 If @option{-finstrument-functions} is given, profiling function calls are
2841 generated at entry and exit of most user-compiled functions.
2842 Functions with this attribute are not so instrumented.
2845 @cindex @code{no_reorder} function attribute
2846 Do not reorder functions or variables marked @code{no_reorder}
2847 against each other or top level assembler statements the executable.
2848 The actual order in the program will depend on the linker command
2849 line. Static variables marked like this are also not removed.
2850 This has a similar effect
2851 as the @option{-fno-toplevel-reorder} option, but only applies to the
2854 @item no_sanitize_address
2855 @itemx no_address_safety_analysis
2856 @cindex @code{no_sanitize_address} function attribute
2857 The @code{no_sanitize_address} attribute on functions is used
2858 to inform the compiler that it should not instrument memory accesses
2859 in the function when compiling with the @option{-fsanitize=address} option.
2860 The @code{no_address_safety_analysis} is a deprecated alias of the
2861 @code{no_sanitize_address} attribute, new code should use
2862 @code{no_sanitize_address}.
2864 @item no_sanitize_thread
2865 @cindex @code{no_sanitize_thread} function attribute
2866 The @code{no_sanitize_thread} attribute on functions is used
2867 to inform the compiler that it should not instrument memory accesses
2868 in the function when compiling with the @option{-fsanitize=thread} option.
2870 @item no_sanitize_undefined
2871 @cindex @code{no_sanitize_undefined} function attribute
2872 The @code{no_sanitize_undefined} attribute on functions is used
2873 to inform the compiler that it should not check for undefined behavior
2874 in the function when compiling with the @option{-fsanitize=undefined} option.
2876 @item no_split_stack
2877 @cindex @code{no_split_stack} function attribute
2878 @opindex fsplit-stack
2879 If @option{-fsplit-stack} is given, functions have a small
2880 prologue which decides whether to split the stack. Functions with the
2881 @code{no_split_stack} attribute do not have that prologue, and thus
2882 may run with only a small amount of stack space available.
2885 @cindex @code{noclone} function attribute
2886 This function attribute prevents a function from being considered for
2887 cloning---a mechanism that produces specialized copies of functions
2888 and which is (currently) performed by interprocedural constant
2892 @cindex @code{noinline} function attribute
2893 This function attribute prevents a function from being considered for
2895 @c Don't enumerate the optimizations by name here; we try to be
2896 @c future-compatible with this mechanism.
2897 If the function does not have side-effects, there are optimizations
2898 other than inlining that cause function calls to be optimized away,
2899 although the function call is live. To keep such calls from being
2906 (@pxref{Extended Asm}) in the called function, to serve as a special
2909 @item nonnull (@var{arg-index}, @dots{})
2910 @cindex @code{nonnull} function attribute
2911 @cindex functions with non-null pointer arguments
2912 The @code{nonnull} attribute specifies that some function parameters should
2913 be non-null pointers. For instance, the declaration:
2917 my_memcpy (void *dest, const void *src, size_t len)
2918 __attribute__((nonnull (1, 2)));
2922 causes the compiler to check that, in calls to @code{my_memcpy},
2923 arguments @var{dest} and @var{src} are non-null. If the compiler
2924 determines that a null pointer is passed in an argument slot marked
2925 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2926 is issued. The compiler may also choose to make optimizations based
2927 on the knowledge that certain function arguments will never be null.
2929 If no argument index list is given to the @code{nonnull} attribute,
2930 all pointer arguments are marked as non-null. To illustrate, the
2931 following declaration is equivalent to the previous example:
2935 my_memcpy (void *dest, const void *src, size_t len)
2936 __attribute__((nonnull));
2940 @cindex @code{noreturn} function attribute
2941 @cindex functions that never return
2942 A few standard library functions, such as @code{abort} and @code{exit},
2943 cannot return. GCC knows this automatically. Some programs define
2944 their own functions that never return. You can declare them
2945 @code{noreturn} to tell the compiler this fact. For example,
2949 void fatal () __attribute__ ((noreturn));
2952 fatal (/* @r{@dots{}} */)
2954 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2960 The @code{noreturn} keyword tells the compiler to assume that
2961 @code{fatal} cannot return. It can then optimize without regard to what
2962 would happen if @code{fatal} ever did return. This makes slightly
2963 better code. More importantly, it helps avoid spurious warnings of
2964 uninitialized variables.
2966 The @code{noreturn} keyword does not affect the exceptional path when that
2967 applies: a @code{noreturn}-marked function may still return to the caller
2968 by throwing an exception or calling @code{longjmp}.
2970 Do not assume that registers saved by the calling function are
2971 restored before calling the @code{noreturn} function.
2973 It does not make sense for a @code{noreturn} function to have a return
2974 type other than @code{void}.
2977 @cindex @code{nothrow} function attribute
2978 The @code{nothrow} attribute is used to inform the compiler that a
2979 function cannot throw an exception. For example, most functions in
2980 the standard C library can be guaranteed not to throw an exception
2981 with the notable exceptions of @code{qsort} and @code{bsearch} that
2982 take function pointer arguments.
2985 @cindex @code{noplt} function attribute
2986 The @code{noplt} attribute is the counterpart to option @option{-fno-plt} and
2987 does not use PLT for calls to functions marked with this attribute in position
2992 /* Externally defined function foo. */
2993 int foo () __attribute__ ((noplt));
2996 main (/* @r{@dots{}} */)
3005 The @code{noplt} attribute on function foo tells the compiler to assume that
3006 the function foo is externally defined and the call to foo must avoid the PLT
3007 in position independent code.
3009 Additionally, a few targets also convert calls to those functions that are
3010 marked to not use the PLT to use the GOT instead for non-position independent
3014 @cindex @code{optimize} function attribute
3015 The @code{optimize} attribute is used to specify that a function is to
3016 be compiled with different optimization options than specified on the
3017 command line. Arguments can either be numbers or strings. Numbers
3018 are assumed to be an optimization level. Strings that begin with
3019 @code{O} are assumed to be an optimization option, while other options
3020 are assumed to be used with a @code{-f} prefix. You can also use the
3021 @samp{#pragma GCC optimize} pragma to set the optimization options
3022 that affect more than one function.
3023 @xref{Function Specific Option Pragmas}, for details about the
3024 @samp{#pragma GCC optimize} pragma.
3026 This can be used for instance to have frequently-executed functions
3027 compiled with more aggressive optimization options that produce faster
3028 and larger code, while other functions can be compiled with less
3032 @cindex @code{pure} function attribute
3033 @cindex functions that have no side effects
3034 Many functions have no effects except the return value and their
3035 return value depends only on the parameters and/or global variables.
3036 Such a function can be subject
3037 to common subexpression elimination and loop optimization just as an
3038 arithmetic operator would be. These functions should be declared
3039 with the attribute @code{pure}. For example,
3042 int square (int) __attribute__ ((pure));
3046 says that the hypothetical function @code{square} is safe to call
3047 fewer times than the program says.
3049 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3050 Interesting non-pure functions are functions with infinite loops or those
3051 depending on volatile memory or other system resource, that may change between
3052 two consecutive calls (such as @code{feof} in a multithreading environment).
3054 @item returns_nonnull
3055 @cindex @code{returns_nonnull} function attribute
3056 The @code{returns_nonnull} attribute specifies that the function
3057 return value should be a non-null pointer. For instance, the declaration:
3061 mymalloc (size_t len) __attribute__((returns_nonnull));
3065 lets the compiler optimize callers based on the knowledge
3066 that the return value will never be null.
3069 @cindex @code{returns_twice} function attribute
3070 @cindex functions that return more than once
3071 The @code{returns_twice} attribute tells the compiler that a function may
3072 return more than one time. The compiler ensures that all registers
3073 are dead before calling such a function and emits a warning about
3074 the variables that may be clobbered after the second return from the
3075 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3076 The @code{longjmp}-like counterpart of such function, if any, might need
3077 to be marked with the @code{noreturn} attribute.
3079 @item section ("@var{section-name}")
3080 @cindex @code{section} function attribute
3081 @cindex functions in arbitrary sections
3082 Normally, the compiler places the code it generates in the @code{text} section.
3083 Sometimes, however, you need additional sections, or you need certain
3084 particular functions to appear in special sections. The @code{section}
3085 attribute specifies that a function lives in a particular section.
3086 For example, the declaration:
3089 extern void foobar (void) __attribute__ ((section ("bar")));
3093 puts the function @code{foobar} in the @code{bar} section.
3095 Some file formats do not support arbitrary sections so the @code{section}
3096 attribute is not available on all platforms.
3097 If you need to map the entire contents of a module to a particular
3098 section, consider using the facilities of the linker instead.
3101 @cindex @code{sentinel} function attribute
3102 This function attribute ensures that a parameter in a function call is
3103 an explicit @code{NULL}. The attribute is only valid on variadic
3104 functions. By default, the sentinel is located at position zero, the
3105 last parameter of the function call. If an optional integer position
3106 argument P is supplied to the attribute, the sentinel must be located at
3107 position P counting backwards from the end of the argument list.
3110 __attribute__ ((sentinel))
3112 __attribute__ ((sentinel(0)))
3115 The attribute is automatically set with a position of 0 for the built-in
3116 functions @code{execl} and @code{execlp}. The built-in function
3117 @code{execle} has the attribute set with a position of 1.
3119 A valid @code{NULL} in this context is defined as zero with any pointer
3120 type. If your system defines the @code{NULL} macro with an integer type
3121 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3122 with a copy that redefines NULL appropriately.
3124 The warnings for missing or incorrect sentinels are enabled with
3128 @cindex @code{stack_protect} function attribute
3129 This function attribute make a stack protection of the function if
3130 flags @option{fstack-protector} or @option{fstack-protector-strong}
3131 or @option{fstack-protector-explicit} are set.
3133 @item target_clones (@var{options})
3134 @cindex @code{target_clones} function attribute
3135 The @code{target_clones} attribute is used to specify that a function is to
3136 be cloned into multiple versions compiled with different target options
3137 than specified on the command line. The supported options and restrictions
3138 are the same as for @code{target} attribute.
3140 For instance on an x86, you could compile a function with
3141 @code{target_clones("sse4.1,avx")}. It will create 2 function clones,
3142 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3143 At the function call it will create resolver @code{ifunc}, that will
3144 dynamically call a clone suitable for current architecture.
3147 @cindex @code{simd} function attribute.
3148 This attribute enables creation of one or more function versions that
3149 can process multiple arguments using SIMD instructions from a
3150 single invocation. Specifying this attribute allows compiler to
3151 assume that such versions are available at link time (provided
3152 in the same or another translation unit). Generated versions are
3153 target dependent and described in corresponding Vector ABI document. For
3154 x86_64 target this document can be found
3155 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3156 The attribute should not be used together with Cilk Plus @code{vector}
3157 attribute on the same function.
3158 If the attribute is specified and @code{#pragma omp declare simd}
3159 present on a declaration and @code{-fopenmp} or @code{-fopenmp-simd}
3160 switch is specified, then the attribute is ignored.
3162 @item target (@var{options})
3163 @cindex @code{target} function attribute
3164 Multiple target back ends implement the @code{target} attribute
3165 to specify that a function is to
3166 be compiled with different target options than specified on the
3167 command line. This can be used for instance to have functions
3168 compiled with a different ISA (instruction set architecture) than the
3169 default. You can also use the @samp{#pragma GCC target} pragma to set
3170 more than one function to be compiled with specific target options.
3171 @xref{Function Specific Option Pragmas}, for details about the
3172 @samp{#pragma GCC target} pragma.
3174 For instance, on an x86, you could declare one function with the
3175 @code{target("sse4.1,arch=core2")} attribute and another with
3176 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3177 compiling the first function with @option{-msse4.1} and
3178 @option{-march=core2} options, and the second function with
3179 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3180 to make sure that a function is only invoked on a machine that
3181 supports the particular ISA it is compiled for (for example by using
3182 @code{cpuid} on x86 to determine what feature bits and architecture
3186 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3187 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3190 You can either use multiple
3191 strings separated by commas to specify multiple options,
3192 or separate the options with a comma (@samp{,}) within a single string.
3194 The options supported are specific to each target; refer to @ref{x86
3195 Function Attributes}, @ref{PowerPC Function Attributes},
3196 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3200 @cindex @code{unused} function attribute
3201 This attribute, attached to a function, means that the function is meant
3202 to be possibly unused. GCC does not produce a warning for this
3206 @cindex @code{used} function attribute
3207 This attribute, attached to a function, means that code must be emitted
3208 for the function even if it appears that the function is not referenced.
3209 This is useful, for example, when the function is referenced only in
3212 When applied to a member function of a C++ class template, the
3213 attribute also means that the function is instantiated if the
3214 class itself is instantiated.
3216 @item visibility ("@var{visibility_type}")
3217 @cindex @code{visibility} function attribute
3218 This attribute affects the linkage of the declaration to which it is attached.
3219 There are four supported @var{visibility_type} values: default,
3220 hidden, protected or internal visibility.
3223 void __attribute__ ((visibility ("protected")))
3224 f () @{ /* @r{Do something.} */; @}
3225 int i __attribute__ ((visibility ("hidden")));
3228 The possible values of @var{visibility_type} correspond to the
3229 visibility settings in the ELF gABI.
3232 @c keep this list of visibilities in alphabetical order.
3235 Default visibility is the normal case for the object file format.
3236 This value is available for the visibility attribute to override other
3237 options that may change the assumed visibility of entities.
3239 On ELF, default visibility means that the declaration is visible to other
3240 modules and, in shared libraries, means that the declared entity may be
3243 On Darwin, default visibility means that the declaration is visible to
3246 Default visibility corresponds to ``external linkage'' in the language.
3249 Hidden visibility indicates that the entity declared has a new
3250 form of linkage, which we call ``hidden linkage''. Two
3251 declarations of an object with hidden linkage refer to the same object
3252 if they are in the same shared object.
3255 Internal visibility is like hidden visibility, but with additional
3256 processor specific semantics. Unless otherwise specified by the
3257 psABI, GCC defines internal visibility to mean that a function is
3258 @emph{never} called from another module. Compare this with hidden
3259 functions which, while they cannot be referenced directly by other
3260 modules, can be referenced indirectly via function pointers. By
3261 indicating that a function cannot be called from outside the module,
3262 GCC may for instance omit the load of a PIC register since it is known
3263 that the calling function loaded the correct value.
3266 Protected visibility is like default visibility except that it
3267 indicates that references within the defining module bind to the
3268 definition in that module. That is, the declared entity cannot be
3269 overridden by another module.
3273 All visibilities are supported on many, but not all, ELF targets
3274 (supported when the assembler supports the @samp{.visibility}
3275 pseudo-op). Default visibility is supported everywhere. Hidden
3276 visibility is supported on Darwin targets.
3278 The visibility attribute should be applied only to declarations that
3279 would otherwise have external linkage. The attribute should be applied
3280 consistently, so that the same entity should not be declared with
3281 different settings of the attribute.
3283 In C++, the visibility attribute applies to types as well as functions
3284 and objects, because in C++ types have linkage. A class must not have
3285 greater visibility than its non-static data member types and bases,
3286 and class members default to the visibility of their class. Also, a
3287 declaration without explicit visibility is limited to the visibility
3290 In C++, you can mark member functions and static member variables of a
3291 class with the visibility attribute. This is useful if you know a
3292 particular method or static member variable should only be used from
3293 one shared object; then you can mark it hidden while the rest of the
3294 class has default visibility. Care must be taken to avoid breaking
3295 the One Definition Rule; for example, it is usually not useful to mark
3296 an inline method as hidden without marking the whole class as hidden.
3298 A C++ namespace declaration can also have the visibility attribute.
3301 namespace nspace1 __attribute__ ((visibility ("protected")))
3302 @{ /* @r{Do something.} */; @}
3305 This attribute applies only to the particular namespace body, not to
3306 other definitions of the same namespace; it is equivalent to using
3307 @samp{#pragma GCC visibility} before and after the namespace
3308 definition (@pxref{Visibility Pragmas}).
3310 In C++, if a template argument has limited visibility, this
3311 restriction is implicitly propagated to the template instantiation.
3312 Otherwise, template instantiations and specializations default to the
3313 visibility of their template.
3315 If both the template and enclosing class have explicit visibility, the
3316 visibility from the template is used.
3318 @item warn_unused_result
3319 @cindex @code{warn_unused_result} function attribute
3320 The @code{warn_unused_result} attribute causes a warning to be emitted
3321 if a caller of the function with this attribute does not use its
3322 return value. This is useful for functions where not checking
3323 the result is either a security problem or always a bug, such as
3327 int fn () __attribute__ ((warn_unused_result));
3330 if (fn () < 0) return -1;
3337 results in warning on line 5.
3340 @cindex @code{weak} function attribute
3341 The @code{weak} attribute causes the declaration to be emitted as a weak
3342 symbol rather than a global. This is primarily useful in defining
3343 library functions that can be overridden in user code, though it can
3344 also be used with non-function declarations. Weak symbols are supported
3345 for ELF targets, and also for a.out targets when using the GNU assembler
3349 @itemx weakref ("@var{target}")
3350 @cindex @code{weakref} function attribute
3351 The @code{weakref} attribute marks a declaration as a weak reference.
3352 Without arguments, it should be accompanied by an @code{alias} attribute
3353 naming the target symbol. Optionally, the @var{target} may be given as
3354 an argument to @code{weakref} itself. In either case, @code{weakref}
3355 implicitly marks the declaration as @code{weak}. Without a
3356 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3357 @code{weakref} is equivalent to @code{weak}.
3360 static int x() __attribute__ ((weakref ("y")));
3361 /* is equivalent to... */
3362 static int x() __attribute__ ((weak, weakref, alias ("y")));
3364 static int x() __attribute__ ((weakref));
3365 static int x() __attribute__ ((alias ("y")));
3368 A weak reference is an alias that does not by itself require a
3369 definition to be given for the target symbol. If the target symbol is
3370 only referenced through weak references, then it becomes a @code{weak}
3371 undefined symbol. If it is directly referenced, however, then such
3372 strong references prevail, and a definition is required for the
3373 symbol, not necessarily in the same translation unit.
3375 The effect is equivalent to moving all references to the alias to a
3376 separate translation unit, renaming the alias to the aliased symbol,
3377 declaring it as weak, compiling the two separate translation units and
3378 performing a reloadable link on them.
3380 At present, a declaration to which @code{weakref} is attached can
3381 only be @code{static}.
3386 @cindex lower memory region on the MSP430
3387 @cindex upper memory region on the MSP430
3388 @cindex either memory region on the MSP430
3389 On the MSP430 target these attributes can be used to specify whether
3390 the function or variable should be placed into low memory, high
3391 memory, or the placement should be left to the linker to decide. The
3392 attributes are only significant if compiling for the MSP430X
3395 The attributes work in conjunction with a linker script that has been
3396 augmented to specify where to place sections with a @code{.lower} and
3397 a @code{.upper} prefix. So for example as well as placing the
3398 @code{.data} section the script would also specify the placement of a
3399 @code{.lower.data} and a @code{.upper.data} section. The intention
3400 being that @code{lower} sections are placed into a small but easier to
3401 access memory region and the upper sections are placed into a larger, but
3402 slower to access region.
3404 The @code{either} attribute is special. It tells the linker to place
3405 the object into the corresponding @code{lower} section if there is
3406 room for it. If there is insufficient room then the object is placed
3407 into the corresponding @code{upper} section instead. Note - the
3408 placement algorithm is not very sophisticated. It will not attempt to
3409 find an optimal packing of the @code{lower} sections. It just makes
3410 one pass over the objects and does the best that it can. Using the
3411 @option{-ffunction-sections} and @option{-fdata-sections} command line
3412 options can help the packing however, since they produce smaller,
3413 easier to pack regions.
3416 On the MSP430 a function can be given the @code{reentant} attribute.
3417 This makes the function disable interrupts upon entry and enable
3418 interrupts upon exit. Reentrant functions cannot be @code{naked}.
3421 On the MSP430 a function can be given the @code{critical} attribute.
3422 This makes the function disable interrupts upon entry and restore the
3423 previous interrupt enabled/disabled state upon exit. A function
3424 cannot have both the @code{reentrant} and @code{critical} attributes.
3425 Critical functions cannot be @code{naked}.
3428 On the MSP430 a function can be given the @code{wakeup} attribute.
3429 Such a function must also have the @code{interrupt} attribute. When a
3430 function with the @code{wakeup} attribute exists the processor will be
3431 woken up from any low-power state in which it may be residing.
3435 @c This is the end of the target-independent attribute table
3437 @node AArch64 Function Attributes
3438 @subsection AArch64 Function Attributes
3440 The following target-specific function attributes are available for the
3441 AArch64 target. For the most part, these options mirror the behavior of
3442 similar command-line options (@pxref{AArch64 Options}), but on a
3446 @item general-regs-only
3447 @cindex @code{general-regs-only} function attribute, AArch64
3448 Indicates that no floating-point or Advanced SIMD registers should be
3449 used when generating code for this function. If the function explicitly
3450 uses floating-point code, then the compiler gives an error. This is
3451 the same behavior as that of the command-line option
3452 @option{-mgeneral-regs-only}.
3454 @item fix-cortex-a53-835769
3455 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3456 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3457 applied to this function. To explicitly disable the workaround for this
3458 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3459 This corresponds to the behavior of the command line options
3460 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3463 @cindex @code{cmodel=} function attribute, AArch64
3464 Indicates that code should be generated for a particular code model for
3465 this function. The behavior and permissible arguments are the same as
3466 for the command line option @option{-mcmodel=}.
3469 @cindex @code{strict-align} function attribute, AArch64
3470 Indicates that the compiler should not assume that unaligned memory references
3471 are handled by the system. The behavior is the same as for the command-line
3472 option @option{-mstrict-align}.
3474 @item omit-leaf-frame-pointer
3475 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3476 Indicates that the frame pointer should be omitted for a leaf function call.
3477 To keep the frame pointer, the inverse attribute
3478 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3479 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3480 and @option{-mno-omit-leaf-frame-pointer}.
3483 @cindex @code{tls-dialect=} function attribute, AArch64
3484 Specifies the TLS dialect to use for this function. The behavior and
3485 permissible arguments are the same as for the command-line option
3486 @option{-mtls-dialect=}.
3489 @cindex @code{arch=} function attribute, AArch64
3490 Specifies the architecture version and architectural extensions to use
3491 for this function. The behavior and permissible arguments are the same as
3492 for the @option{-march=} command-line option.
3495 @cindex @code{tune=} function attribute, AArch64
3496 Specifies the core for which to tune the performance of this function.
3497 The behavior and permissible arguments are the same as for the @option{-mtune=}
3498 command-line option.
3501 @cindex @code{cpu=} function attribute, AArch64
3502 Specifies the core for which to tune the performance of this function and also
3503 whose architectural features to use. The behavior and valid arguments are the
3504 same as for the @option{-mcpu=} command-line option.
3508 The above target attributes can be specified as follows:
3511 __attribute__((target("@var{attr-string}")))
3519 where @code{@var{attr-string}} is one of the attribute strings specified above.
3521 Additionally, the architectural extension string may be specified on its
3522 own. This can be used to turn on and off particular architectural extensions
3523 without having to specify a particular architecture version or core. Example:
3526 __attribute__((target("+crc+nocrypto")))
3534 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3535 extension and disables the @code{crypto} extension for the function @code{foo}
3536 without modifying an existing @option{-march=} or @option{-mcpu} option.
3538 Multiple target function attributes can be specified by separating them with
3539 a comma. For example:
3541 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3549 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3550 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3552 @subsubsection Inlining rules
3553 Specifying target attributes on individual functions or performing link-time
3554 optimization across translation units compiled with different target options
3555 can affect function inlining rules:
3557 In particular, a caller function can inline a callee function only if the
3558 architectural features available to the callee are a subset of the features
3559 available to the caller.
3560 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3561 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3562 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3563 because the all the architectural features that function @code{bar} requires
3564 are available to function @code{foo}. Conversely, function @code{bar} cannot
3565 inline function @code{foo}.
3567 Additionally inlining a function compiled with @option{-mstrict-align} into a
3568 function compiled without @code{-mstrict-align} is not allowed.
3569 However, inlining a function compiled without @option{-mstrict-align} into a
3570 function compiled with @option{-mstrict-align} is allowed.
3572 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3573 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3574 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3575 architectural feature rules specified above.
3577 @node ARC Function Attributes
3578 @subsection ARC Function Attributes
3580 These function attributes are supported by the ARC back end:
3584 @cindex @code{interrupt} function attribute, ARC
3585 Use this attribute to indicate
3586 that the specified function is an interrupt handler. The compiler generates
3587 function entry and exit sequences suitable for use in an interrupt handler
3588 when this attribute is present.
3590 On the ARC, you must specify the kind of interrupt to be handled
3591 in a parameter to the interrupt attribute like this:
3594 void f () __attribute__ ((interrupt ("ilink1")));
3597 Permissible values for this parameter are: @w{@code{ilink1}} and
3603 @cindex @code{long_call} function attribute, ARC
3604 @cindex @code{medium_call} function attribute, ARC
3605 @cindex @code{short_call} function attribute, ARC
3606 @cindex indirect calls, ARC
3607 These attributes specify how a particular function is called.
3608 These attributes override the
3609 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3610 command-line switches and @code{#pragma long_calls} settings.
3612 For ARC, a function marked with the @code{long_call} attribute is
3613 always called using register-indirect jump-and-link instructions,
3614 thereby enabling the called function to be placed anywhere within the
3615 32-bit address space. A function marked with the @code{medium_call}
3616 attribute will always be close enough to be called with an unconditional
3617 branch-and-link instruction, which has a 25-bit offset from
3618 the call site. A function marked with the @code{short_call}
3619 attribute will always be close enough to be called with a conditional
3620 branch-and-link instruction, which has a 21-bit offset from
3624 @node ARM Function Attributes
3625 @subsection ARM Function Attributes
3627 These function attributes are supported for ARM targets:
3631 @cindex @code{interrupt} function attribute, ARM
3632 Use this attribute to indicate
3633 that the specified function is an interrupt handler. The compiler generates
3634 function entry and exit sequences suitable for use in an interrupt handler
3635 when this attribute is present.
3637 You can specify the kind of interrupt to be handled by
3638 adding an optional parameter to the interrupt attribute like this:
3641 void f () __attribute__ ((interrupt ("IRQ")));
3645 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3646 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3648 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3649 may be called with a word-aligned stack pointer.
3652 @cindex @code{isr} function attribute, ARM
3653 Use this attribute on ARM to write Interrupt Service Routines. This is an
3654 alias to the @code{interrupt} attribute above.
3658 @cindex @code{long_call} function attribute, ARM
3659 @cindex @code{short_call} function attribute, ARM
3660 @cindex indirect calls, ARM
3661 These attributes specify how a particular function is called.
3662 These attributes override the
3663 @option{-mlong-calls} (@pxref{ARM Options})
3664 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3665 @code{long_call} attribute indicates that the function might be far
3666 away from the call site and require a different (more expensive)
3667 calling sequence. The @code{short_call} attribute always places
3668 the offset to the function from the call site into the @samp{BL}
3669 instruction directly.
3672 @cindex @code{naked} function attribute, ARM
3673 This attribute allows the compiler to construct the
3674 requisite function declaration, while allowing the body of the
3675 function to be assembly code. The specified function will not have
3676 prologue/epilogue sequences generated by the compiler. Only basic
3677 @code{asm} statements can safely be included in naked functions
3678 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3679 basic @code{asm} and C code may appear to work, they cannot be
3680 depended upon to work reliably and are not supported.
3683 @cindex @code{pcs} function attribute, ARM
3685 The @code{pcs} attribute can be used to control the calling convention
3686 used for a function on ARM. The attribute takes an argument that specifies
3687 the calling convention to use.
3689 When compiling using the AAPCS ABI (or a variant of it) then valid
3690 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3691 order to use a variant other than @code{"aapcs"} then the compiler must
3692 be permitted to use the appropriate co-processor registers (i.e., the
3693 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3697 /* Argument passed in r0, and result returned in r0+r1. */
3698 double f2d (float) __attribute__((pcs("aapcs")));
3701 Variadic functions always use the @code{"aapcs"} calling convention and
3702 the compiler rejects attempts to specify an alternative.
3704 @item target (@var{options})
3705 @cindex @code{target} function attribute
3706 As discussed in @ref{Common Function Attributes}, this attribute
3707 allows specification of target-specific compilation options.
3709 On ARM, the following options are allowed:
3713 @cindex @code{target("thumb")} function attribute, ARM
3714 Force code generation in the Thumb (T16/T32) ISA, depending on the
3718 @cindex @code{target("arm")} function attribute, ARM
3719 Force code generation in the ARM (A32) ISA.
3721 Functions from different modes can be inlined in the caller's mode.
3724 @cindex @code{target("fpu=")} function attribute, ARM
3725 Specifies the fpu for which to tune the performance of this function.
3726 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3727 command-line option.
3733 @node AVR Function Attributes
3734 @subsection AVR Function Attributes
3736 These function attributes are supported by the AVR back end:
3740 @cindex @code{interrupt} function attribute, AVR
3741 Use this attribute to indicate
3742 that the specified function is an interrupt handler. The compiler generates
3743 function entry and exit sequences suitable for use in an interrupt handler
3744 when this attribute is present.
3746 On the AVR, the hardware globally disables interrupts when an
3747 interrupt is executed. The first instruction of an interrupt handler
3748 declared with this attribute is a @code{SEI} instruction to
3749 re-enable interrupts. See also the @code{signal} function attribute
3750 that does not insert a @code{SEI} instruction. If both @code{signal} and
3751 @code{interrupt} are specified for the same function, @code{signal}
3752 is silently ignored.
3755 @cindex @code{naked} function attribute, AVR
3756 This attribute allows the compiler to construct the
3757 requisite function declaration, while allowing the body of the
3758 function to be assembly code. The specified function will not have
3759 prologue/epilogue sequences generated by the compiler. Only basic
3760 @code{asm} statements can safely be included in naked functions
3761 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3762 basic @code{asm} and C code may appear to work, they cannot be
3763 depended upon to work reliably and are not supported.
3767 @cindex @code{OS_main} function attribute, AVR
3768 @cindex @code{OS_task} function attribute, AVR
3769 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3770 do not save/restore any call-saved register in their prologue/epilogue.
3772 The @code{OS_main} attribute can be used when there @emph{is
3773 guarantee} that interrupts are disabled at the time when the function
3774 is entered. This saves resources when the stack pointer has to be
3775 changed to set up a frame for local variables.
3777 The @code{OS_task} attribute can be used when there is @emph{no
3778 guarantee} that interrupts are disabled at that time when the function
3779 is entered like for, e@.g@. task functions in a multi-threading operating
3780 system. In that case, changing the stack pointer register is
3781 guarded by save/clear/restore of the global interrupt enable flag.
3783 The differences to the @code{naked} function attribute are:
3785 @item @code{naked} functions do not have a return instruction whereas
3786 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3787 @code{RETI} return instruction.
3788 @item @code{naked} functions do not set up a frame for local variables
3789 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3794 @cindex @code{signal} function attribute, AVR
3795 Use this attribute on the AVR to indicate that the specified
3796 function is an interrupt handler. The compiler generates function
3797 entry and exit sequences suitable for use in an interrupt handler when this
3798 attribute is present.
3800 See also the @code{interrupt} function attribute.
3802 The AVR hardware globally disables interrupts when an interrupt is executed.
3803 Interrupt handler functions defined with the @code{signal} attribute
3804 do not re-enable interrupts. It is save to enable interrupts in a
3805 @code{signal} handler. This ``save'' only applies to the code
3806 generated by the compiler and not to the IRQ layout of the
3807 application which is responsibility of the application.
3809 If both @code{signal} and @code{interrupt} are specified for the same
3810 function, @code{signal} is silently ignored.
3813 @node Blackfin Function Attributes
3814 @subsection Blackfin Function Attributes
3816 These function attributes are supported by the Blackfin back end:
3820 @item exception_handler
3821 @cindex @code{exception_handler} function attribute
3822 @cindex exception handler functions, Blackfin
3823 Use this attribute on the Blackfin to indicate that the specified function
3824 is an exception handler. The compiler generates function entry and
3825 exit sequences suitable for use in an exception handler when this
3826 attribute is present.
3828 @item interrupt_handler
3829 @cindex @code{interrupt_handler} function attribute, Blackfin
3830 Use this attribute to
3831 indicate that the specified function is an interrupt handler. The compiler
3832 generates function entry and exit sequences suitable for use in an
3833 interrupt handler when this attribute is present.
3836 @cindex @code{kspisusp} function attribute, Blackfin
3837 @cindex User stack pointer in interrupts on the Blackfin
3838 When used together with @code{interrupt_handler}, @code{exception_handler}
3839 or @code{nmi_handler}, code is generated to load the stack pointer
3840 from the USP register in the function prologue.
3843 @cindex @code{l1_text} function attribute, Blackfin
3844 This attribute specifies a function to be placed into L1 Instruction
3845 SRAM@. The function is put into a specific section named @code{.l1.text}.
3846 With @option{-mfdpic}, function calls with a such function as the callee
3847 or caller uses inlined PLT.
3850 @cindex @code{l2} function attribute, Blackfin
3851 This attribute specifies a function to be placed into L2
3852 SRAM. The function is put into a specific section named
3853 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3858 @cindex indirect calls, Blackfin
3859 @cindex @code{longcall} function attribute, Blackfin
3860 @cindex @code{shortcall} function attribute, Blackfin
3861 The @code{longcall} attribute
3862 indicates that the function might be far away from the call site and
3863 require a different (more expensive) calling sequence. The
3864 @code{shortcall} attribute indicates that the function is always close
3865 enough for the shorter calling sequence to be used. These attributes
3866 override the @option{-mlongcall} switch.
3869 @cindex @code{nesting} function attribute, Blackfin
3870 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3871 Use this attribute together with @code{interrupt_handler},
3872 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3873 entry code should enable nested interrupts or exceptions.
3876 @cindex @code{nmi_handler} function attribute, Blackfin
3877 @cindex NMI handler functions on the Blackfin processor
3878 Use this attribute on the Blackfin to indicate that the specified function
3879 is an NMI handler. The compiler generates function entry and
3880 exit sequences suitable for use in an NMI handler when this
3881 attribute is present.
3884 @cindex @code{saveall} function attribute, Blackfin
3885 @cindex save all registers on the Blackfin
3886 Use this attribute to indicate that
3887 all registers except the stack pointer should be saved in the prologue
3888 regardless of whether they are used or not.
3891 @node CR16 Function Attributes
3892 @subsection CR16 Function Attributes
3894 These function attributes are supported by the CR16 back end:
3898 @cindex @code{interrupt} function attribute, CR16
3899 Use this attribute to indicate
3900 that the specified function is an interrupt handler. The compiler generates
3901 function entry and exit sequences suitable for use in an interrupt handler
3902 when this attribute is present.
3905 @node Epiphany Function Attributes
3906 @subsection Epiphany Function Attributes
3908 These function attributes are supported by the Epiphany back end:
3912 @cindex @code{disinterrupt} function attribute, Epiphany
3913 This attribute causes the compiler to emit
3914 instructions to disable interrupts for the duration of the given
3917 @item forwarder_section
3918 @cindex @code{forwarder_section} function attribute, Epiphany
3919 This attribute modifies the behavior of an interrupt handler.
3920 The interrupt handler may be in external memory which cannot be
3921 reached by a branch instruction, so generate a local memory trampoline
3922 to transfer control. The single parameter identifies the section where
3923 the trampoline is placed.
3926 @cindex @code{interrupt} function attribute, Epiphany
3927 Use this attribute to indicate
3928 that the specified function is an interrupt handler. The compiler generates
3929 function entry and exit sequences suitable for use in an interrupt handler
3930 when this attribute is present. It may also generate
3931 a special section with code to initialize the interrupt vector table.
3933 On Epiphany targets one or more optional parameters can be added like this:
3936 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3939 Permissible values for these parameters are: @w{@code{reset}},
3940 @w{@code{software_exception}}, @w{@code{page_miss}},
3941 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3942 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3943 Multiple parameters indicate that multiple entries in the interrupt
3944 vector table should be initialized for this function, i.e.@: for each
3945 parameter @w{@var{name}}, a jump to the function is emitted in
3946 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3947 entirely, in which case no interrupt vector table entry is provided.
3949 Note that interrupts are enabled inside the function
3950 unless the @code{disinterrupt} attribute is also specified.
3952 The following examples are all valid uses of these attributes on
3955 void __attribute__ ((interrupt)) universal_handler ();
3956 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3957 void __attribute__ ((interrupt ("dma0, dma1")))
3958 universal_dma_handler ();
3959 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3960 fast_timer_handler ();
3961 void __attribute__ ((interrupt ("dma0, dma1"),
3962 forwarder_section ("tramp")))
3963 external_dma_handler ();
3968 @cindex @code{long_call} function attribute, Epiphany
3969 @cindex @code{short_call} function attribute, Epiphany
3970 @cindex indirect calls, Epiphany
3971 These attributes specify how a particular function is called.
3972 These attributes override the
3973 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3974 command-line switch and @code{#pragma long_calls} settings.
3978 @node H8/300 Function Attributes
3979 @subsection H8/300 Function Attributes
3981 These function attributes are available for H8/300 targets:
3984 @item function_vector
3985 @cindex @code{function_vector} function attribute, H8/300
3986 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3987 that the specified function should be called through the function vector.
3988 Calling a function through the function vector reduces code size; however,
3989 the function vector has a limited size (maximum 128 entries on the H8/300
3990 and 64 entries on the H8/300H and H8S)
3991 and shares space with the interrupt vector.
3993 @item interrupt_handler
3994 @cindex @code{interrupt_handler} function attribute, H8/300
3995 Use this attribute on the H8/300, H8/300H, and H8S to
3996 indicate that the specified function is an interrupt handler. The compiler
3997 generates function entry and exit sequences suitable for use in an
3998 interrupt handler when this attribute is present.
4001 @cindex @code{saveall} function attribute, H8/300
4002 @cindex save all registers on the H8/300, H8/300H, and H8S
4003 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
4004 all registers except the stack pointer should be saved in the prologue
4005 regardless of whether they are used or not.
4008 @node IA-64 Function Attributes
4009 @subsection IA-64 Function Attributes
4011 These function attributes are supported on IA-64 targets:
4014 @item syscall_linkage
4015 @cindex @code{syscall_linkage} function attribute, IA-64
4016 This attribute is used to modify the IA-64 calling convention by marking
4017 all input registers as live at all function exits. This makes it possible
4018 to restart a system call after an interrupt without having to save/restore
4019 the input registers. This also prevents kernel data from leaking into
4023 @cindex @code{version_id} function attribute, IA-64
4024 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4025 symbol to contain a version string, thus allowing for function level
4026 versioning. HP-UX system header files may use function level versioning
4027 for some system calls.
4030 extern int foo () __attribute__((version_id ("20040821")));
4034 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4037 @node M32C Function Attributes
4038 @subsection M32C Function Attributes
4040 These function attributes are supported by the M32C back end:
4044 @cindex @code{bank_switch} function attribute, M32C
4045 When added to an interrupt handler with the M32C port, causes the
4046 prologue and epilogue to use bank switching to preserve the registers
4047 rather than saving them on the stack.
4049 @item fast_interrupt
4050 @cindex @code{fast_interrupt} function attribute, M32C
4051 Use this attribute on the M32C port to indicate that the specified
4052 function is a fast interrupt handler. This is just like the
4053 @code{interrupt} attribute, except that @code{freit} is used to return
4054 instead of @code{reit}.
4056 @item function_vector
4057 @cindex @code{function_vector} function attribute, M16C/M32C
4058 On M16C/M32C targets, the @code{function_vector} attribute declares a
4059 special page subroutine call function. Use of this attribute reduces
4060 the code size by 2 bytes for each call generated to the
4061 subroutine. The argument to the attribute is the vector number entry
4062 from the special page vector table which contains the 16 low-order
4063 bits of the subroutine's entry address. Each vector table has special
4064 page number (18 to 255) that is used in @code{jsrs} instructions.
4065 Jump addresses of the routines are generated by adding 0x0F0000 (in
4066 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4067 2-byte addresses set in the vector table. Therefore you need to ensure
4068 that all the special page vector routines should get mapped within the
4069 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4072 In the following example 2 bytes are saved for each call to
4073 function @code{foo}.
4076 void foo (void) __attribute__((function_vector(0x18)));
4087 If functions are defined in one file and are called in another file,
4088 then be sure to write this declaration in both files.
4090 This attribute is ignored for R8C target.
4093 @cindex @code{interrupt} function attribute, M32C
4094 Use this attribute to indicate
4095 that the specified function is an interrupt handler. The compiler generates
4096 function entry and exit sequences suitable for use in an interrupt handler
4097 when this attribute is present.
4100 @node M32R/D Function Attributes
4101 @subsection M32R/D Function Attributes
4103 These function attributes are supported by the M32R/D back end:
4107 @cindex @code{interrupt} function attribute, M32R/D
4108 Use this attribute to indicate
4109 that the specified function is an interrupt handler. The compiler generates
4110 function entry and exit sequences suitable for use in an interrupt handler
4111 when this attribute is present.
4113 @item model (@var{model-name})
4114 @cindex @code{model} function attribute, M32R/D
4115 @cindex function addressability on the M32R/D
4117 On the M32R/D, use this attribute to set the addressability of an
4118 object, and of the code generated for a function. The identifier
4119 @var{model-name} is one of @code{small}, @code{medium}, or
4120 @code{large}, representing each of the code models.
4122 Small model objects live in the lower 16MB of memory (so that their
4123 addresses can be loaded with the @code{ld24} instruction), and are
4124 callable with the @code{bl} instruction.
4126 Medium model objects may live anywhere in the 32-bit address space (the
4127 compiler generates @code{seth/add3} instructions to load their addresses),
4128 and are callable with the @code{bl} instruction.
4130 Large model objects may live anywhere in the 32-bit address space (the
4131 compiler generates @code{seth/add3} instructions to load their addresses),
4132 and may not be reachable with the @code{bl} instruction (the compiler
4133 generates the much slower @code{seth/add3/jl} instruction sequence).
4136 @node m68k Function Attributes
4137 @subsection m68k Function Attributes
4139 These function attributes are supported by the m68k back end:
4143 @itemx interrupt_handler
4144 @cindex @code{interrupt} function attribute, m68k
4145 @cindex @code{interrupt_handler} function attribute, m68k
4146 Use this attribute to
4147 indicate that the specified function is an interrupt handler. The compiler
4148 generates function entry and exit sequences suitable for use in an
4149 interrupt handler when this attribute is present. Either name may be used.
4151 @item interrupt_thread
4152 @cindex @code{interrupt_thread} function attribute, fido
4153 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4154 that the specified function is an interrupt handler that is designed
4155 to run as a thread. The compiler omits generate prologue/epilogue
4156 sequences and replaces the return instruction with a @code{sleep}
4157 instruction. This attribute is available only on fido.
4160 @node MCORE Function Attributes
4161 @subsection MCORE Function Attributes
4163 These function attributes are supported by the MCORE back end:
4167 @cindex @code{naked} function attribute, MCORE
4168 This attribute allows the compiler to construct the
4169 requisite function declaration, while allowing the body of the
4170 function to be assembly code. The specified function will not have
4171 prologue/epilogue sequences generated by the compiler. Only basic
4172 @code{asm} statements can safely be included in naked functions
4173 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4174 basic @code{asm} and C code may appear to work, they cannot be
4175 depended upon to work reliably and are not supported.
4178 @node MeP Function Attributes
4179 @subsection MeP Function Attributes
4181 These function attributes are supported by the MeP back end:
4185 @cindex @code{disinterrupt} function attribute, MeP
4186 On MeP targets, this attribute causes the compiler to emit
4187 instructions to disable interrupts for the duration of the given
4191 @cindex @code{interrupt} function attribute, MeP
4192 Use this attribute to indicate
4193 that the specified function is an interrupt handler. The compiler generates
4194 function entry and exit sequences suitable for use in an interrupt handler
4195 when this attribute is present.
4198 @cindex @code{near} function attribute, MeP
4199 This attribute causes the compiler to assume the called
4200 function is close enough to use the normal calling convention,
4201 overriding the @option{-mtf} command-line option.
4204 @cindex @code{far} function attribute, MeP
4205 On MeP targets this causes the compiler to use a calling convention
4206 that assumes the called function is too far away for the built-in
4210 @cindex @code{vliw} function attribute, MeP
4211 The @code{vliw} attribute tells the compiler to emit
4212 instructions in VLIW mode instead of core mode. Note that this
4213 attribute is not allowed unless a VLIW coprocessor has been configured
4214 and enabled through command-line options.
4217 @node MicroBlaze Function Attributes
4218 @subsection MicroBlaze Function Attributes
4220 These function attributes are supported on MicroBlaze targets:
4223 @item save_volatiles
4224 @cindex @code{save_volatiles} function attribute, MicroBlaze
4225 Use this attribute to indicate that the function is
4226 an interrupt handler. All volatile registers (in addition to non-volatile
4227 registers) are saved in the function prologue. If the function is a leaf
4228 function, only volatiles used by the function are saved. A normal function
4229 return is generated instead of a return from interrupt.
4232 @cindex @code{break_handler} function attribute, MicroBlaze
4233 @cindex break handler functions
4234 Use this attribute to indicate that
4235 the specified function is a break handler. The compiler generates function
4236 entry and exit sequences suitable for use in an break handler when this
4237 attribute is present. The return from @code{break_handler} is done through
4238 the @code{rtbd} instead of @code{rtsd}.
4241 void f () __attribute__ ((break_handler));
4245 @node Microsoft Windows Function Attributes
4246 @subsection Microsoft Windows Function Attributes
4248 The following attributes are available on Microsoft Windows and Symbian OS
4253 @cindex @code{dllexport} function attribute
4254 @cindex @code{__declspec(dllexport)}
4255 On Microsoft Windows targets and Symbian OS targets the
4256 @code{dllexport} attribute causes the compiler to provide a global
4257 pointer to a pointer in a DLL, so that it can be referenced with the
4258 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4259 name is formed by combining @code{_imp__} and the function or variable
4262 You can use @code{__declspec(dllexport)} as a synonym for
4263 @code{__attribute__ ((dllexport))} for compatibility with other
4266 On systems that support the @code{visibility} attribute, this
4267 attribute also implies ``default'' visibility. It is an error to
4268 explicitly specify any other visibility.
4270 GCC's default behavior is to emit all inline functions with the
4271 @code{dllexport} attribute. Since this can cause object file-size bloat,
4272 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4273 ignore the attribute for inlined functions unless the
4274 @option{-fkeep-inline-functions} flag is used instead.
4276 The attribute is ignored for undefined symbols.
4278 When applied to C++ classes, the attribute marks defined non-inlined
4279 member functions and static data members as exports. Static consts
4280 initialized in-class are not marked unless they are also defined
4283 For Microsoft Windows targets there are alternative methods for
4284 including the symbol in the DLL's export table such as using a
4285 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4286 the @option{--export-all} linker flag.
4289 @cindex @code{dllimport} function attribute
4290 @cindex @code{__declspec(dllimport)}
4291 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4292 attribute causes the compiler to reference a function or variable via
4293 a global pointer to a pointer that is set up by the DLL exporting the
4294 symbol. The attribute implies @code{extern}. On Microsoft Windows
4295 targets, the pointer name is formed by combining @code{_imp__} and the
4296 function or variable name.
4298 You can use @code{__declspec(dllimport)} as a synonym for
4299 @code{__attribute__ ((dllimport))} for compatibility with other
4302 On systems that support the @code{visibility} attribute, this
4303 attribute also implies ``default'' visibility. It is an error to
4304 explicitly specify any other visibility.
4306 Currently, the attribute is ignored for inlined functions. If the
4307 attribute is applied to a symbol @emph{definition}, an error is reported.
4308 If a symbol previously declared @code{dllimport} is later defined, the
4309 attribute is ignored in subsequent references, and a warning is emitted.
4310 The attribute is also overridden by a subsequent declaration as
4313 When applied to C++ classes, the attribute marks non-inlined
4314 member functions and static data members as imports. However, the
4315 attribute is ignored for virtual methods to allow creation of vtables
4318 On the SH Symbian OS target the @code{dllimport} attribute also has
4319 another affect---it can cause the vtable and run-time type information
4320 for a class to be exported. This happens when the class has a
4321 dllimported constructor or a non-inline, non-pure virtual function
4322 and, for either of those two conditions, the class also has an inline
4323 constructor or destructor and has a key function that is defined in
4324 the current translation unit.
4326 For Microsoft Windows targets the use of the @code{dllimport}
4327 attribute on functions is not necessary, but provides a small
4328 performance benefit by eliminating a thunk in the DLL@. The use of the
4329 @code{dllimport} attribute on imported variables can be avoided by passing the
4330 @option{--enable-auto-import} switch to the GNU linker. As with
4331 functions, using the attribute for a variable eliminates a thunk in
4334 One drawback to using this attribute is that a pointer to a
4335 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4336 address. However, a pointer to a @emph{function} with the
4337 @code{dllimport} attribute can be used as a constant initializer; in
4338 this case, the address of a stub function in the import lib is
4339 referenced. On Microsoft Windows targets, the attribute can be disabled
4340 for functions by setting the @option{-mnop-fun-dllimport} flag.
4343 @node MIPS Function Attributes
4344 @subsection MIPS Function Attributes
4346 These function attributes are supported by the MIPS back end:
4350 @cindex @code{interrupt} function attribute, MIPS
4351 Use this attribute to indicate that the specified function is an interrupt
4352 handler. The compiler generates function entry and exit sequences suitable
4353 for use in an interrupt handler when this attribute is present.
4354 An optional argument is supported for the interrupt attribute which allows
4355 the interrupt mode to be described. By default GCC assumes the external
4356 interrupt controller (EIC) mode is in use, this can be explicitly set using
4357 @code{eic}. When interrupts are non-masked then the requested Interrupt
4358 Priority Level (IPL) is copied to the current IPL which has the effect of only
4359 enabling higher priority interrupts. To use vectored interrupt mode use
4360 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4361 the behaviour of the non-masked interrupt support and GCC will arrange to mask
4362 all interrupts from sw0 up to and including the specified interrupt vector.
4364 You can use the following attributes to modify the behavior
4365 of an interrupt handler:
4367 @item use_shadow_register_set
4368 @cindex @code{use_shadow_register_set} function attribute, MIPS
4369 Assume that the handler uses a shadow register set, instead of
4370 the main general-purpose registers. An optional argument @code{intstack} is
4371 supported to indicate that the shadow register set contains a valid stack
4374 @item keep_interrupts_masked
4375 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4376 Keep interrupts masked for the whole function. Without this attribute,
4377 GCC tries to reenable interrupts for as much of the function as it can.
4379 @item use_debug_exception_return
4380 @cindex @code{use_debug_exception_return} function attribute, MIPS
4381 Return using the @code{deret} instruction. Interrupt handlers that don't
4382 have this attribute return using @code{eret} instead.
4385 You can use any combination of these attributes, as shown below:
4387 void __attribute__ ((interrupt)) v0 ();
4388 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4389 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4390 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4391 void __attribute__ ((interrupt, use_shadow_register_set,
4392 keep_interrupts_masked)) v4 ();
4393 void __attribute__ ((interrupt, use_shadow_register_set,
4394 use_debug_exception_return)) v5 ();
4395 void __attribute__ ((interrupt, keep_interrupts_masked,
4396 use_debug_exception_return)) v6 ();
4397 void __attribute__ ((interrupt, use_shadow_register_set,
4398 keep_interrupts_masked,
4399 use_debug_exception_return)) v7 ();
4400 void __attribute__ ((interrupt("eic"))) v8 ();
4401 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4407 @cindex indirect calls, MIPS
4408 @cindex @code{long_call} function attribute, MIPS
4409 @cindex @code{near} function attribute, MIPS
4410 @cindex @code{far} function attribute, MIPS
4411 These attributes specify how a particular function is called on MIPS@.
4412 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4413 command-line switch. The @code{long_call} and @code{far} attributes are
4414 synonyms, and cause the compiler to always call
4415 the function by first loading its address into a register, and then using
4416 the contents of that register. The @code{near} attribute has the opposite
4417 effect; it specifies that non-PIC calls should be made using the more
4418 efficient @code{jal} instruction.
4422 @cindex @code{mips16} function attribute, MIPS
4423 @cindex @code{nomips16} function attribute, MIPS
4425 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4426 function attributes to locally select or turn off MIPS16 code generation.
4427 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4428 while MIPS16 code generation is disabled for functions with the
4429 @code{nomips16} attribute. These attributes override the
4430 @option{-mips16} and @option{-mno-mips16} options on the command line
4431 (@pxref{MIPS Options}).
4433 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4434 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4435 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4436 may interact badly with some GCC extensions such as @code{__builtin_apply}
4437 (@pxref{Constructing Calls}).
4439 @item micromips, MIPS
4440 @itemx nomicromips, MIPS
4441 @cindex @code{micromips} function attribute
4442 @cindex @code{nomicromips} function attribute
4444 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4445 function attributes to locally select or turn off microMIPS code generation.
4446 A function with the @code{micromips} attribute is emitted as microMIPS code,
4447 while microMIPS code generation is disabled for functions with the
4448 @code{nomicromips} attribute. These attributes override the
4449 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4450 (@pxref{MIPS Options}).
4452 When compiling files containing mixed microMIPS and non-microMIPS code, the
4453 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4455 not that within individual functions. Mixed microMIPS and non-microMIPS code
4456 may interact badly with some GCC extensions such as @code{__builtin_apply}
4457 (@pxref{Constructing Calls}).
4460 @cindex @code{nocompression} function attribute, MIPS
4461 On MIPS targets, you can use the @code{nocompression} function attribute
4462 to locally turn off MIPS16 and microMIPS code generation. This attribute
4463 overrides the @option{-mips16} and @option{-mmicromips} options on the
4464 command line (@pxref{MIPS Options}).
4467 @node MSP430 Function Attributes
4468 @subsection MSP430 Function Attributes
4470 These function attributes are supported by the MSP430 back end:
4474 @cindex @code{critical} function attribute, MSP430
4475 Critical functions disable interrupts upon entry and restore the
4476 previous interrupt state upon exit. Critical functions cannot also
4477 have the @code{naked} or @code{reentrant} attributes. They can have
4478 the @code{interrupt} attribute.
4481 @cindex @code{interrupt} function attribute, MSP430
4482 Use this attribute to indicate
4483 that the specified function is an interrupt handler. The compiler generates
4484 function entry and exit sequences suitable for use in an interrupt handler
4485 when this attribute is present.
4487 You can provide an argument to the interrupt
4488 attribute which specifies a name or number. If the argument is a
4489 number it indicates the slot in the interrupt vector table (0 - 31) to
4490 which this handler should be assigned. If the argument is a name it
4491 is treated as a symbolic name for the vector slot. These names should
4492 match up with appropriate entries in the linker script. By default
4493 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4494 @code{reset} for vector 31 are recognized.
4497 @cindex @code{naked} function attribute, MSP430
4498 This attribute allows the compiler to construct the
4499 requisite function declaration, while allowing the body of the
4500 function to be assembly code. The specified function will not have
4501 prologue/epilogue sequences generated by the compiler. Only basic
4502 @code{asm} statements can safely be included in naked functions
4503 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4504 basic @code{asm} and C code may appear to work, they cannot be
4505 depended upon to work reliably and are not supported.
4508 @cindex @code{reentrant} function attribute, MSP430
4509 Reentrant functions disable interrupts upon entry and enable them
4510 upon exit. Reentrant functions cannot also have the @code{naked}
4511 or @code{critical} attributes. They can have the @code{interrupt}
4515 @cindex @code{wakeup} function attribute, MSP430
4516 This attribute only applies to interrupt functions. It is silently
4517 ignored if applied to a non-interrupt function. A wakeup interrupt
4518 function will rouse the processor from any low-power state that it
4519 might be in when the function exits.
4522 @node NDS32 Function Attributes
4523 @subsection NDS32 Function Attributes
4525 These function attributes are supported by the NDS32 back end:
4529 @cindex @code{exception} function attribute
4530 @cindex exception handler functions, NDS32
4531 Use this attribute on the NDS32 target to indicate that the specified function
4532 is an exception handler. The compiler will generate corresponding sections
4533 for use in an exception handler.
4536 @cindex @code{interrupt} function attribute, NDS32
4537 On NDS32 target, this attribute indicates that the specified function
4538 is an interrupt handler. The compiler generates corresponding sections
4539 for use in an interrupt handler. You can use the following attributes
4540 to modify the behavior:
4543 @cindex @code{nested} function attribute, NDS32
4544 This interrupt service routine is interruptible.
4546 @cindex @code{not_nested} function attribute, NDS32
4547 This interrupt service routine is not interruptible.
4549 @cindex @code{nested_ready} function attribute, NDS32
4550 This interrupt service routine is interruptible after @code{PSW.GIE}
4551 (global interrupt enable) is set. This allows interrupt service routine to
4552 finish some short critical code before enabling interrupts.
4554 @cindex @code{save_all} function attribute, NDS32
4555 The system will help save all registers into stack before entering
4558 @cindex @code{partial_save} function attribute, NDS32
4559 The system will help save caller registers into stack before entering
4564 @cindex @code{naked} function attribute, NDS32
4565 This attribute allows the compiler to construct the
4566 requisite function declaration, while allowing the body of the
4567 function to be assembly code. The specified function will not have
4568 prologue/epilogue sequences generated by the compiler. Only basic
4569 @code{asm} statements can safely be included in naked functions
4570 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4571 basic @code{asm} and C code may appear to work, they cannot be
4572 depended upon to work reliably and are not supported.
4575 @cindex @code{reset} function attribute, NDS32
4576 @cindex reset handler functions
4577 Use this attribute on the NDS32 target to indicate that the specified function
4578 is a reset handler. The compiler will generate corresponding sections
4579 for use in a reset handler. You can use the following attributes
4580 to provide extra exception handling:
4583 @cindex @code{nmi} function attribute, NDS32
4584 Provide a user-defined function to handle NMI exception.
4586 @cindex @code{warm} function attribute, NDS32
4587 Provide a user-defined function to handle warm reset exception.
4591 @node Nios II Function Attributes
4592 @subsection Nios II Function Attributes
4594 These function attributes are supported by the Nios II back end:
4597 @item target (@var{options})
4598 @cindex @code{target} function attribute
4599 As discussed in @ref{Common Function Attributes}, this attribute
4600 allows specification of target-specific compilation options.
4602 When compiling for Nios II, the following options are allowed:
4605 @item custom-@var{insn}=@var{N}
4606 @itemx no-custom-@var{insn}
4607 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4608 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4609 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4610 custom instruction with encoding @var{N} when generating code that uses
4611 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4612 the custom instruction @var{insn}.
4613 These target attributes correspond to the
4614 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4615 command-line options, and support the same set of @var{insn} keywords.
4616 @xref{Nios II Options}, for more information.
4618 @item custom-fpu-cfg=@var{name}
4619 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4620 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4621 command-line option, to select a predefined set of custom instructions
4623 @xref{Nios II Options}, for more information.
4627 @node PowerPC Function Attributes
4628 @subsection PowerPC Function Attributes
4630 These function attributes are supported by the PowerPC back end:
4635 @cindex indirect calls, PowerPC
4636 @cindex @code{longcall} function attribute, PowerPC
4637 @cindex @code{shortcall} function attribute, PowerPC
4638 The @code{longcall} attribute
4639 indicates that the function might be far away from the call site and
4640 require a different (more expensive) calling sequence. The
4641 @code{shortcall} attribute indicates that the function is always close
4642 enough for the shorter calling sequence to be used. These attributes
4643 override both the @option{-mlongcall} switch and
4644 the @code{#pragma longcall} setting.
4646 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4647 calls are necessary.
4649 @item target (@var{options})
4650 @cindex @code{target} function attribute
4651 As discussed in @ref{Common Function Attributes}, this attribute
4652 allows specification of target-specific compilation options.
4654 On the PowerPC, the following options are allowed:
4659 @cindex @code{target("altivec")} function attribute, PowerPC
4660 Generate code that uses (does not use) AltiVec instructions. In
4661 32-bit code, you cannot enable AltiVec instructions unless
4662 @option{-mabi=altivec} is used on the command line.
4666 @cindex @code{target("cmpb")} function attribute, PowerPC
4667 Generate code that uses (does not use) the compare bytes instruction
4668 implemented on the POWER6 processor and other processors that support
4669 the PowerPC V2.05 architecture.
4673 @cindex @code{target("dlmzb")} function attribute, PowerPC
4674 Generate code that uses (does not use) the string-search @samp{dlmzb}
4675 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4676 generated by default when targeting those processors.
4680 @cindex @code{target("fprnd")} function attribute, PowerPC
4681 Generate code that uses (does not use) the FP round to integer
4682 instructions implemented on the POWER5+ processor and other processors
4683 that support the PowerPC V2.03 architecture.
4687 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4688 Generate code that uses (does not use) the decimal floating-point
4689 instructions implemented on some POWER processors.
4693 @cindex @code{target("isel")} function attribute, PowerPC
4694 Generate code that uses (does not use) ISEL instruction.
4698 @cindex @code{target("mfcrf")} function attribute, PowerPC
4699 Generate code that uses (does not use) the move from condition
4700 register field instruction implemented on the POWER4 processor and
4701 other processors that support the PowerPC V2.01 architecture.
4705 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4706 Generate code that uses (does not use) the FP move to/from general
4707 purpose register instructions implemented on the POWER6X processor and
4708 other processors that support the extended PowerPC V2.05 architecture.
4712 @cindex @code{target("mulhw")} function attribute, PowerPC
4713 Generate code that uses (does not use) the half-word multiply and
4714 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4715 These instructions are generated by default when targeting those
4720 @cindex @code{target("multiple")} function attribute, PowerPC
4721 Generate code that uses (does not use) the load multiple word
4722 instructions and the store multiple word instructions.
4726 @cindex @code{target("update")} function attribute, PowerPC
4727 Generate code that uses (does not use) the load or store instructions
4728 that update the base register to the address of the calculated memory
4733 @cindex @code{target("popcntb")} function attribute, PowerPC
4734 Generate code that uses (does not use) the popcount and double-precision
4735 FP reciprocal estimate instruction implemented on the POWER5
4736 processor and other processors that support the PowerPC V2.02
4741 @cindex @code{target("popcntd")} function attribute, PowerPC
4742 Generate code that uses (does not use) the popcount instruction
4743 implemented on the POWER7 processor and other processors that support
4744 the PowerPC V2.06 architecture.
4746 @item powerpc-gfxopt
4747 @itemx no-powerpc-gfxopt
4748 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4749 Generate code that uses (does not use) the optional PowerPC
4750 architecture instructions in the Graphics group, including
4751 floating-point select.
4754 @itemx no-powerpc-gpopt
4755 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4756 Generate code that uses (does not use) the optional PowerPC
4757 architecture instructions in the General Purpose group, including
4758 floating-point square root.
4760 @item recip-precision
4761 @itemx no-recip-precision
4762 @cindex @code{target("recip-precision")} function attribute, PowerPC
4763 Assume (do not assume) that the reciprocal estimate instructions
4764 provide higher-precision estimates than is mandated by the PowerPC
4769 @cindex @code{target("string")} function attribute, PowerPC
4770 Generate code that uses (does not use) the load string instructions
4771 and the store string word instructions to save multiple registers and
4772 do small block moves.
4776 @cindex @code{target("vsx")} function attribute, PowerPC
4777 Generate code that uses (does not use) vector/scalar (VSX)
4778 instructions, and also enable the use of built-in functions that allow
4779 more direct access to the VSX instruction set. In 32-bit code, you
4780 cannot enable VSX or AltiVec instructions unless
4781 @option{-mabi=altivec} is used on the command line.
4785 @cindex @code{target("friz")} function attribute, PowerPC
4786 Generate (do not generate) the @code{friz} instruction when the
4787 @option{-funsafe-math-optimizations} option is used to optimize
4788 rounding a floating-point value to 64-bit integer and back to floating
4789 point. The @code{friz} instruction does not return the same value if
4790 the floating-point number is too large to fit in an integer.
4792 @item avoid-indexed-addresses
4793 @itemx no-avoid-indexed-addresses
4794 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4795 Generate code that tries to avoid (not avoid) the use of indexed load
4796 or store instructions.
4800 @cindex @code{target("paired")} function attribute, PowerPC
4801 Generate code that uses (does not use) the generation of PAIRED simd
4806 @cindex @code{target("longcall")} function attribute, PowerPC
4807 Generate code that assumes (does not assume) that all calls are far
4808 away so that a longer more expensive calling sequence is required.
4811 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4812 Specify the architecture to generate code for when compiling the
4813 function. If you select the @code{target("cpu=power7")} attribute when
4814 generating 32-bit code, VSX and AltiVec instructions are not generated
4815 unless you use the @option{-mabi=altivec} option on the command line.
4817 @item tune=@var{TUNE}
4818 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4819 Specify the architecture to tune for when compiling the function. If
4820 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4821 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4822 compilation tunes for the @var{CPU} architecture, and not the
4823 default tuning specified on the command line.
4826 On the PowerPC, the inliner does not inline a
4827 function that has different target options than the caller, unless the
4828 callee has a subset of the target options of the caller.
4831 @node RL78 Function Attributes
4832 @subsection RL78 Function Attributes
4834 These function attributes are supported by the RL78 back end:
4838 @itemx brk_interrupt
4839 @cindex @code{interrupt} function attribute, RL78
4840 @cindex @code{brk_interrupt} function attribute, RL78
4841 These attributes indicate
4842 that the specified function is an interrupt handler. The compiler generates
4843 function entry and exit sequences suitable for use in an interrupt handler
4844 when this attribute is present.
4846 Use @code{brk_interrupt} instead of @code{interrupt} for
4847 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4848 that must end with @code{RETB} instead of @code{RETI}).
4851 @cindex @code{naked} function attribute, RL78
4852 This attribute allows the compiler to construct the
4853 requisite function declaration, while allowing the body of the
4854 function to be assembly code. The specified function will not have
4855 prologue/epilogue sequences generated by the compiler. Only basic
4856 @code{asm} statements can safely be included in naked functions
4857 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4858 basic @code{asm} and C code may appear to work, they cannot be
4859 depended upon to work reliably and are not supported.
4862 @node RX Function Attributes
4863 @subsection RX Function Attributes
4865 These function attributes are supported by the RX back end:
4868 @item fast_interrupt
4869 @cindex @code{fast_interrupt} function attribute, RX
4870 Use this attribute on the RX port to indicate that the specified
4871 function is a fast interrupt handler. This is just like the
4872 @code{interrupt} attribute, except that @code{freit} is used to return
4873 instead of @code{reit}.
4876 @cindex @code{interrupt} function attribute, RX
4877 Use this attribute to indicate
4878 that the specified function is an interrupt handler. The compiler generates
4879 function entry and exit sequences suitable for use in an interrupt handler
4880 when this attribute is present.
4882 On RX targets, you may specify one or more vector numbers as arguments
4883 to the attribute, as well as naming an alternate table name.
4884 Parameters are handled sequentially, so one handler can be assigned to
4885 multiple entries in multiple tables. One may also pass the magic
4886 string @code{"$default"} which causes the function to be used for any
4887 unfilled slots in the current table.
4889 This example shows a simple assignment of a function to one vector in
4890 the default table (note that preprocessor macros may be used for
4891 chip-specific symbolic vector names):
4893 void __attribute__ ((interrupt (5))) txd1_handler ();
4896 This example assigns a function to two slots in the default table
4897 (using preprocessor macros defined elsewhere) and makes it the default
4898 for the @code{dct} table:
4900 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4905 @cindex @code{naked} function attribute, RX
4906 This attribute allows the compiler to construct the
4907 requisite function declaration, while allowing the body of the
4908 function to be assembly code. The specified function will not have
4909 prologue/epilogue sequences generated by the compiler. Only basic
4910 @code{asm} statements can safely be included in naked functions
4911 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4912 basic @code{asm} and C code may appear to work, they cannot be
4913 depended upon to work reliably and are not supported.
4916 @cindex @code{vector} function attribute, RX
4917 This RX attribute is similar to the @code{interrupt} attribute, including its
4918 parameters, but does not make the function an interrupt-handler type
4919 function (i.e. it retains the normal C function calling ABI). See the
4920 @code{interrupt} attribute for a description of its arguments.
4923 @node S/390 Function Attributes
4924 @subsection S/390 Function Attributes
4926 These function attributes are supported on the S/390:
4929 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4930 @cindex @code{hotpatch} function attribute, S/390
4932 On S/390 System z targets, you can use this function attribute to
4933 make GCC generate a ``hot-patching'' function prologue. If the
4934 @option{-mhotpatch=} command-line option is used at the same time,
4935 the @code{hotpatch} attribute takes precedence. The first of the
4936 two arguments specifies the number of halfwords to be added before
4937 the function label. A second argument can be used to specify the
4938 number of halfwords to be added after the function label. For
4939 both arguments the maximum allowed value is 1000000.
4941 If both arguments are zero, hotpatching is disabled.
4944 @node SH Function Attributes
4945 @subsection SH Function Attributes
4947 These function attributes are supported on the SH family of processors:
4950 @item function_vector
4951 @cindex @code{function_vector} function attribute, SH
4952 @cindex calling functions through the function vector on SH2A
4953 On SH2A targets, this attribute declares a function to be called using the
4954 TBR relative addressing mode. The argument to this attribute is the entry
4955 number of the same function in a vector table containing all the TBR
4956 relative addressable functions. For correct operation the TBR must be setup
4957 accordingly to point to the start of the vector table before any functions with
4958 this attribute are invoked. Usually a good place to do the initialization is
4959 the startup routine. The TBR relative vector table can have at max 256 function
4960 entries. The jumps to these functions are generated using a SH2A specific,
4961 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
4962 from GNU binutils version 2.7 or later for this attribute to work correctly.
4964 In an application, for a function being called once, this attribute
4965 saves at least 8 bytes of code; and if other successive calls are being
4966 made to the same function, it saves 2 bytes of code per each of these
4969 @item interrupt_handler
4970 @cindex @code{interrupt_handler} function attribute, SH
4971 Use this attribute to
4972 indicate that the specified function is an interrupt handler. The compiler
4973 generates function entry and exit sequences suitable for use in an
4974 interrupt handler when this attribute is present.
4976 @item nosave_low_regs
4977 @cindex @code{nosave_low_regs} function attribute, SH
4978 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
4979 function should not save and restore registers R0..R7. This can be used on SH3*
4980 and SH4* targets that have a second R0..R7 register bank for non-reentrant
4984 @cindex @code{renesas} function attribute, SH
4985 On SH targets this attribute specifies that the function or struct follows the
4989 @cindex @code{resbank} function attribute, SH
4990 On the SH2A target, this attribute enables the high-speed register
4991 saving and restoration using a register bank for @code{interrupt_handler}
4992 routines. Saving to the bank is performed automatically after the CPU
4993 accepts an interrupt that uses a register bank.
4995 The nineteen 32-bit registers comprising general register R0 to R14,
4996 control register GBR, and system registers MACH, MACL, and PR and the
4997 vector table address offset are saved into a register bank. Register
4998 banks are stacked in first-in last-out (FILO) sequence. Restoration
4999 from the bank is executed by issuing a RESBANK instruction.
5002 @cindex @code{sp_switch} function attribute, SH
5003 Use this attribute on the SH to indicate an @code{interrupt_handler}
5004 function should switch to an alternate stack. It expects a string
5005 argument that names a global variable holding the address of the
5010 void f () __attribute__ ((interrupt_handler,
5011 sp_switch ("alt_stack")));
5015 @cindex @code{trap_exit} function attribute, SH
5016 Use this attribute on the SH for an @code{interrupt_handler} to return using
5017 @code{trapa} instead of @code{rte}. This attribute expects an integer
5018 argument specifying the trap number to be used.
5021 @cindex @code{trapa_handler} function attribute, SH
5022 On SH targets this function attribute is similar to @code{interrupt_handler}
5023 but it does not save and restore all registers.
5026 @node SPU Function Attributes
5027 @subsection SPU Function Attributes
5029 These function attributes are supported by the SPU back end:
5033 @cindex @code{naked} function attribute, SPU
5034 This attribute allows the compiler to construct the
5035 requisite function declaration, while allowing the body of the
5036 function to be assembly code. The specified function will not have
5037 prologue/epilogue sequences generated by the compiler. Only basic
5038 @code{asm} statements can safely be included in naked functions
5039 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5040 basic @code{asm} and C code may appear to work, they cannot be
5041 depended upon to work reliably and are not supported.
5044 @node Symbian OS Function Attributes
5045 @subsection Symbian OS Function Attributes
5047 @xref{Microsoft Windows Function Attributes}, for discussion of the
5048 @code{dllexport} and @code{dllimport} attributes.
5050 @node Visium Function Attributes
5051 @subsection Visium Function Attributes
5053 These function attributes are supported by the Visium back end:
5057 @cindex @code{interrupt} function attribute, Visium
5058 Use this attribute to indicate
5059 that the specified function is an interrupt handler. The compiler generates
5060 function entry and exit sequences suitable for use in an interrupt handler
5061 when this attribute is present.
5064 @node x86 Function Attributes
5065 @subsection x86 Function Attributes
5067 These function attributes are supported by the x86 back end:
5071 @cindex @code{cdecl} function attribute, x86-32
5072 @cindex functions that pop the argument stack on x86-32
5074 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5075 assume that the calling function pops off the stack space used to
5076 pass arguments. This is
5077 useful to override the effects of the @option{-mrtd} switch.
5080 @cindex @code{fastcall} function attribute, x86-32
5081 @cindex functions that pop the argument stack on x86-32
5082 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5083 pass the first argument (if of integral type) in the register ECX and
5084 the second argument (if of integral type) in the register EDX@. Subsequent
5085 and other typed arguments are passed on the stack. The called function
5086 pops the arguments off the stack. If the number of arguments is variable all
5087 arguments are pushed on the stack.
5090 @cindex @code{thiscall} function attribute, x86-32
5091 @cindex functions that pop the argument stack on x86-32
5092 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5093 pass the first argument (if of integral type) in the register ECX.
5094 Subsequent and other typed arguments are passed on the stack. The called
5095 function pops the arguments off the stack.
5096 If the number of arguments is variable all arguments are pushed on the
5098 The @code{thiscall} attribute is intended for C++ non-static member functions.
5099 As a GCC extension, this calling convention can be used for C functions
5100 and for static member methods.
5104 @cindex @code{ms_abi} function attribute, x86
5105 @cindex @code{sysv_abi} function attribute, x86
5107 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5108 to indicate which calling convention should be used for a function. The
5109 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5110 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5111 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5112 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5114 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5115 requires the @option{-maccumulate-outgoing-args} option.
5117 @item callee_pop_aggregate_return (@var{number})
5118 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5120 On x86-32 targets, you can use this attribute to control how
5121 aggregates are returned in memory. If the caller is responsible for
5122 popping the hidden pointer together with the rest of the arguments, specify
5123 @var{number} equal to zero. If callee is responsible for popping the
5124 hidden pointer, specify @var{number} equal to one.
5126 The default x86-32 ABI assumes that the callee pops the
5127 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5128 the compiler assumes that the
5129 caller pops the stack for hidden pointer.
5131 @item ms_hook_prologue
5132 @cindex @code{ms_hook_prologue} function attribute, x86
5134 On 32-bit and 64-bit x86 targets, you can use
5135 this function attribute to make GCC generate the ``hot-patching'' function
5136 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5139 @item regparm (@var{number})
5140 @cindex @code{regparm} function attribute, x86
5141 @cindex functions that are passed arguments in registers on x86-32
5142 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5143 pass arguments number one to @var{number} if they are of integral type
5144 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5145 take a variable number of arguments continue to be passed all of their
5146 arguments on the stack.
5148 Beware that on some ELF systems this attribute is unsuitable for
5149 global functions in shared libraries with lazy binding (which is the
5150 default). Lazy binding sends the first call via resolving code in
5151 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5152 per the standard calling conventions. Solaris 8 is affected by this.
5153 Systems with the GNU C Library version 2.1 or higher
5154 and FreeBSD are believed to be
5155 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5156 disabled with the linker or the loader if desired, to avoid the
5160 @cindex @code{sseregparm} function attribute, x86
5161 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5162 causes the compiler to pass up to 3 floating-point arguments in
5163 SSE registers instead of on the stack. Functions that take a
5164 variable number of arguments continue to pass all of their
5165 floating-point arguments on the stack.
5167 @item force_align_arg_pointer
5168 @cindex @code{force_align_arg_pointer} function attribute, x86
5169 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5170 applied to individual function definitions, generating an alternate
5171 prologue and epilogue that realigns the run-time stack if necessary.
5172 This supports mixing legacy codes that run with a 4-byte aligned stack
5173 with modern codes that keep a 16-byte stack for SSE compatibility.
5176 @cindex @code{stdcall} function attribute, x86-32
5177 @cindex functions that pop the argument stack on x86-32
5178 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5179 assume that the called function pops off the stack space used to
5180 pass arguments, unless it takes a variable number of arguments.
5182 @item target (@var{options})
5183 @cindex @code{target} function attribute
5184 As discussed in @ref{Common Function Attributes}, this attribute
5185 allows specification of target-specific compilation options.
5187 On the x86, the following options are allowed:
5191 @cindex @code{target("abm")} function attribute, x86
5192 Enable/disable the generation of the advanced bit instructions.
5196 @cindex @code{target("aes")} function attribute, x86
5197 Enable/disable the generation of the AES instructions.
5200 @cindex @code{target("default")} function attribute, x86
5201 @xref{Function Multiversioning}, where it is used to specify the
5202 default function version.
5206 @cindex @code{target("mmx")} function attribute, x86
5207 Enable/disable the generation of the MMX instructions.
5211 @cindex @code{target("pclmul")} function attribute, x86
5212 Enable/disable the generation of the PCLMUL instructions.
5216 @cindex @code{target("popcnt")} function attribute, x86
5217 Enable/disable the generation of the POPCNT instruction.
5221 @cindex @code{target("sse")} function attribute, x86
5222 Enable/disable the generation of the SSE instructions.
5226 @cindex @code{target("sse2")} function attribute, x86
5227 Enable/disable the generation of the SSE2 instructions.
5231 @cindex @code{target("sse3")} function attribute, x86
5232 Enable/disable the generation of the SSE3 instructions.
5236 @cindex @code{target("sse4")} function attribute, x86
5237 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5242 @cindex @code{target("sse4.1")} function attribute, x86
5243 Enable/disable the generation of the sse4.1 instructions.
5247 @cindex @code{target("sse4.2")} function attribute, x86
5248 Enable/disable the generation of the sse4.2 instructions.
5252 @cindex @code{target("sse4a")} function attribute, x86
5253 Enable/disable the generation of the SSE4A instructions.
5257 @cindex @code{target("fma4")} function attribute, x86
5258 Enable/disable the generation of the FMA4 instructions.
5262 @cindex @code{target("xop")} function attribute, x86
5263 Enable/disable the generation of the XOP instructions.
5267 @cindex @code{target("lwp")} function attribute, x86
5268 Enable/disable the generation of the LWP instructions.
5272 @cindex @code{target("ssse3")} function attribute, x86
5273 Enable/disable the generation of the SSSE3 instructions.
5277 @cindex @code{target("cld")} function attribute, x86
5278 Enable/disable the generation of the CLD before string moves.
5280 @item fancy-math-387
5281 @itemx no-fancy-math-387
5282 @cindex @code{target("fancy-math-387")} function attribute, x86
5283 Enable/disable the generation of the @code{sin}, @code{cos}, and
5284 @code{sqrt} instructions on the 387 floating-point unit.
5287 @itemx no-fused-madd
5288 @cindex @code{target("fused-madd")} function attribute, x86
5289 Enable/disable the generation of the fused multiply/add instructions.
5293 @cindex @code{target("ieee-fp")} function attribute, x86
5294 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5296 @item inline-all-stringops
5297 @itemx no-inline-all-stringops
5298 @cindex @code{target("inline-all-stringops")} function attribute, x86
5299 Enable/disable inlining of string operations.
5301 @item inline-stringops-dynamically
5302 @itemx no-inline-stringops-dynamically
5303 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5304 Enable/disable the generation of the inline code to do small string
5305 operations and calling the library routines for large operations.
5307 @item align-stringops
5308 @itemx no-align-stringops
5309 @cindex @code{target("align-stringops")} function attribute, x86
5310 Do/do not align destination of inlined string operations.
5314 @cindex @code{target("recip")} function attribute, x86
5315 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5316 instructions followed an additional Newton-Raphson step instead of
5317 doing a floating-point division.
5319 @item arch=@var{ARCH}
5320 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5321 Specify the architecture to generate code for in compiling the function.
5323 @item tune=@var{TUNE}
5324 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5325 Specify the architecture to tune for in compiling the function.
5327 @item fpmath=@var{FPMATH}
5328 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5329 Specify which floating-point unit to use. You must specify the
5330 @code{target("fpmath=sse,387")} option as
5331 @code{target("fpmath=sse+387")} because the comma would separate
5335 On the x86, the inliner does not inline a
5336 function that has different target options than the caller, unless the
5337 callee has a subset of the target options of the caller. For example
5338 a function declared with @code{target("sse3")} can inline a function
5339 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5342 @node Xstormy16 Function Attributes
5343 @subsection Xstormy16 Function Attributes
5345 These function attributes are supported by the Xstormy16 back end:
5349 @cindex @code{interrupt} function attribute, Xstormy16
5350 Use this attribute to indicate
5351 that the specified function is an interrupt handler. The compiler generates
5352 function entry and exit sequences suitable for use in an interrupt handler
5353 when this attribute is present.
5356 @node Variable Attributes
5357 @section Specifying Attributes of Variables
5358 @cindex attribute of variables
5359 @cindex variable attributes
5361 The keyword @code{__attribute__} allows you to specify special
5362 attributes of variables or structure fields. This keyword is followed
5363 by an attribute specification inside double parentheses. Some
5364 attributes are currently defined generically for variables.
5365 Other attributes are defined for variables on particular target
5366 systems. Other attributes are available for functions
5367 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5368 enumerators (@pxref{Enumerator Attributes}), and for types
5369 (@pxref{Type Attributes}).
5370 Other front ends might define more attributes
5371 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5373 @xref{Attribute Syntax}, for details of the exact syntax for using
5377 * Common Variable Attributes::
5378 * AVR Variable Attributes::
5379 * Blackfin Variable Attributes::
5380 * H8/300 Variable Attributes::
5381 * IA-64 Variable Attributes::
5382 * M32R/D Variable Attributes::
5383 * MeP Variable Attributes::
5384 * Microsoft Windows Variable Attributes::
5385 * MSP430 Variable Attributes::
5386 * PowerPC Variable Attributes::
5387 * SPU Variable Attributes::
5388 * x86 Variable Attributes::
5389 * Xstormy16 Variable Attributes::
5392 @node Common Variable Attributes
5393 @subsection Common Variable Attributes
5395 The following attributes are supported on most targets.
5398 @cindex @code{aligned} variable attribute
5399 @item aligned (@var{alignment})
5400 This attribute specifies a minimum alignment for the variable or
5401 structure field, measured in bytes. For example, the declaration:
5404 int x __attribute__ ((aligned (16))) = 0;
5408 causes the compiler to allocate the global variable @code{x} on a
5409 16-byte boundary. On a 68040, this could be used in conjunction with
5410 an @code{asm} expression to access the @code{move16} instruction which
5411 requires 16-byte aligned operands.
5413 You can also specify the alignment of structure fields. For example, to
5414 create a double-word aligned @code{int} pair, you could write:
5417 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5421 This is an alternative to creating a union with a @code{double} member,
5422 which forces the union to be double-word aligned.
5424 As in the preceding examples, you can explicitly specify the alignment
5425 (in bytes) that you wish the compiler to use for a given variable or
5426 structure field. Alternatively, you can leave out the alignment factor
5427 and just ask the compiler to align a variable or field to the
5428 default alignment for the target architecture you are compiling for.
5429 The default alignment is sufficient for all scalar types, but may not be
5430 enough for all vector types on a target that supports vector operations.
5431 The default alignment is fixed for a particular target ABI.
5433 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5434 which is the largest alignment ever used for any data type on the
5435 target machine you are compiling for. For example, you could write:
5438 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5441 The compiler automatically sets the alignment for the declared
5442 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5443 often make copy operations more efficient, because the compiler can
5444 use whatever instructions copy the biggest chunks of memory when
5445 performing copies to or from the variables or fields that you have
5446 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5447 may change depending on command-line options.
5449 When used on a struct, or struct member, the @code{aligned} attribute can
5450 only increase the alignment; in order to decrease it, the @code{packed}
5451 attribute must be specified as well. When used as part of a typedef, the
5452 @code{aligned} attribute can both increase and decrease alignment, and
5453 specifying the @code{packed} attribute generates a warning.
5455 Note that the effectiveness of @code{aligned} attributes may be limited
5456 by inherent limitations in your linker. On many systems, the linker is
5457 only able to arrange for variables to be aligned up to a certain maximum
5458 alignment. (For some linkers, the maximum supported alignment may
5459 be very very small.) If your linker is only able to align variables
5460 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5461 in an @code{__attribute__} still only provides you with 8-byte
5462 alignment. See your linker documentation for further information.
5464 The @code{aligned} attribute can also be used for functions
5465 (@pxref{Common Function Attributes}.)
5467 @item cleanup (@var{cleanup_function})
5468 @cindex @code{cleanup} variable attribute
5469 The @code{cleanup} attribute runs a function when the variable goes
5470 out of scope. This attribute can only be applied to auto function
5471 scope variables; it may not be applied to parameters or variables
5472 with static storage duration. The function must take one parameter,
5473 a pointer to a type compatible with the variable. The return value
5474 of the function (if any) is ignored.
5476 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5477 is run during the stack unwinding that happens during the
5478 processing of the exception. Note that the @code{cleanup} attribute
5479 does not allow the exception to be caught, only to perform an action.
5480 It is undefined what happens if @var{cleanup_function} does not
5485 @cindex @code{common} variable attribute
5486 @cindex @code{nocommon} variable attribute
5489 The @code{common} attribute requests GCC to place a variable in
5490 ``common'' storage. The @code{nocommon} attribute requests the
5491 opposite---to allocate space for it directly.
5493 These attributes override the default chosen by the
5494 @option{-fno-common} and @option{-fcommon} flags respectively.
5497 @itemx deprecated (@var{msg})
5498 @cindex @code{deprecated} variable attribute
5499 The @code{deprecated} attribute results in a warning if the variable
5500 is used anywhere in the source file. This is useful when identifying
5501 variables that are expected to be removed in a future version of a
5502 program. The warning also includes the location of the declaration
5503 of the deprecated variable, to enable users to easily find further
5504 information about why the variable is deprecated, or what they should
5505 do instead. Note that the warning only occurs for uses:
5508 extern int old_var __attribute__ ((deprecated));
5510 int new_fn () @{ return old_var; @}
5514 results in a warning on line 3 but not line 2. The optional @var{msg}
5515 argument, which must be a string, is printed in the warning if
5518 The @code{deprecated} attribute can also be used for functions and
5519 types (@pxref{Common Function Attributes},
5520 @pxref{Common Type Attributes}).
5522 @item mode (@var{mode})
5523 @cindex @code{mode} variable attribute
5524 This attribute specifies the data type for the declaration---whichever
5525 type corresponds to the mode @var{mode}. This in effect lets you
5526 request an integer or floating-point type according to its width.
5528 You may also specify a mode of @code{byte} or @code{__byte__} to
5529 indicate the mode corresponding to a one-byte integer, @code{word} or
5530 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5531 or @code{__pointer__} for the mode used to represent pointers.
5534 @cindex @code{packed} variable attribute
5535 The @code{packed} attribute specifies that a variable or structure field
5536 should have the smallest possible alignment---one byte for a variable,
5537 and one bit for a field, unless you specify a larger value with the
5538 @code{aligned} attribute.
5540 Here is a structure in which the field @code{x} is packed, so that it
5541 immediately follows @code{a}:
5547 int x[2] __attribute__ ((packed));
5551 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5552 @code{packed} attribute on bit-fields of type @code{char}. This has
5553 been fixed in GCC 4.4 but the change can lead to differences in the
5554 structure layout. See the documentation of
5555 @option{-Wpacked-bitfield-compat} for more information.
5557 @item section ("@var{section-name}")
5558 @cindex @code{section} variable attribute
5559 Normally, the compiler places the objects it generates in sections like
5560 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5561 or you need certain particular variables to appear in special sections,
5562 for example to map to special hardware. The @code{section}
5563 attribute specifies that a variable (or function) lives in a particular
5564 section. For example, this small program uses several specific section names:
5567 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5568 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5569 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5570 int init_data __attribute__ ((section ("INITDATA")));
5574 /* @r{Initialize stack pointer} */
5575 init_sp (stack + sizeof (stack));
5577 /* @r{Initialize initialized data} */
5578 memcpy (&init_data, &data, &edata - &data);
5580 /* @r{Turn on the serial ports} */
5587 Use the @code{section} attribute with
5588 @emph{global} variables and not @emph{local} variables,
5589 as shown in the example.
5591 You may use the @code{section} attribute with initialized or
5592 uninitialized global variables but the linker requires
5593 each object be defined once, with the exception that uninitialized
5594 variables tentatively go in the @code{common} (or @code{bss}) section
5595 and can be multiply ``defined''. Using the @code{section} attribute
5596 changes what section the variable goes into and may cause the
5597 linker to issue an error if an uninitialized variable has multiple
5598 definitions. You can force a variable to be initialized with the
5599 @option{-fno-common} flag or the @code{nocommon} attribute.
5601 Some file formats do not support arbitrary sections so the @code{section}
5602 attribute is not available on all platforms.
5603 If you need to map the entire contents of a module to a particular
5604 section, consider using the facilities of the linker instead.
5606 @item tls_model ("@var{tls_model}")
5607 @cindex @code{tls_model} variable attribute
5608 The @code{tls_model} attribute sets thread-local storage model
5609 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5610 overriding @option{-ftls-model=} command-line switch on a per-variable
5612 The @var{tls_model} argument should be one of @code{global-dynamic},
5613 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5615 Not all targets support this attribute.
5618 @cindex @code{unused} variable attribute
5619 This attribute, attached to a variable, means that the variable is meant
5620 to be possibly unused. GCC does not produce a warning for this
5624 @cindex @code{used} variable attribute
5625 This attribute, attached to a variable with static storage, means that
5626 the variable must be emitted even if it appears that the variable is not
5629 When applied to a static data member of a C++ class template, the
5630 attribute also means that the member is instantiated if the
5631 class itself is instantiated.
5633 @item vector_size (@var{bytes})
5634 @cindex @code{vector_size} variable attribute
5635 This attribute specifies the vector size for the variable, measured in
5636 bytes. For example, the declaration:
5639 int foo __attribute__ ((vector_size (16)));
5643 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5644 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5645 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5647 This attribute is only applicable to integral and float scalars,
5648 although arrays, pointers, and function return values are allowed in
5649 conjunction with this construct.
5651 Aggregates with this attribute are invalid, even if they are of the same
5652 size as a corresponding scalar. For example, the declaration:
5655 struct S @{ int a; @};
5656 struct S __attribute__ ((vector_size (16))) foo;
5660 is invalid even if the size of the structure is the same as the size of
5664 @cindex @code{weak} variable attribute
5665 The @code{weak} attribute is described in
5666 @ref{Common Function Attributes}.
5670 @node AVR Variable Attributes
5671 @subsection AVR Variable Attributes
5675 @cindex @code{progmem} variable attribute, AVR
5676 The @code{progmem} attribute is used on the AVR to place read-only
5677 data in the non-volatile program memory (flash). The @code{progmem}
5678 attribute accomplishes this by putting respective variables into a
5679 section whose name starts with @code{.progmem}.
5681 This attribute works similar to the @code{section} attribute
5682 but adds additional checking. Notice that just like the
5683 @code{section} attribute, @code{progmem} affects the location
5684 of the data but not how this data is accessed.
5686 In order to read data located with the @code{progmem} attribute
5687 (inline) assembler must be used.
5689 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5690 #include <avr/pgmspace.h>
5692 /* Locate var in flash memory */
5693 const int var[2] PROGMEM = @{ 1, 2 @};
5695 int read_var (int i)
5697 /* Access var[] by accessor macro from avr/pgmspace.h */
5698 return (int) pgm_read_word (& var[i]);
5702 AVR is a Harvard architecture processor and data and read-only data
5703 normally resides in the data memory (RAM).
5705 See also the @ref{AVR Named Address Spaces} section for
5706 an alternate way to locate and access data in flash memory.
5709 @itemx io (@var{addr})
5710 @cindex @code{io} variable attribute, AVR
5711 Variables with the @code{io} attribute are used to address
5712 memory-mapped peripherals in the io address range.
5713 If an address is specified, the variable
5714 is assigned that address, and the value is interpreted as an
5715 address in the data address space.
5719 volatile int porta __attribute__((io (0x22)));
5722 The address specified in the address in the data address range.
5724 Otherwise, the variable it is not assigned an address, but the
5725 compiler will still use in/out instructions where applicable,
5726 assuming some other module assigns an address in the io address range.
5730 extern volatile int porta __attribute__((io));
5734 @itemx io_low (@var{addr})
5735 @cindex @code{io_low} variable attribute, AVR
5736 This is like the @code{io} attribute, but additionally it informs the
5737 compiler that the object lies in the lower half of the I/O area,
5738 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5742 @itemx address (@var{addr})
5743 @cindex @code{address} variable attribute, AVR
5744 Variables with the @code{address} attribute are used to address
5745 memory-mapped peripherals that may lie outside the io address range.
5748 volatile int porta __attribute__((address (0x600)));
5753 @node Blackfin Variable Attributes
5754 @subsection Blackfin Variable Attributes
5756 Three attributes are currently defined for the Blackfin.
5762 @cindex @code{l1_data} variable attribute, Blackfin
5763 @cindex @code{l1_data_A} variable attribute, Blackfin
5764 @cindex @code{l1_data_B} variable attribute, Blackfin
5765 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5766 Variables with @code{l1_data} attribute are put into the specific section
5767 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5768 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5769 attribute are put into the specific section named @code{.l1.data.B}.
5772 @cindex @code{l2} variable attribute, Blackfin
5773 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5774 Variables with @code{l2} attribute are put into the specific section
5775 named @code{.l2.data}.
5778 @node H8/300 Variable Attributes
5779 @subsection H8/300 Variable Attributes
5781 These variable attributes are available for H8/300 targets:
5785 @cindex @code{eightbit_data} variable attribute, H8/300
5786 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5787 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5788 variable should be placed into the eight-bit data section.
5789 The compiler generates more efficient code for certain operations
5790 on data in the eight-bit data area. Note the eight-bit data area is limited to
5793 You must use GAS and GLD from GNU binutils version 2.7 or later for
5794 this attribute to work correctly.
5797 @cindex @code{tiny_data} variable attribute, H8/300
5798 @cindex tiny data section on the H8/300H and H8S
5799 Use this attribute on the H8/300H and H8S to indicate that the specified
5800 variable should be placed into the tiny data section.
5801 The compiler generates more efficient code for loads and stores
5802 on data in the tiny data section. Note the tiny data area is limited to
5803 slightly under 32KB of data.
5807 @node IA-64 Variable Attributes
5808 @subsection IA-64 Variable Attributes
5810 The IA-64 back end supports the following variable attribute:
5813 @item model (@var{model-name})
5814 @cindex @code{model} variable attribute, IA-64
5816 On IA-64, use this attribute to set the addressability of an object.
5817 At present, the only supported identifier for @var{model-name} is
5818 @code{small}, indicating addressability via ``small'' (22-bit)
5819 addresses (so that their addresses can be loaded with the @code{addl}
5820 instruction). Caveat: such addressing is by definition not position
5821 independent and hence this attribute must not be used for objects
5822 defined by shared libraries.
5826 @node M32R/D Variable Attributes
5827 @subsection M32R/D Variable Attributes
5829 One attribute is currently defined for the M32R/D@.
5832 @item model (@var{model-name})
5833 @cindex @code{model-name} variable attribute, M32R/D
5834 @cindex variable addressability on the M32R/D
5835 Use this attribute on the M32R/D to set the addressability of an object.
5836 The identifier @var{model-name} is one of @code{small}, @code{medium},
5837 or @code{large}, representing each of the code models.
5839 Small model objects live in the lower 16MB of memory (so that their
5840 addresses can be loaded with the @code{ld24} instruction).
5842 Medium and large model objects may live anywhere in the 32-bit address space
5843 (the compiler generates @code{seth/add3} instructions to load their
5847 @node MeP Variable Attributes
5848 @subsection MeP Variable Attributes
5850 The MeP target has a number of addressing modes and busses. The
5851 @code{near} space spans the standard memory space's first 16 megabytes
5852 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5853 The @code{based} space is a 128-byte region in the memory space that
5854 is addressed relative to the @code{$tp} register. The @code{tiny}
5855 space is a 65536-byte region relative to the @code{$gp} register. In
5856 addition to these memory regions, the MeP target has a separate 16-bit
5857 control bus which is specified with @code{cb} attributes.
5862 @cindex @code{based} variable attribute, MeP
5863 Any variable with the @code{based} attribute is assigned to the
5864 @code{.based} section, and is accessed with relative to the
5865 @code{$tp} register.
5868 @cindex @code{tiny} variable attribute, MeP
5869 Likewise, the @code{tiny} attribute assigned variables to the
5870 @code{.tiny} section, relative to the @code{$gp} register.
5873 @cindex @code{near} variable attribute, MeP
5874 Variables with the @code{near} attribute are assumed to have addresses
5875 that fit in a 24-bit addressing mode. This is the default for large
5876 variables (@code{-mtiny=4} is the default) but this attribute can
5877 override @code{-mtiny=} for small variables, or override @code{-ml}.
5880 @cindex @code{far} variable attribute, MeP
5881 Variables with the @code{far} attribute are addressed using a full
5882 32-bit address. Since this covers the entire memory space, this
5883 allows modules to make no assumptions about where variables might be
5887 @cindex @code{io} variable attribute, MeP
5888 @itemx io (@var{addr})
5889 Variables with the @code{io} attribute are used to address
5890 memory-mapped peripherals. If an address is specified, the variable
5891 is assigned that address, else it is not assigned an address (it is
5892 assumed some other module assigns an address). Example:
5895 int timer_count __attribute__((io(0x123)));
5899 @itemx cb (@var{addr})
5900 @cindex @code{cb} variable attribute, MeP
5901 Variables with the @code{cb} attribute are used to access the control
5902 bus, using special instructions. @code{addr} indicates the control bus
5906 int cpu_clock __attribute__((cb(0x123)));
5911 @node Microsoft Windows Variable Attributes
5912 @subsection Microsoft Windows Variable Attributes
5914 You can use these attributes on Microsoft Windows targets.
5915 @ref{x86 Variable Attributes} for additional Windows compatibility
5916 attributes available on all x86 targets.
5921 @cindex @code{dllimport} variable attribute
5922 @cindex @code{dllexport} variable attribute
5923 The @code{dllimport} and @code{dllexport} attributes are described in
5924 @ref{Microsoft Windows Function Attributes}.
5927 @cindex @code{selectany} variable attribute
5928 The @code{selectany} attribute causes an initialized global variable to
5929 have link-once semantics. When multiple definitions of the variable are
5930 encountered by the linker, the first is selected and the remainder are
5931 discarded. Following usage by the Microsoft compiler, the linker is told
5932 @emph{not} to warn about size or content differences of the multiple
5935 Although the primary usage of this attribute is for POD types, the
5936 attribute can also be applied to global C++ objects that are initialized
5937 by a constructor. In this case, the static initialization and destruction
5938 code for the object is emitted in each translation defining the object,
5939 but the calls to the constructor and destructor are protected by a
5940 link-once guard variable.
5942 The @code{selectany} attribute is only available on Microsoft Windows
5943 targets. You can use @code{__declspec (selectany)} as a synonym for
5944 @code{__attribute__ ((selectany))} for compatibility with other
5948 @cindex @code{shared} variable attribute
5949 On Microsoft Windows, in addition to putting variable definitions in a named
5950 section, the section can also be shared among all running copies of an
5951 executable or DLL@. For example, this small program defines shared data
5952 by putting it in a named section @code{shared} and marking the section
5956 int foo __attribute__((section ("shared"), shared)) = 0;
5961 /* @r{Read and write foo. All running
5962 copies see the same value.} */
5968 You may only use the @code{shared} attribute along with @code{section}
5969 attribute with a fully-initialized global definition because of the way
5970 linkers work. See @code{section} attribute for more information.
5972 The @code{shared} attribute is only available on Microsoft Windows@.
5976 @node MSP430 Variable Attributes
5977 @subsection MSP430 Variable Attributes
5981 @cindex @code{noinit} MSP430 variable attribute
5982 Any data with the @code{noinit} attribute will not be initialised by
5983 the C runtime startup code, or the program loader. Not initialising
5984 data in this way can reduce program startup times.
5987 @cindex @code{persistent} MSP430 variable attribute
5988 Any variable with the @code{persistent} attribute will not be
5989 initialised by the C runtime startup code. Instead its value will be
5990 set once, when the application is loaded, and then never initialised
5991 again, even if the processor is reset or the program restarts.
5992 Persistent data is intended to be placed into FLASH RAM, where its
5993 value will be retained across resets. The linker script being used to
5994 create the application should ensure that persistent data is correctly
6000 @cindex @code{lower} memory region on the MSP430
6001 @cindex @code{upper} memory region on the MSP430
6002 @cindex @code{either} memory region on the MSP430
6003 These attributes are the same as the MSP430 function attributes of the
6004 same name. These attributes can be applied to both functions and
6008 @node PowerPC Variable Attributes
6009 @subsection PowerPC Variable Attributes
6011 Three attributes currently are defined for PowerPC configurations:
6012 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6014 @cindex @code{ms_struct} variable attribute, PowerPC
6015 @cindex @code{gcc_struct} variable attribute, PowerPC
6016 For full documentation of the struct attributes please see the
6017 documentation in @ref{x86 Variable Attributes}.
6019 @cindex @code{altivec} variable attribute, PowerPC
6020 For documentation of @code{altivec} attribute please see the
6021 documentation in @ref{PowerPC Type Attributes}.
6023 @node SPU Variable Attributes
6024 @subsection SPU Variable Attributes
6026 @cindex @code{spu_vector} variable attribute, SPU
6027 The SPU supports the @code{spu_vector} attribute for variables. For
6028 documentation of this attribute please see the documentation in
6029 @ref{SPU Type Attributes}.
6031 @node x86 Variable Attributes
6032 @subsection x86 Variable Attributes
6034 Two attributes are currently defined for x86 configurations:
6035 @code{ms_struct} and @code{gcc_struct}.
6040 @cindex @code{ms_struct} variable attribute, x86
6041 @cindex @code{gcc_struct} variable attribute, x86
6043 If @code{packed} is used on a structure, or if bit-fields are used,
6044 it may be that the Microsoft ABI lays out the structure differently
6045 than the way GCC normally does. Particularly when moving packed
6046 data between functions compiled with GCC and the native Microsoft compiler
6047 (either via function call or as data in a file), it may be necessary to access
6050 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
6051 compilers to match the native Microsoft compiler.
6053 The Microsoft structure layout algorithm is fairly simple with the exception
6054 of the bit-field packing.
6055 The padding and alignment of members of structures and whether a bit-field
6056 can straddle a storage-unit boundary are determine by these rules:
6059 @item Structure members are stored sequentially in the order in which they are
6060 declared: the first member has the lowest memory address and the last member
6063 @item Every data object has an alignment requirement. The alignment requirement
6064 for all data except structures, unions, and arrays is either the size of the
6065 object or the current packing size (specified with either the
6066 @code{aligned} attribute or the @code{pack} pragma),
6067 whichever is less. For structures, unions, and arrays,
6068 the alignment requirement is the largest alignment requirement of its members.
6069 Every object is allocated an offset so that:
6072 offset % alignment_requirement == 0
6075 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
6076 unit if the integral types are the same size and if the next bit-field fits
6077 into the current allocation unit without crossing the boundary imposed by the
6078 common alignment requirements of the bit-fields.
6081 MSVC interprets zero-length bit-fields in the following ways:
6084 @item If a zero-length bit-field is inserted between two bit-fields that
6085 are normally coalesced, the bit-fields are not coalesced.
6092 unsigned long bf_1 : 12;
6094 unsigned long bf_2 : 12;
6099 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
6100 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
6102 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
6103 alignment of the zero-length bit-field is greater than the member that follows it,
6104 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
6125 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
6126 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
6127 bit-field does not affect the alignment of @code{bar} or, as a result, the size
6130 Taking this into account, it is important to note the following:
6133 @item If a zero-length bit-field follows a normal bit-field, the type of the
6134 zero-length bit-field may affect the alignment of the structure as whole. For
6135 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
6136 normal bit-field, and is of type short.
6138 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
6139 still affect the alignment of the structure:
6150 Here, @code{t4} takes up 4 bytes.
6153 @item Zero-length bit-fields following non-bit-field members are ignored:
6165 Here, @code{t5} takes up 2 bytes.
6169 @node Xstormy16 Variable Attributes
6170 @subsection Xstormy16 Variable Attributes
6172 One attribute is currently defined for xstormy16 configurations:
6177 @cindex @code{below100} variable attribute, Xstormy16
6179 If a variable has the @code{below100} attribute (@code{BELOW100} is
6180 allowed also), GCC places the variable in the first 0x100 bytes of
6181 memory and use special opcodes to access it. Such variables are
6182 placed in either the @code{.bss_below100} section or the
6183 @code{.data_below100} section.
6187 @node Type Attributes
6188 @section Specifying Attributes of Types
6189 @cindex attribute of types
6190 @cindex type attributes
6192 The keyword @code{__attribute__} allows you to specify special
6193 attributes of types. Some type attributes apply only to @code{struct}
6194 and @code{union} types, while others can apply to any type defined
6195 via a @code{typedef} declaration. Other attributes are defined for
6196 functions (@pxref{Function Attributes}), labels (@pxref{Label
6197 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6198 variables (@pxref{Variable Attributes}).
6200 The @code{__attribute__} keyword is followed by an attribute specification
6201 inside double parentheses.
6203 You may specify type attributes in an enum, struct or union type
6204 declaration or definition by placing them immediately after the
6205 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6206 syntax is to place them just past the closing curly brace of the
6209 You can also include type attributes in a @code{typedef} declaration.
6210 @xref{Attribute Syntax}, for details of the exact syntax for using
6214 * Common Type Attributes::
6215 * ARM Type Attributes::
6216 * MeP Type Attributes::
6217 * PowerPC Type Attributes::
6218 * SPU Type Attributes::
6219 * x86 Type Attributes::
6222 @node Common Type Attributes
6223 @subsection Common Type Attributes
6225 The following type attributes are supported on most targets.
6228 @cindex @code{aligned} type attribute
6229 @item aligned (@var{alignment})
6230 This attribute specifies a minimum alignment (in bytes) for variables
6231 of the specified type. For example, the declarations:
6234 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6235 typedef int more_aligned_int __attribute__ ((aligned (8)));
6239 force the compiler to ensure (as far as it can) that each variable whose
6240 type is @code{struct S} or @code{more_aligned_int} is allocated and
6241 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6242 variables of type @code{struct S} aligned to 8-byte boundaries allows
6243 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6244 store) instructions when copying one variable of type @code{struct S} to
6245 another, thus improving run-time efficiency.
6247 Note that the alignment of any given @code{struct} or @code{union} type
6248 is required by the ISO C standard to be at least a perfect multiple of
6249 the lowest common multiple of the alignments of all of the members of
6250 the @code{struct} or @code{union} in question. This means that you @emph{can}
6251 effectively adjust the alignment of a @code{struct} or @code{union}
6252 type by attaching an @code{aligned} attribute to any one of the members
6253 of such a type, but the notation illustrated in the example above is a
6254 more obvious, intuitive, and readable way to request the compiler to
6255 adjust the alignment of an entire @code{struct} or @code{union} type.
6257 As in the preceding example, you can explicitly specify the alignment
6258 (in bytes) that you wish the compiler to use for a given @code{struct}
6259 or @code{union} type. Alternatively, you can leave out the alignment factor
6260 and just ask the compiler to align a type to the maximum
6261 useful alignment for the target machine you are compiling for. For
6262 example, you could write:
6265 struct S @{ short f[3]; @} __attribute__ ((aligned));
6268 Whenever you leave out the alignment factor in an @code{aligned}
6269 attribute specification, the compiler automatically sets the alignment
6270 for the type to the largest alignment that is ever used for any data
6271 type on the target machine you are compiling for. Doing this can often
6272 make copy operations more efficient, because the compiler can use
6273 whatever instructions copy the biggest chunks of memory when performing
6274 copies to or from the variables that have types that you have aligned
6277 In the example above, if the size of each @code{short} is 2 bytes, then
6278 the size of the entire @code{struct S} type is 6 bytes. The smallest
6279 power of two that is greater than or equal to that is 8, so the
6280 compiler sets the alignment for the entire @code{struct S} type to 8
6283 Note that although you can ask the compiler to select a time-efficient
6284 alignment for a given type and then declare only individual stand-alone
6285 objects of that type, the compiler's ability to select a time-efficient
6286 alignment is primarily useful only when you plan to create arrays of
6287 variables having the relevant (efficiently aligned) type. If you
6288 declare or use arrays of variables of an efficiently-aligned type, then
6289 it is likely that your program also does pointer arithmetic (or
6290 subscripting, which amounts to the same thing) on pointers to the
6291 relevant type, and the code that the compiler generates for these
6292 pointer arithmetic operations is often more efficient for
6293 efficiently-aligned types than for other types.
6295 The @code{aligned} attribute can only increase the alignment; but you
6296 can decrease it by specifying @code{packed} as well. See below.
6298 Note that the effectiveness of @code{aligned} attributes may be limited
6299 by inherent limitations in your linker. On many systems, the linker is
6300 only able to arrange for variables to be aligned up to a certain maximum
6301 alignment. (For some linkers, the maximum supported alignment may
6302 be very very small.) If your linker is only able to align variables
6303 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6304 in an @code{__attribute__} still only provides you with 8-byte
6305 alignment. See your linker documentation for further information.
6307 @opindex fshort-enums
6308 Specifying this attribute for @code{struct} and @code{union} types is
6309 equivalent to specifying the @code{packed} attribute on each of the
6310 structure or union members. Specifying the @option{-fshort-enums}
6311 flag on the line is equivalent to specifying the @code{packed}
6312 attribute on all @code{enum} definitions.
6314 In the following example @code{struct my_packed_struct}'s members are
6315 packed closely together, but the internal layout of its @code{s} member
6316 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6320 struct my_unpacked_struct
6326 struct __attribute__ ((__packed__)) my_packed_struct
6330 struct my_unpacked_struct s;
6334 You may only specify this attribute on the definition of an @code{enum},
6335 @code{struct} or @code{union}, not on a @code{typedef} that does not
6336 also define the enumerated type, structure or union.
6338 @item bnd_variable_size
6339 @cindex @code{bnd_variable_size} type attribute
6340 @cindex Pointer Bounds Checker attributes
6341 When applied to a structure field, this attribute tells Pointer
6342 Bounds Checker that the size of this field should not be computed
6343 using static type information. It may be used to mark variably-sized
6344 static array fields placed at the end of a structure.
6352 S *p = (S *)malloc (sizeof(S) + 100);
6353 p->data[10] = 0; //Bounds violation
6357 By using an attribute for the field we may avoid unwanted bound
6364 char data[1] __attribute__((bnd_variable_size));
6366 S *p = (S *)malloc (sizeof(S) + 100);
6367 p->data[10] = 0; //OK
6371 @itemx deprecated (@var{msg})
6372 @cindex @code{deprecated} type attribute
6373 The @code{deprecated} attribute results in a warning if the type
6374 is used anywhere in the source file. This is useful when identifying
6375 types that are expected to be removed in a future version of a program.
6376 If possible, the warning also includes the location of the declaration
6377 of the deprecated type, to enable users to easily find further
6378 information about why the type is deprecated, or what they should do
6379 instead. Note that the warnings only occur for uses and then only
6380 if the type is being applied to an identifier that itself is not being
6381 declared as deprecated.
6384 typedef int T1 __attribute__ ((deprecated));
6388 typedef T1 T3 __attribute__ ((deprecated));
6389 T3 z __attribute__ ((deprecated));
6393 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6394 warning is issued for line 4 because T2 is not explicitly
6395 deprecated. Line 5 has no warning because T3 is explicitly
6396 deprecated. Similarly for line 6. The optional @var{msg}
6397 argument, which must be a string, is printed in the warning if
6400 The @code{deprecated} attribute can also be used for functions and
6401 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6403 @item designated_init
6404 @cindex @code{designated_init} type attribute
6405 This attribute may only be applied to structure types. It indicates
6406 that any initialization of an object of this type must use designated
6407 initializers rather than positional initializers. The intent of this
6408 attribute is to allow the programmer to indicate that a structure's
6409 layout may change, and that therefore relying on positional
6410 initialization will result in future breakage.
6412 GCC emits warnings based on this attribute by default; use
6413 @option{-Wno-designated-init} to suppress them.
6416 @cindex @code{may_alias} type attribute
6417 Accesses through pointers to types with this attribute are not subject
6418 to type-based alias analysis, but are instead assumed to be able to alias
6419 any other type of objects.
6420 In the context of section 6.5 paragraph 7 of the C99 standard,
6421 an lvalue expression
6422 dereferencing such a pointer is treated like having a character type.
6423 See @option{-fstrict-aliasing} for more information on aliasing issues.
6424 This extension exists to support some vector APIs, in which pointers to
6425 one vector type are permitted to alias pointers to a different vector type.
6427 Note that an object of a type with this attribute does not have any
6433 typedef short __attribute__((__may_alias__)) short_a;
6439 short_a *b = (short_a *) &a;
6443 if (a == 0x12345678)
6451 If you replaced @code{short_a} with @code{short} in the variable
6452 declaration, the above program would abort when compiled with
6453 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6457 @cindex @code{packed} type attribute
6458 This attribute, attached to @code{struct} or @code{union} type
6459 definition, specifies that each member (other than zero-width bit-fields)
6460 of the structure or union is placed to minimize the memory required. When
6461 attached to an @code{enum} definition, it indicates that the smallest
6462 integral type should be used.
6464 @item scalar_storage_order ("@var{endianness}")
6465 @cindex @code{scalar_storage_order} type attribute
6466 When attached to a @code{union} or a @code{struct}, this attribute sets
6467 the storage order, aka endianness, of the scalar fields of the type, as
6468 well as the array fields whose component is scalar. The supported
6469 endianness are @code{big-endian} and @code{little-endian}. The attribute
6470 has no effects on fields which are themselves a @code{union}, a @code{struct}
6471 or an array whose component is a @code{union} or a @code{struct}, and it is
6472 possible to have fields with a different scalar storage order than the
6475 This attribute is supported only for targets that use a uniform default
6476 scalar storage order (fortunately, most of them), i.e. targets that store
6477 the scalars either all in big-endian or all in little-endian.
6479 Additional restrictions are enforced for types with the reverse scalar
6480 storage order with regard to the scalar storage order of the target:
6483 @item Taking the address of a scalar field of a @code{union} or a
6484 @code{struct} with reverse scalar storage order is not permitted and will
6486 @item Taking the address of an array field, whose component is scalar, of
6487 a @code{union} or a @code{struct} with reverse scalar storage order is
6488 permitted but will yield a warning, unless @option{-Wno-scalar-storage-order}
6490 @item Taking the address of a @code{union} or a @code{struct} with reverse
6491 scalar storage order is permitted.
6494 These restrictions exist because the storage order attribute is lost when
6495 the address of a scalar or the address of an array with scalar component
6496 is taken, so storing indirectly through this address will generally not work.
6497 The second case is nevertheless allowed to be able to perform a block copy
6498 from or to the array.
6500 @item transparent_union
6501 @cindex @code{transparent_union} type attribute
6503 This attribute, attached to a @code{union} type definition, indicates
6504 that any function parameter having that union type causes calls to that
6505 function to be treated in a special way.
6507 First, the argument corresponding to a transparent union type can be of
6508 any type in the union; no cast is required. Also, if the union contains
6509 a pointer type, the corresponding argument can be a null pointer
6510 constant or a void pointer expression; and if the union contains a void
6511 pointer type, the corresponding argument can be any pointer expression.
6512 If the union member type is a pointer, qualifiers like @code{const} on
6513 the referenced type must be respected, just as with normal pointer
6516 Second, the argument is passed to the function using the calling
6517 conventions of the first member of the transparent union, not the calling
6518 conventions of the union itself. All members of the union must have the
6519 same machine representation; this is necessary for this argument passing
6522 Transparent unions are designed for library functions that have multiple
6523 interfaces for compatibility reasons. For example, suppose the
6524 @code{wait} function must accept either a value of type @code{int *} to
6525 comply with POSIX, or a value of type @code{union wait *} to comply with
6526 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6527 @code{wait} would accept both kinds of arguments, but it would also
6528 accept any other pointer type and this would make argument type checking
6529 less useful. Instead, @code{<sys/wait.h>} might define the interface
6533 typedef union __attribute__ ((__transparent_union__))
6537 @} wait_status_ptr_t;
6539 pid_t wait (wait_status_ptr_t);
6543 This interface allows either @code{int *} or @code{union wait *}
6544 arguments to be passed, using the @code{int *} calling convention.
6545 The program can call @code{wait} with arguments of either type:
6548 int w1 () @{ int w; return wait (&w); @}
6549 int w2 () @{ union wait w; return wait (&w); @}
6553 With this interface, @code{wait}'s implementation might look like this:
6556 pid_t wait (wait_status_ptr_t p)
6558 return waitpid (-1, p.__ip, 0);
6563 @cindex @code{unused} type attribute
6564 When attached to a type (including a @code{union} or a @code{struct}),
6565 this attribute means that variables of that type are meant to appear
6566 possibly unused. GCC does not produce a warning for any variables of
6567 that type, even if the variable appears to do nothing. This is often
6568 the case with lock or thread classes, which are usually defined and then
6569 not referenced, but contain constructors and destructors that have
6570 nontrivial bookkeeping functions.
6573 @cindex @code{visibility} type attribute
6574 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6575 applied to class, struct, union and enum types. Unlike other type
6576 attributes, the attribute must appear between the initial keyword and
6577 the name of the type; it cannot appear after the body of the type.
6579 Note that the type visibility is applied to vague linkage entities
6580 associated with the class (vtable, typeinfo node, etc.). In
6581 particular, if a class is thrown as an exception in one shared object
6582 and caught in another, the class must have default visibility.
6583 Otherwise the two shared objects are unable to use the same
6584 typeinfo node and exception handling will break.
6588 To specify multiple attributes, separate them by commas within the
6589 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6592 @node ARM Type Attributes
6593 @subsection ARM Type Attributes
6595 @cindex @code{notshared} type attribute, ARM
6596 On those ARM targets that support @code{dllimport} (such as Symbian
6597 OS), you can use the @code{notshared} attribute to indicate that the
6598 virtual table and other similar data for a class should not be
6599 exported from a DLL@. For example:
6602 class __declspec(notshared) C @{
6604 __declspec(dllimport) C();
6608 __declspec(dllexport)
6613 In this code, @code{C::C} is exported from the current DLL, but the
6614 virtual table for @code{C} is not exported. (You can use
6615 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6616 most Symbian OS code uses @code{__declspec}.)
6618 @node MeP Type Attributes
6619 @subsection MeP Type Attributes
6621 @cindex @code{based} type attribute, MeP
6622 @cindex @code{tiny} type attribute, MeP
6623 @cindex @code{near} type attribute, MeP
6624 @cindex @code{far} type attribute, MeP
6625 Many of the MeP variable attributes may be applied to types as well.
6626 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6627 @code{far} attributes may be applied to either. The @code{io} and
6628 @code{cb} attributes may not be applied to types.
6630 @node PowerPC Type Attributes
6631 @subsection PowerPC Type Attributes
6633 Three attributes currently are defined for PowerPC configurations:
6634 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6636 @cindex @code{ms_struct} type attribute, PowerPC
6637 @cindex @code{gcc_struct} type attribute, PowerPC
6638 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6639 attributes please see the documentation in @ref{x86 Type Attributes}.
6641 @cindex @code{altivec} type attribute, PowerPC
6642 The @code{altivec} attribute allows one to declare AltiVec vector data
6643 types supported by the AltiVec Programming Interface Manual. The
6644 attribute requires an argument to specify one of three vector types:
6645 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6646 and @code{bool__} (always followed by unsigned).
6649 __attribute__((altivec(vector__)))
6650 __attribute__((altivec(pixel__))) unsigned short
6651 __attribute__((altivec(bool__))) unsigned
6654 These attributes mainly are intended to support the @code{__vector},
6655 @code{__pixel}, and @code{__bool} AltiVec keywords.
6657 @node SPU Type Attributes
6658 @subsection SPU Type Attributes
6660 @cindex @code{spu_vector} type attribute, SPU
6661 The SPU supports the @code{spu_vector} attribute for types. This attribute
6662 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6663 Language Extensions Specification. It is intended to support the
6664 @code{__vector} keyword.
6666 @node x86 Type Attributes
6667 @subsection x86 Type Attributes
6669 Two attributes are currently defined for x86 configurations:
6670 @code{ms_struct} and @code{gcc_struct}.
6676 @cindex @code{ms_struct} type attribute, x86
6677 @cindex @code{gcc_struct} type attribute, x86
6679 If @code{packed} is used on a structure, or if bit-fields are used
6680 it may be that the Microsoft ABI packs them differently
6681 than GCC normally packs them. Particularly when moving packed
6682 data between functions compiled with GCC and the native Microsoft compiler
6683 (either via function call or as data in a file), it may be necessary to access
6686 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
6687 compilers to match the native Microsoft compiler.
6690 @node Label Attributes
6691 @section Label Attributes
6692 @cindex Label Attributes
6694 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6695 details of the exact syntax for using attributes. Other attributes are
6696 available for functions (@pxref{Function Attributes}), variables
6697 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6698 and for types (@pxref{Type Attributes}).
6700 This example uses the @code{cold} label attribute to indicate the
6701 @code{ErrorHandling} branch is unlikely to be taken and that the
6702 @code{ErrorHandling} label is unused:
6706 asm goto ("some asm" : : : : NoError);
6708 /* This branch (the fall-through from the asm) is less commonly used */
6710 __attribute__((cold, unused)); /* Semi-colon is required here */
6715 printf("no error\n");
6721 @cindex @code{unused} label attribute
6722 This feature is intended for program-generated code that may contain
6723 unused labels, but which is compiled with @option{-Wall}. It is
6724 not normally appropriate to use in it human-written code, though it
6725 could be useful in cases where the code that jumps to the label is
6726 contained within an @code{#ifdef} conditional.
6729 @cindex @code{hot} label attribute
6730 The @code{hot} attribute on a label is used to inform the compiler that
6731 the path following the label is more likely than paths that are not so
6732 annotated. This attribute is used in cases where @code{__builtin_expect}
6733 cannot be used, for instance with computed goto or @code{asm goto}.
6736 @cindex @code{cold} label attribute
6737 The @code{cold} attribute on labels is used to inform the compiler that
6738 the path following the label is unlikely to be executed. This attribute
6739 is used in cases where @code{__builtin_expect} cannot be used, for instance
6740 with computed goto or @code{asm goto}.
6744 @node Enumerator Attributes
6745 @section Enumerator Attributes
6746 @cindex Enumerator Attributes
6748 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6749 details of the exact syntax for using attributes. Other attributes are
6750 available for functions (@pxref{Function Attributes}), variables
6751 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6752 and for types (@pxref{Type Attributes}).
6754 This example uses the @code{deprecated} enumerator attribute to indicate the
6755 @code{oldval} enumerator is deprecated:
6759 oldval __attribute__((deprecated)),
6772 @cindex @code{deprecated} enumerator attribute
6773 The @code{deprecated} attribute results in a warning if the enumerator
6774 is used anywhere in the source file. This is useful when identifying
6775 enumerators that are expected to be removed in a future version of a
6776 program. The warning also includes the location of the declaration
6777 of the deprecated enumerator, to enable users to easily find further
6778 information about why the enumerator is deprecated, or what they should
6779 do instead. Note that the warnings only occurs for uses.
6783 @node Attribute Syntax
6784 @section Attribute Syntax
6785 @cindex attribute syntax
6787 This section describes the syntax with which @code{__attribute__} may be
6788 used, and the constructs to which attribute specifiers bind, for the C
6789 language. Some details may vary for C++ and Objective-C@. Because of
6790 infelicities in the grammar for attributes, some forms described here
6791 may not be successfully parsed in all cases.
6793 There are some problems with the semantics of attributes in C++. For
6794 example, there are no manglings for attributes, although they may affect
6795 code generation, so problems may arise when attributed types are used in
6796 conjunction with templates or overloading. Similarly, @code{typeid}
6797 does not distinguish between types with different attributes. Support
6798 for attributes in C++ may be restricted in future to attributes on
6799 declarations only, but not on nested declarators.
6801 @xref{Function Attributes}, for details of the semantics of attributes
6802 applying to functions. @xref{Variable Attributes}, for details of the
6803 semantics of attributes applying to variables. @xref{Type Attributes},
6804 for details of the semantics of attributes applying to structure, union
6805 and enumerated types.
6806 @xref{Label Attributes}, for details of the semantics of attributes
6808 @xref{Enumerator Attributes}, for details of the semantics of attributes
6809 applying to enumerators.
6811 An @dfn{attribute specifier} is of the form
6812 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6813 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6814 each attribute is one of the following:
6818 Empty. Empty attributes are ignored.
6822 (which may be an identifier such as @code{unused}, or a reserved
6823 word such as @code{const}).
6826 An attribute name followed by a parenthesized list of
6827 parameters for the attribute.
6828 These parameters take one of the following forms:
6832 An identifier. For example, @code{mode} attributes use this form.
6835 An identifier followed by a comma and a non-empty comma-separated list
6836 of expressions. For example, @code{format} attributes use this form.
6839 A possibly empty comma-separated list of expressions. For example,
6840 @code{format_arg} attributes use this form with the list being a single
6841 integer constant expression, and @code{alias} attributes use this form
6842 with the list being a single string constant.
6846 An @dfn{attribute specifier list} is a sequence of one or more attribute
6847 specifiers, not separated by any other tokens.
6849 You may optionally specify attribute names with @samp{__}
6850 preceding and following the name.
6851 This allows you to use them in header files without
6852 being concerned about a possible macro of the same name. For example,
6853 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6856 @subsubheading Label Attributes
6858 In GNU C, an attribute specifier list may appear after the colon following a
6859 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6860 attributes on labels if the attribute specifier is immediately
6861 followed by a semicolon (i.e., the label applies to an empty
6862 statement). If the semicolon is missing, C++ label attributes are
6863 ambiguous, as it is permissible for a declaration, which could begin
6864 with an attribute list, to be labelled in C++. Declarations cannot be
6865 labelled in C90 or C99, so the ambiguity does not arise there.
6867 @subsubheading Enumerator Attributes
6869 In GNU C, an attribute specifier list may appear as part of an enumerator.
6870 The attribute goes after the enumeration constant, before @code{=}, if
6871 present. The optional attribute in the enumerator appertains to the
6872 enumeration constant. It is not possible to place the attribute after
6873 the constant expression, if present.
6875 @subsubheading Type Attributes
6877 An attribute specifier list may appear as part of a @code{struct},
6878 @code{union} or @code{enum} specifier. It may go either immediately
6879 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6880 the closing brace. The former syntax is preferred.
6881 Where attribute specifiers follow the closing brace, they are considered
6882 to relate to the structure, union or enumerated type defined, not to any
6883 enclosing declaration the type specifier appears in, and the type
6884 defined is not complete until after the attribute specifiers.
6885 @c Otherwise, there would be the following problems: a shift/reduce
6886 @c conflict between attributes binding the struct/union/enum and
6887 @c binding to the list of specifiers/qualifiers; and "aligned"
6888 @c attributes could use sizeof for the structure, but the size could be
6889 @c changed later by "packed" attributes.
6892 @subsubheading All other attributes
6894 Otherwise, an attribute specifier appears as part of a declaration,
6895 counting declarations of unnamed parameters and type names, and relates
6896 to that declaration (which may be nested in another declaration, for
6897 example in the case of a parameter declaration), or to a particular declarator
6898 within a declaration. Where an
6899 attribute specifier is applied to a parameter declared as a function or
6900 an array, it should apply to the function or array rather than the
6901 pointer to which the parameter is implicitly converted, but this is not
6902 yet correctly implemented.
6904 Any list of specifiers and qualifiers at the start of a declaration may
6905 contain attribute specifiers, whether or not such a list may in that
6906 context contain storage class specifiers. (Some attributes, however,
6907 are essentially in the nature of storage class specifiers, and only make
6908 sense where storage class specifiers may be used; for example,
6909 @code{section}.) There is one necessary limitation to this syntax: the
6910 first old-style parameter declaration in a function definition cannot
6911 begin with an attribute specifier, because such an attribute applies to
6912 the function instead by syntax described below (which, however, is not
6913 yet implemented in this case). In some other cases, attribute
6914 specifiers are permitted by this grammar but not yet supported by the
6915 compiler. All attribute specifiers in this place relate to the
6916 declaration as a whole. In the obsolescent usage where a type of
6917 @code{int} is implied by the absence of type specifiers, such a list of
6918 specifiers and qualifiers may be an attribute specifier list with no
6919 other specifiers or qualifiers.
6921 At present, the first parameter in a function prototype must have some
6922 type specifier that is not an attribute specifier; this resolves an
6923 ambiguity in the interpretation of @code{void f(int
6924 (__attribute__((foo)) x))}, but is subject to change. At present, if
6925 the parentheses of a function declarator contain only attributes then
6926 those attributes are ignored, rather than yielding an error or warning
6927 or implying a single parameter of type int, but this is subject to
6930 An attribute specifier list may appear immediately before a declarator
6931 (other than the first) in a comma-separated list of declarators in a
6932 declaration of more than one identifier using a single list of
6933 specifiers and qualifiers. Such attribute specifiers apply
6934 only to the identifier before whose declarator they appear. For
6938 __attribute__((noreturn)) void d0 (void),
6939 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6944 the @code{noreturn} attribute applies to all the functions
6945 declared; the @code{format} attribute only applies to @code{d1}.
6947 An attribute specifier list may appear immediately before the comma,
6948 @code{=} or semicolon terminating the declaration of an identifier other
6949 than a function definition. Such attribute specifiers apply
6950 to the declared object or function. Where an
6951 assembler name for an object or function is specified (@pxref{Asm
6952 Labels}), the attribute must follow the @code{asm}
6955 An attribute specifier list may, in future, be permitted to appear after
6956 the declarator in a function definition (before any old-style parameter
6957 declarations or the function body).
6959 Attribute specifiers may be mixed with type qualifiers appearing inside
6960 the @code{[]} of a parameter array declarator, in the C99 construct by
6961 which such qualifiers are applied to the pointer to which the array is
6962 implicitly converted. Such attribute specifiers apply to the pointer,
6963 not to the array, but at present this is not implemented and they are
6966 An attribute specifier list may appear at the start of a nested
6967 declarator. At present, there are some limitations in this usage: the
6968 attributes correctly apply to the declarator, but for most individual
6969 attributes the semantics this implies are not implemented.
6970 When attribute specifiers follow the @code{*} of a pointer
6971 declarator, they may be mixed with any type qualifiers present.
6972 The following describes the formal semantics of this syntax. It makes the
6973 most sense if you are familiar with the formal specification of
6974 declarators in the ISO C standard.
6976 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
6977 D1}, where @code{T} contains declaration specifiers that specify a type
6978 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
6979 contains an identifier @var{ident}. The type specified for @var{ident}
6980 for derived declarators whose type does not include an attribute
6981 specifier is as in the ISO C standard.
6983 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
6984 and the declaration @code{T D} specifies the type
6985 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6986 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6987 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
6989 If @code{D1} has the form @code{*
6990 @var{type-qualifier-and-attribute-specifier-list} D}, and the
6991 declaration @code{T D} specifies the type
6992 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6993 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6994 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7000 void (__attribute__((noreturn)) ****f) (void);
7004 specifies the type ``pointer to pointer to pointer to pointer to
7005 non-returning function returning @code{void}''. As another example,
7008 char *__attribute__((aligned(8))) *f;
7012 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7013 Note again that this does not work with most attributes; for example,
7014 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7015 is not yet supported.
7017 For compatibility with existing code written for compiler versions that
7018 did not implement attributes on nested declarators, some laxity is
7019 allowed in the placing of attributes. If an attribute that only applies
7020 to types is applied to a declaration, it is treated as applying to
7021 the type of that declaration. If an attribute that only applies to
7022 declarations is applied to the type of a declaration, it is treated
7023 as applying to that declaration; and, for compatibility with code
7024 placing the attributes immediately before the identifier declared, such
7025 an attribute applied to a function return type is treated as
7026 applying to the function type, and such an attribute applied to an array
7027 element type is treated as applying to the array type. If an
7028 attribute that only applies to function types is applied to a
7029 pointer-to-function type, it is treated as applying to the pointer
7030 target type; if such an attribute is applied to a function return type
7031 that is not a pointer-to-function type, it is treated as applying
7032 to the function type.
7034 @node Function Prototypes
7035 @section Prototypes and Old-Style Function Definitions
7036 @cindex function prototype declarations
7037 @cindex old-style function definitions
7038 @cindex promotion of formal parameters
7040 GNU C extends ISO C to allow a function prototype to override a later
7041 old-style non-prototype definition. Consider the following example:
7044 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7051 /* @r{Prototype function declaration.} */
7052 int isroot P((uid_t));
7054 /* @r{Old-style function definition.} */
7056 isroot (x) /* @r{??? lossage here ???} */
7063 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7064 not allow this example, because subword arguments in old-style
7065 non-prototype definitions are promoted. Therefore in this example the
7066 function definition's argument is really an @code{int}, which does not
7067 match the prototype argument type of @code{short}.
7069 This restriction of ISO C makes it hard to write code that is portable
7070 to traditional C compilers, because the programmer does not know
7071 whether the @code{uid_t} type is @code{short}, @code{int}, or
7072 @code{long}. Therefore, in cases like these GNU C allows a prototype
7073 to override a later old-style definition. More precisely, in GNU C, a
7074 function prototype argument type overrides the argument type specified
7075 by a later old-style definition if the former type is the same as the
7076 latter type before promotion. Thus in GNU C the above example is
7077 equivalent to the following:
7090 GNU C++ does not support old-style function definitions, so this
7091 extension is irrelevant.
7094 @section C++ Style Comments
7096 @cindex C++ comments
7097 @cindex comments, C++ style
7099 In GNU C, you may use C++ style comments, which start with @samp{//} and
7100 continue until the end of the line. Many other C implementations allow
7101 such comments, and they are included in the 1999 C standard. However,
7102 C++ style comments are not recognized if you specify an @option{-std}
7103 option specifying a version of ISO C before C99, or @option{-ansi}
7104 (equivalent to @option{-std=c90}).
7107 @section Dollar Signs in Identifier Names
7109 @cindex dollar signs in identifier names
7110 @cindex identifier names, dollar signs in
7112 In GNU C, you may normally use dollar signs in identifier names.
7113 This is because many traditional C implementations allow such identifiers.
7114 However, dollar signs in identifiers are not supported on a few target
7115 machines, typically because the target assembler does not allow them.
7117 @node Character Escapes
7118 @section The Character @key{ESC} in Constants
7120 You can use the sequence @samp{\e} in a string or character constant to
7121 stand for the ASCII character @key{ESC}.
7124 @section Inquiring on Alignment of Types or Variables
7126 @cindex type alignment
7127 @cindex variable alignment
7129 The keyword @code{__alignof__} allows you to inquire about how an object
7130 is aligned, or the minimum alignment usually required by a type. Its
7131 syntax is just like @code{sizeof}.
7133 For example, if the target machine requires a @code{double} value to be
7134 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7135 This is true on many RISC machines. On more traditional machine
7136 designs, @code{__alignof__ (double)} is 4 or even 2.
7138 Some machines never actually require alignment; they allow reference to any
7139 data type even at an odd address. For these machines, @code{__alignof__}
7140 reports the smallest alignment that GCC gives the data type, usually as
7141 mandated by the target ABI.
7143 If the operand of @code{__alignof__} is an lvalue rather than a type,
7144 its value is the required alignment for its type, taking into account
7145 any minimum alignment specified with GCC's @code{__attribute__}
7146 extension (@pxref{Variable Attributes}). For example, after this
7150 struct foo @{ int x; char y; @} foo1;
7154 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7155 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7157 It is an error to ask for the alignment of an incomplete type.
7161 @section An Inline Function is As Fast As a Macro
7162 @cindex inline functions
7163 @cindex integrating function code
7165 @cindex macros, inline alternative
7167 By declaring a function inline, you can direct GCC to make
7168 calls to that function faster. One way GCC can achieve this is to
7169 integrate that function's code into the code for its callers. This
7170 makes execution faster by eliminating the function-call overhead; in
7171 addition, if any of the actual argument values are constant, their
7172 known values may permit simplifications at compile time so that not
7173 all of the inline function's code needs to be included. The effect on
7174 code size is less predictable; object code may be larger or smaller
7175 with function inlining, depending on the particular case. You can
7176 also direct GCC to try to integrate all ``simple enough'' functions
7177 into their callers with the option @option{-finline-functions}.
7179 GCC implements three different semantics of declaring a function
7180 inline. One is available with @option{-std=gnu89} or
7181 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7182 on all inline declarations, another when
7183 @option{-std=c99}, @option{-std=c11},
7184 @option{-std=gnu99} or @option{-std=gnu11}
7185 (without @option{-fgnu89-inline}), and the third
7186 is used when compiling C++.
7188 To declare a function inline, use the @code{inline} keyword in its
7189 declaration, like this:
7199 If you are writing a header file to be included in ISO C90 programs, write
7200 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7202 The three types of inlining behave similarly in two important cases:
7203 when the @code{inline} keyword is used on a @code{static} function,
7204 like the example above, and when a function is first declared without
7205 using the @code{inline} keyword and then is defined with
7206 @code{inline}, like this:
7209 extern int inc (int *a);
7217 In both of these common cases, the program behaves the same as if you
7218 had not used the @code{inline} keyword, except for its speed.
7220 @cindex inline functions, omission of
7221 @opindex fkeep-inline-functions
7222 When a function is both inline and @code{static}, if all calls to the
7223 function are integrated into the caller, and the function's address is
7224 never used, then the function's own assembler code is never referenced.
7225 In this case, GCC does not actually output assembler code for the
7226 function, unless you specify the option @option{-fkeep-inline-functions}.
7227 If there is a nonintegrated call, then the function is compiled to
7228 assembler code as usual. The function must also be compiled as usual if
7229 the program refers to its address, because that can't be inlined.
7232 Note that certain usages in a function definition can make it unsuitable
7233 for inline substitution. Among these usages are: variadic functions,
7234 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7235 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7236 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7237 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7238 function marked @code{inline} could not be substituted, and gives the
7239 reason for the failure.
7241 @cindex automatic @code{inline} for C++ member fns
7242 @cindex @code{inline} automatic for C++ member fns
7243 @cindex member fns, automatically @code{inline}
7244 @cindex C++ member fns, automatically @code{inline}
7245 @opindex fno-default-inline
7246 As required by ISO C++, GCC considers member functions defined within
7247 the body of a class to be marked inline even if they are
7248 not explicitly declared with the @code{inline} keyword. You can
7249 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7250 Options,,Options Controlling C++ Dialect}.
7252 GCC does not inline any functions when not optimizing unless you specify
7253 the @samp{always_inline} attribute for the function, like this:
7256 /* @r{Prototype.} */
7257 inline void foo (const char) __attribute__((always_inline));
7260 The remainder of this section is specific to GNU C90 inlining.
7262 @cindex non-static inline function
7263 When an inline function is not @code{static}, then the compiler must assume
7264 that there may be calls from other source files; since a global symbol can
7265 be defined only once in any program, the function must not be defined in
7266 the other source files, so the calls therein cannot be integrated.
7267 Therefore, a non-@code{static} inline function is always compiled on its
7268 own in the usual fashion.
7270 If you specify both @code{inline} and @code{extern} in the function
7271 definition, then the definition is used only for inlining. In no case
7272 is the function compiled on its own, not even if you refer to its
7273 address explicitly. Such an address becomes an external reference, as
7274 if you had only declared the function, and had not defined it.
7276 This combination of @code{inline} and @code{extern} has almost the
7277 effect of a macro. The way to use it is to put a function definition in
7278 a header file with these keywords, and put another copy of the
7279 definition (lacking @code{inline} and @code{extern}) in a library file.
7280 The definition in the header file causes most calls to the function
7281 to be inlined. If any uses of the function remain, they refer to
7282 the single copy in the library.
7285 @section When is a Volatile Object Accessed?
7286 @cindex accessing volatiles
7287 @cindex volatile read
7288 @cindex volatile write
7289 @cindex volatile access
7291 C has the concept of volatile objects. These are normally accessed by
7292 pointers and used for accessing hardware or inter-thread
7293 communication. The standard encourages compilers to refrain from
7294 optimizations concerning accesses to volatile objects, but leaves it
7295 implementation defined as to what constitutes a volatile access. The
7296 minimum requirement is that at a sequence point all previous accesses
7297 to volatile objects have stabilized and no subsequent accesses have
7298 occurred. Thus an implementation is free to reorder and combine
7299 volatile accesses that occur between sequence points, but cannot do
7300 so for accesses across a sequence point. The use of volatile does
7301 not allow you to violate the restriction on updating objects multiple
7302 times between two sequence points.
7304 Accesses to non-volatile objects are not ordered with respect to
7305 volatile accesses. You cannot use a volatile object as a memory
7306 barrier to order a sequence of writes to non-volatile memory. For
7310 int *ptr = @var{something};
7312 *ptr = @var{something};
7317 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7318 that the write to @var{*ptr} occurs by the time the update
7319 of @var{vobj} happens. If you need this guarantee, you must use
7320 a stronger memory barrier such as:
7323 int *ptr = @var{something};
7325 *ptr = @var{something};
7326 asm volatile ("" : : : "memory");
7330 A scalar volatile object is read when it is accessed in a void context:
7333 volatile int *src = @var{somevalue};
7337 Such expressions are rvalues, and GCC implements this as a
7338 read of the volatile object being pointed to.
7340 Assignments are also expressions and have an rvalue. However when
7341 assigning to a scalar volatile, the volatile object is not reread,
7342 regardless of whether the assignment expression's rvalue is used or
7343 not. If the assignment's rvalue is used, the value is that assigned
7344 to the volatile object. For instance, there is no read of @var{vobj}
7345 in all the following cases:
7350 vobj = @var{something};
7351 obj = vobj = @var{something};
7352 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7353 obj = (@var{something}, vobj = @var{anotherthing});
7356 If you need to read the volatile object after an assignment has
7357 occurred, you must use a separate expression with an intervening
7360 As bit-fields are not individually addressable, volatile bit-fields may
7361 be implicitly read when written to, or when adjacent bit-fields are
7362 accessed. Bit-field operations may be optimized such that adjacent
7363 bit-fields are only partially accessed, if they straddle a storage unit
7364 boundary. For these reasons it is unwise to use volatile bit-fields to
7367 @node Using Assembly Language with C
7368 @section How to Use Inline Assembly Language in C Code
7369 @cindex @code{asm} keyword
7370 @cindex assembly language in C
7371 @cindex inline assembly language
7372 @cindex mixing assembly language and C
7374 The @code{asm} keyword allows you to embed assembler instructions
7375 within C code. GCC provides two forms of inline @code{asm}
7376 statements. A @dfn{basic @code{asm}} statement is one with no
7377 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7378 statement (@pxref{Extended Asm}) includes one or more operands.
7379 The extended form is preferred for mixing C and assembly language
7380 within a function, but to include assembly language at
7381 top level you must use basic @code{asm}.
7383 You can also use the @code{asm} keyword to override the assembler name
7384 for a C symbol, or to place a C variable in a specific register.
7387 * Basic Asm:: Inline assembler without operands.
7388 * Extended Asm:: Inline assembler with operands.
7389 * Constraints:: Constraints for @code{asm} operands
7390 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7391 * Explicit Register Variables:: Defining variables residing in specified
7393 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7397 @subsection Basic Asm --- Assembler Instructions Without Operands
7398 @cindex basic @code{asm}
7399 @cindex assembly language in C, basic
7401 A basic @code{asm} statement has the following syntax:
7404 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7407 The @code{asm} keyword is a GNU extension.
7408 When writing code that can be compiled with @option{-ansi} and the
7409 various @option{-std} options, use @code{__asm__} instead of
7410 @code{asm} (@pxref{Alternate Keywords}).
7412 @subsubheading Qualifiers
7415 The optional @code{volatile} qualifier has no effect.
7416 All basic @code{asm} blocks are implicitly volatile.
7419 @subsubheading Parameters
7422 @item AssemblerInstructions
7423 This is a literal string that specifies the assembler code. The string can
7424 contain any instructions recognized by the assembler, including directives.
7425 GCC does not parse the assembler instructions themselves and
7426 does not know what they mean or even whether they are valid assembler input.
7428 You may place multiple assembler instructions together in a single @code{asm}
7429 string, separated by the characters normally used in assembly code for the
7430 system. A combination that works in most places is a newline to break the
7431 line, plus a tab character (written as @samp{\n\t}).
7432 Some assemblers allow semicolons as a line separator. However,
7433 note that some assembler dialects use semicolons to start a comment.
7436 @subsubheading Remarks
7437 Using extended @code{asm} typically produces smaller, safer, and more
7438 efficient code, and in most cases it is a better solution than basic
7439 @code{asm}. However, there are two situations where only basic @code{asm}
7444 Extended @code{asm} statements have to be inside a C
7445 function, so to write inline assembly language at file scope (``top-level''),
7446 outside of C functions, you must use basic @code{asm}.
7447 You can use this technique to emit assembler directives,
7448 define assembly language macros that can be invoked elsewhere in the file,
7449 or write entire functions in assembly language.
7453 with the @code{naked} attribute also require basic @code{asm}
7454 (@pxref{Function Attributes}).
7457 Safely accessing C data and calling functions from basic @code{asm} is more
7458 complex than it may appear. To access C data, it is better to use extended
7461 Do not expect a sequence of @code{asm} statements to remain perfectly
7462 consecutive after compilation. If certain instructions need to remain
7463 consecutive in the output, put them in a single multi-instruction @code{asm}
7464 statement. Note that GCC's optimizers can move @code{asm} statements
7465 relative to other code, including across jumps.
7467 @code{asm} statements may not perform jumps into other @code{asm} statements.
7468 GCC does not know about these jumps, and therefore cannot take
7469 account of them when deciding how to optimize. Jumps from @code{asm} to C
7470 labels are only supported in extended @code{asm}.
7472 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7473 assembly code when optimizing. This can lead to unexpected duplicate
7474 symbol errors during compilation if your assembly code defines symbols or
7477 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7478 visibility of any symbols it references. This may result in GCC discarding
7479 those symbols as unreferenced.
7481 The compiler copies the assembler instructions in a basic @code{asm}
7482 verbatim to the assembly language output file, without
7483 processing dialects or any of the @samp{%} operators that are available with
7484 extended @code{asm}. This results in minor differences between basic
7485 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7486 registers you might use @samp{%eax} in basic @code{asm} and
7487 @samp{%%eax} in extended @code{asm}.
7489 On targets such as x86 that support multiple assembler dialects,
7490 all basic @code{asm} blocks use the assembler dialect specified by the
7491 @option{-masm} command-line option (@pxref{x86 Options}).
7492 Basic @code{asm} provides no
7493 mechanism to provide different assembler strings for different dialects.
7495 Here is an example of basic @code{asm} for i386:
7498 /* Note that this code will not compile with -masm=intel */
7499 #define DebugBreak() asm("int $3")
7503 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7504 @cindex extended @code{asm}
7505 @cindex assembly language in C, extended
7507 With extended @code{asm} you can read and write C variables from
7508 assembler and perform jumps from assembler code to C labels.
7509 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7510 the operand parameters after the assembler template:
7513 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7514 : @var{OutputOperands}
7515 @r{[} : @var{InputOperands}
7516 @r{[} : @var{Clobbers} @r{]} @r{]})
7518 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7520 : @var{InputOperands}
7525 The @code{asm} keyword is a GNU extension.
7526 When writing code that can be compiled with @option{-ansi} and the
7527 various @option{-std} options, use @code{__asm__} instead of
7528 @code{asm} (@pxref{Alternate Keywords}).
7530 @subsubheading Qualifiers
7534 The typical use of extended @code{asm} statements is to manipulate input
7535 values to produce output values. However, your @code{asm} statements may
7536 also produce side effects. If so, you may need to use the @code{volatile}
7537 qualifier to disable certain optimizations. @xref{Volatile}.
7540 This qualifier informs the compiler that the @code{asm} statement may
7541 perform a jump to one of the labels listed in the @var{GotoLabels}.
7545 @subsubheading Parameters
7547 @item AssemblerTemplate
7548 This is a literal string that is the template for the assembler code. It is a
7549 combination of fixed text and tokens that refer to the input, output,
7550 and goto parameters. @xref{AssemblerTemplate}.
7552 @item OutputOperands
7553 A comma-separated list of the C variables modified by the instructions in the
7554 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7557 A comma-separated list of C expressions read by the instructions in the
7558 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7561 A comma-separated list of registers or other values changed by the
7562 @var{AssemblerTemplate}, beyond those listed as outputs.
7563 An empty list is permitted. @xref{Clobbers}.
7566 When you are using the @code{goto} form of @code{asm}, this section contains
7567 the list of all C labels to which the code in the
7568 @var{AssemblerTemplate} may jump.
7571 @code{asm} statements may not perform jumps into other @code{asm} statements,
7572 only to the listed @var{GotoLabels}.
7573 GCC's optimizers do not know about other jumps; therefore they cannot take
7574 account of them when deciding how to optimize.
7577 The total number of input + output + goto operands is limited to 30.
7579 @subsubheading Remarks
7580 The @code{asm} statement allows you to include assembly instructions directly
7581 within C code. This may help you to maximize performance in time-sensitive
7582 code or to access assembly instructions that are not readily available to C
7585 Note that extended @code{asm} statements must be inside a function. Only
7586 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7587 Functions declared with the @code{naked} attribute also require basic
7588 @code{asm} (@pxref{Function Attributes}).
7590 While the uses of @code{asm} are many and varied, it may help to think of an
7591 @code{asm} statement as a series of low-level instructions that convert input
7592 parameters to output parameters. So a simple (if not particularly useful)
7593 example for i386 using @code{asm} might look like this:
7599 asm ("mov %1, %0\n\t"
7604 printf("%d\n", dst);
7607 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7610 @subsubsection Volatile
7611 @cindex volatile @code{asm}
7612 @cindex @code{asm} volatile
7614 GCC's optimizers sometimes discard @code{asm} statements if they determine
7615 there is no need for the output variables. Also, the optimizers may move
7616 code out of loops if they believe that the code will always return the same
7617 result (i.e. none of its input values change between calls). Using the
7618 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7619 that have no output operands, including @code{asm goto} statements,
7620 are implicitly volatile.
7622 This i386 code demonstrates a case that does not use (or require) the
7623 @code{volatile} qualifier. If it is performing assertion checking, this code
7624 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7625 unreferenced by any code. As a result, the optimizers can discard the
7626 @code{asm} statement, which in turn removes the need for the entire
7627 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7628 isn't needed you allow the optimizers to produce the most efficient code
7632 void DoCheck(uint32_t dwSomeValue)
7636 // Assumes dwSomeValue is not zero.
7646 The next example shows a case where the optimizers can recognize that the input
7647 (@code{dwSomeValue}) never changes during the execution of the function and can
7648 therefore move the @code{asm} outside the loop to produce more efficient code.
7649 Again, using @code{volatile} disables this type of optimization.
7652 void do_print(uint32_t dwSomeValue)
7656 for (uint32_t x=0; x < 5; x++)
7658 // Assumes dwSomeValue is not zero.
7664 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7669 The following example demonstrates a case where you need to use the
7670 @code{volatile} qualifier.
7671 It uses the x86 @code{rdtsc} instruction, which reads
7672 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7673 the optimizers might assume that the @code{asm} block will always return the
7674 same value and therefore optimize away the second call.
7679 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7680 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7681 "or %%rdx, %0" // 'Or' in the lower bits.
7686 printf("msr: %llx\n", msr);
7690 // Reprint the timestamp
7691 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7692 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7693 "or %%rdx, %0" // 'Or' in the lower bits.
7698 printf("msr: %llx\n", msr);
7701 GCC's optimizers do not treat this code like the non-volatile code in the
7702 earlier examples. They do not move it out of loops or omit it on the
7703 assumption that the result from a previous call is still valid.
7705 Note that the compiler can move even volatile @code{asm} instructions relative
7706 to other code, including across jump instructions. For example, on many
7707 targets there is a system register that controls the rounding mode of
7708 floating-point operations. Setting it with a volatile @code{asm}, as in the
7709 following PowerPC example, does not work reliably.
7712 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7716 The compiler may move the addition back before the volatile @code{asm}. To
7717 make it work as expected, add an artificial dependency to the @code{asm} by
7718 referencing a variable in the subsequent code, for example:
7721 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7725 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7726 assembly code when optimizing. This can lead to unexpected duplicate symbol
7727 errors during compilation if your asm code defines symbols or labels.
7729 (@pxref{AssemblerTemplate}) may help resolve this problem.
7731 @anchor{AssemblerTemplate}
7732 @subsubsection Assembler Template
7733 @cindex @code{asm} assembler template
7735 An assembler template is a literal string containing assembler instructions.
7736 The compiler replaces tokens in the template that refer
7737 to inputs, outputs, and goto labels,
7738 and then outputs the resulting string to the assembler. The
7739 string can contain any instructions recognized by the assembler, including
7740 directives. GCC does not parse the assembler instructions
7741 themselves and does not know what they mean or even whether they are valid
7742 assembler input. However, it does count the statements
7743 (@pxref{Size of an asm}).
7745 You may place multiple assembler instructions together in a single @code{asm}
7746 string, separated by the characters normally used in assembly code for the
7747 system. A combination that works in most places is a newline to break the
7748 line, plus a tab character to move to the instruction field (written as
7750 Some assemblers allow semicolons as a line separator. However, note
7751 that some assembler dialects use semicolons to start a comment.
7753 Do not expect a sequence of @code{asm} statements to remain perfectly
7754 consecutive after compilation, even when you are using the @code{volatile}
7755 qualifier. If certain instructions need to remain consecutive in the output,
7756 put them in a single multi-instruction asm statement.
7758 Accessing data from C programs without using input/output operands (such as
7759 by using global symbols directly from the assembler template) may not work as
7760 expected. Similarly, calling functions directly from an assembler template
7761 requires a detailed understanding of the target assembler and ABI.
7763 Since GCC does not parse the assembler template,
7764 it has no visibility of any
7765 symbols it references. This may result in GCC discarding those symbols as
7766 unreferenced unless they are also listed as input, output, or goto operands.
7768 @subsubheading Special format strings
7770 In addition to the tokens described by the input, output, and goto operands,
7771 these tokens have special meanings in the assembler template:
7775 Outputs a single @samp{%} into the assembler code.
7778 Outputs a number that is unique to each instance of the @code{asm}
7779 statement in the entire compilation. This option is useful when creating local
7780 labels and referring to them multiple times in a single template that
7781 generates multiple assembler instructions.
7786 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7787 into the assembler code. When unescaped, these characters have special
7788 meaning to indicate multiple assembler dialects, as described below.
7791 @subsubheading Multiple assembler dialects in @code{asm} templates
7793 On targets such as x86, GCC supports multiple assembler dialects.
7794 The @option{-masm} option controls which dialect GCC uses as its
7795 default for inline assembler. The target-specific documentation for the
7796 @option{-masm} option contains the list of supported dialects, as well as the
7797 default dialect if the option is not specified. This information may be
7798 important to understand, since assembler code that works correctly when
7799 compiled using one dialect will likely fail if compiled using another.
7802 If your code needs to support multiple assembler dialects (for example, if
7803 you are writing public headers that need to support a variety of compilation
7804 options), use constructs of this form:
7807 @{ dialect0 | dialect1 | dialect2... @}
7810 This construct outputs @code{dialect0}
7811 when using dialect #0 to compile the code,
7812 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7813 braces than the number of dialects the compiler supports, the construct
7816 For example, if an x86 compiler supports two dialects
7817 (@samp{att}, @samp{intel}), an
7818 assembler template such as this:
7821 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7825 is equivalent to one of
7828 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7829 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7832 Using that same compiler, this code:
7835 "xchg@{l@}\t@{%%@}ebx, %1"
7839 corresponds to either
7842 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7843 "xchg\tebx, %1" @r{/* intel dialect */}
7846 There is no support for nesting dialect alternatives.
7848 @anchor{OutputOperands}
7849 @subsubsection Output Operands
7850 @cindex @code{asm} output operands
7852 An @code{asm} statement has zero or more output operands indicating the names
7853 of C variables modified by the assembler code.
7855 In this i386 example, @code{old} (referred to in the template string as
7856 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7857 (@code{%2}) is an input:
7862 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7863 "sbb %0,%0" // Use the CF to calculate old.
7864 : "=r" (old), "+rm" (*Base)
7871 Operands are separated by commas. Each operand has this format:
7874 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7878 @item asmSymbolicName
7879 Specifies a symbolic name for the operand.
7880 Reference the name in the assembler template
7881 by enclosing it in square brackets
7882 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7883 that contains the definition. Any valid C variable name is acceptable,
7884 including names already defined in the surrounding code. No two operands
7885 within the same @code{asm} statement can use the same symbolic name.
7887 When not using an @var{asmSymbolicName}, use the (zero-based) position
7889 in the list of operands in the assembler template. For example if there are
7890 three output operands, use @samp{%0} in the template to refer to the first,
7891 @samp{%1} for the second, and @samp{%2} for the third.
7894 A string constant specifying constraints on the placement of the operand;
7895 @xref{Constraints}, for details.
7897 Output constraints must begin with either @samp{=} (a variable overwriting an
7898 existing value) or @samp{+} (when reading and writing). When using
7899 @samp{=}, do not assume the location contains the existing value
7900 on entry to the @code{asm}, except
7901 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7903 After the prefix, there must be one or more additional constraints
7904 (@pxref{Constraints}) that describe where the value resides. Common
7905 constraints include @samp{r} for register and @samp{m} for memory.
7906 When you list more than one possible location (for example, @code{"=rm"}),
7907 the compiler chooses the most efficient one based on the current context.
7908 If you list as many alternates as the @code{asm} statement allows, you permit
7909 the optimizers to produce the best possible code.
7910 If you must use a specific register, but your Machine Constraints do not
7911 provide sufficient control to select the specific register you want,
7912 local register variables may provide a solution (@pxref{Local Register
7916 Specifies a C lvalue expression to hold the output, typically a variable name.
7917 The enclosing parentheses are a required part of the syntax.
7921 When the compiler selects the registers to use to
7922 represent the output operands, it does not use any of the clobbered registers
7925 Output operand expressions must be lvalues. The compiler cannot check whether
7926 the operands have data types that are reasonable for the instruction being
7927 executed. For output expressions that are not directly addressable (for
7928 example a bit-field), the constraint must allow a register. In that case, GCC
7929 uses the register as the output of the @code{asm}, and then stores that
7930 register into the output.
7932 Operands using the @samp{+} constraint modifier count as two operands
7933 (that is, both as input and output) towards the total maximum of 30 operands
7934 per @code{asm} statement.
7936 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7937 operands that must not overlap an input. Otherwise,
7938 GCC may allocate the output operand in the same register as an unrelated
7939 input operand, on the assumption that the assembler code consumes its
7940 inputs before producing outputs. This assumption may be false if the assembler
7941 code actually consists of more than one instruction.
7943 The same problem can occur if one output parameter (@var{a}) allows a register
7944 constraint and another output parameter (@var{b}) allows a memory constraint.
7945 The code generated by GCC to access the memory address in @var{b} can contain
7946 registers which @emph{might} be shared by @var{a}, and GCC considers those
7947 registers to be inputs to the asm. As above, GCC assumes that such input
7948 registers are consumed before any outputs are written. This assumption may
7949 result in incorrect behavior if the asm writes to @var{a} before using
7950 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7951 ensures that modifying @var{a} does not affect the address referenced by
7952 @var{b}. Otherwise, the location of @var{b}
7953 is undefined if @var{a} is modified before using @var{b}.
7955 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7956 instead of simply @samp{%2}). Typically these qualifiers are hardware
7957 dependent. The list of supported modifiers for x86 is found at
7958 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7960 If the C code that follows the @code{asm} makes no use of any of the output
7961 operands, use @code{volatile} for the @code{asm} statement to prevent the
7962 optimizers from discarding the @code{asm} statement as unneeded
7963 (see @ref{Volatile}).
7965 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7966 references the first output operand as @code{%0} (were there a second, it
7967 would be @code{%1}, etc). The number of the first input operand is one greater
7968 than that of the last output operand. In this i386 example, that makes
7969 @code{Mask} referenced as @code{%1}:
7972 uint32_t Mask = 1234;
7981 That code overwrites the variable @code{Index} (@samp{=}),
7982 placing the value in a register (@samp{r}).
7983 Using the generic @samp{r} constraint instead of a constraint for a specific
7984 register allows the compiler to pick the register to use, which can result
7985 in more efficient code. This may not be possible if an assembler instruction
7986 requires a specific register.
7988 The following i386 example uses the @var{asmSymbolicName} syntax.
7990 same result as the code above, but some may consider it more readable or more
7991 maintainable since reordering index numbers is not necessary when adding or
7992 removing operands. The names @code{aIndex} and @code{aMask}
7993 are only used in this example to emphasize which
7994 names get used where.
7995 It is acceptable to reuse the names @code{Index} and @code{Mask}.
7998 uint32_t Mask = 1234;
8001 asm ("bsfl %[aMask], %[aIndex]"
8002 : [aIndex] "=r" (Index)
8003 : [aMask] "r" (Mask)
8007 Here are some more examples of output operands.
8014 asm ("mov %[e], %[d]"
8019 Here, @code{d} may either be in a register or in memory. Since the compiler
8020 might already have the current value of the @code{uint32_t} location
8021 pointed to by @code{e}
8022 in a register, you can enable it to choose the best location
8023 for @code{d} by specifying both constraints.
8025 @anchor{FlagOutputOperands}
8026 @subsection Flag Output Operands
8027 @cindex @code{asm} flag output operands
8029 Some targets have a special register that holds the ``flags'' for the
8030 result of an operation or comparison. Normally, the contents of that
8031 register are either unmodifed by the asm, or the asm is considered to
8032 clobber the contents.
8034 On some targets, a special form of output operand exists by which
8035 conditions in the flags register may be outputs of the asm. The set of
8036 conditions supported are target specific, but the general rule is that
8037 the output variable must be a scalar integer, and the value will be boolean.
8038 When supported, the target will define the preprocessor symbol
8039 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8041 Because of the special nature of the flag output operands, the constraint
8042 may not include alternatives.
8044 Most often, the target has only one flags register, and thus is an implied
8045 operand of many instructions. In this case, the operand should not be
8046 referenced within the assembler template via @code{%0} etc, as there's
8047 no corresponding text in the assembly language.
8051 The flag output constraints for the x86 family are of the form
8052 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8053 conditions defined in the ISA manual for @code{j@var{cc}} or
8058 ``above'' or unsigned greater than
8060 ``above or equal'' or unsigned greater than or equal
8062 ``below'' or unsigned less than
8064 ``below or equal'' or unsigned less than or equal
8069 ``equal'' or zero flag set
8073 signed greater than or equal
8077 signed less than or equal
8098 ``not'' @var{flag}, or inverted versions of those above
8103 @anchor{InputOperands}
8104 @subsubsection Input Operands
8105 @cindex @code{asm} input operands
8106 @cindex @code{asm} expressions
8108 Input operands make values from C variables and expressions available to the
8111 Operands are separated by commas. Each operand has this format:
8114 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8118 @item asmSymbolicName
8119 Specifies a symbolic name for the operand.
8120 Reference the name in the assembler template
8121 by enclosing it in square brackets
8122 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8123 that contains the definition. Any valid C variable name is acceptable,
8124 including names already defined in the surrounding code. No two operands
8125 within the same @code{asm} statement can use the same symbolic name.
8127 When not using an @var{asmSymbolicName}, use the (zero-based) position
8129 in the list of operands in the assembler template. For example if there are
8130 two output operands and three inputs,
8131 use @samp{%2} in the template to refer to the first input operand,
8132 @samp{%3} for the second, and @samp{%4} for the third.
8135 A string constant specifying constraints on the placement of the operand;
8136 @xref{Constraints}, for details.
8138 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8139 When you list more than one possible location (for example, @samp{"irm"}),
8140 the compiler chooses the most efficient one based on the current context.
8141 If you must use a specific register, but your Machine Constraints do not
8142 provide sufficient control to select the specific register you want,
8143 local register variables may provide a solution (@pxref{Local Register
8146 Input constraints can also be digits (for example, @code{"0"}). This indicates
8147 that the specified input must be in the same place as the output constraint
8148 at the (zero-based) index in the output constraint list.
8149 When using @var{asmSymbolicName} syntax for the output operands,
8150 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8153 This is the C variable or expression being passed to the @code{asm} statement
8154 as input. The enclosing parentheses are a required part of the syntax.
8158 When the compiler selects the registers to use to represent the input
8159 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8161 If there are no output operands but there are input operands, place two
8162 consecutive colons where the output operands would go:
8165 __asm__ ("some instructions"
8167 : "r" (Offset / 8));
8170 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8171 (except for inputs tied to outputs). The compiler assumes that on exit from
8172 the @code{asm} statement these operands contain the same values as they
8173 had before executing the statement.
8174 It is @emph{not} possible to use clobbers
8175 to inform the compiler that the values in these inputs are changing. One
8176 common work-around is to tie the changing input variable to an output variable
8177 that never gets used. Note, however, that if the code that follows the
8178 @code{asm} statement makes no use of any of the output operands, the GCC
8179 optimizers may discard the @code{asm} statement as unneeded
8180 (see @ref{Volatile}).
8182 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8183 instead of simply @samp{%2}). Typically these qualifiers are hardware
8184 dependent. The list of supported modifiers for x86 is found at
8185 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8187 In this example using the fictitious @code{combine} instruction, the
8188 constraint @code{"0"} for input operand 1 says that it must occupy the same
8189 location as output operand 0. Only input operands may use numbers in
8190 constraints, and they must each refer to an output operand. Only a number (or
8191 the symbolic assembler name) in the constraint can guarantee that one operand
8192 is in the same place as another. The mere fact that @code{foo} is the value of
8193 both operands is not enough to guarantee that they are in the same place in
8194 the generated assembler code.
8197 asm ("combine %2, %0"
8199 : "0" (foo), "g" (bar));
8202 Here is an example using symbolic names.
8205 asm ("cmoveq %1, %2, %[result]"
8206 : [result] "=r"(result)
8207 : "r" (test), "r" (new), "[result]" (old));
8211 @subsubsection Clobbers
8212 @cindex @code{asm} clobbers
8214 While the compiler is aware of changes to entries listed in the output
8215 operands, the inline @code{asm} code may modify more than just the outputs. For
8216 example, calculations may require additional registers, or the processor may
8217 overwrite a register as a side effect of a particular assembler instruction.
8218 In order to inform the compiler of these changes, list them in the clobber
8219 list. Clobber list items are either register names or the special clobbers
8220 (listed below). Each clobber list item is a string constant
8221 enclosed in double quotes and separated by commas.
8223 Clobber descriptions may not in any way overlap with an input or output
8224 operand. For example, you may not have an operand describing a register class
8225 with one member when listing that register in the clobber list. Variables
8226 declared to live in specific registers (@pxref{Explicit Register
8227 Variables}) and used
8228 as @code{asm} input or output operands must have no part mentioned in the
8229 clobber description. In particular, there is no way to specify that input
8230 operands get modified without also specifying them as output operands.
8232 When the compiler selects which registers to use to represent input and output
8233 operands, it does not use any of the clobbered registers. As a result,
8234 clobbered registers are available for any use in the assembler code.
8236 Here is a realistic example for the VAX showing the use of clobbered
8240 asm volatile ("movc3 %0, %1, %2"
8242 : "g" (from), "g" (to), "g" (count)
8243 : "r0", "r1", "r2", "r3", "r4", "r5");
8246 Also, there are two special clobber arguments:
8250 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8251 register. On some machines, GCC represents the condition codes as a specific
8252 hardware register; @code{"cc"} serves to name this register.
8253 On other machines, condition code handling is different,
8254 and specifying @code{"cc"} has no effect. But
8255 it is valid no matter what the target.
8258 The @code{"memory"} clobber tells the compiler that the assembly code
8260 reads or writes to items other than those listed in the input and output
8261 operands (for example, accessing the memory pointed to by one of the input
8262 parameters). To ensure memory contains correct values, GCC may need to flush
8263 specific register values to memory before executing the @code{asm}. Further,
8264 the compiler does not assume that any values read from memory before an
8265 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8267 Using the @code{"memory"} clobber effectively forms a read/write
8268 memory barrier for the compiler.
8270 Note that this clobber does not prevent the @emph{processor} from doing
8271 speculative reads past the @code{asm} statement. To prevent that, you need
8272 processor-specific fence instructions.
8274 Flushing registers to memory has performance implications and may be an issue
8275 for time-sensitive code. You can use a trick to avoid this if the size of
8276 the memory being accessed is known at compile time. For example, if accessing
8277 ten bytes of a string, use a memory input like:
8279 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8284 @subsubsection Goto Labels
8285 @cindex @code{asm} goto labels
8287 @code{asm goto} allows assembly code to jump to one or more C labels. The
8288 @var{GotoLabels} section in an @code{asm goto} statement contains
8290 list of all C labels to which the assembler code may jump. GCC assumes that
8291 @code{asm} execution falls through to the next statement (if this is not the
8292 case, consider using the @code{__builtin_unreachable} intrinsic after the
8293 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8294 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8297 An @code{asm goto} statement cannot have outputs.
8298 This is due to an internal restriction of
8299 the compiler: control transfer instructions cannot have outputs.
8300 If the assembler code does modify anything, use the @code{"memory"} clobber
8302 optimizers to flush all register values to memory and reload them if
8303 necessary after the @code{asm} statement.
8305 Also note that an @code{asm goto} statement is always implicitly
8306 considered volatile.
8308 To reference a label in the assembler template,
8309 prefix it with @samp{%l} (lowercase @samp{L}) followed
8310 by its (zero-based) position in @var{GotoLabels} plus the number of input
8311 operands. For example, if the @code{asm} has three inputs and references two
8312 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8314 Alternately, you can reference labels using the actual C label name enclosed
8315 in brackets. For example, to reference a label named @code{carry}, you can
8316 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8317 section when using this approach.
8319 Here is an example of @code{asm goto} for i386:
8326 : "r" (p1), "r" (p2)
8336 The following example shows an @code{asm goto} that uses a memory clobber.
8342 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8353 @anchor{x86Operandmodifiers}
8354 @subsubsection x86 Operand Modifiers
8356 References to input, output, and goto operands in the assembler template
8357 of extended @code{asm} statements can use
8358 modifiers to affect the way the operands are formatted in
8359 the code output to the assembler. For example, the
8360 following code uses the @samp{h} and @samp{b} modifiers for x86:
8364 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8368 These modifiers generate this assembler code:
8374 The rest of this discussion uses the following code for illustrative purposes.
8383 asm volatile goto ("some assembler instructions here"
8385 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8386 : /* No clobbers. */
8391 With no modifiers, this is what the output from the operands would be for the
8392 @samp{att} and @samp{intel} dialects of assembler:
8394 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8395 @headitem Operand @tab masm=att @tab masm=intel
8404 @tab @code{OFFSET FLAT:.L2}
8407 The table below shows the list of supported modifiers and their effects.
8409 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8410 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8412 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8417 @tab Print the QImode name of the register.
8422 @tab Print the QImode name for a ``high'' register.
8427 @tab Print the HImode name of the register.
8432 @tab Print the SImode name of the register.
8437 @tab Print the DImode name of the register.
8442 @tab Print the label name with no punctuation.
8447 @tab Require a constant operand and print the constant expression with no punctuation.
8453 @anchor{x86floatingpointasmoperands}
8454 @subsubsection x86 Floating-Point @code{asm} Operands
8456 On x86 targets, there are several rules on the usage of stack-like registers
8457 in the operands of an @code{asm}. These rules apply only to the operands
8458 that are stack-like registers:
8462 Given a set of input registers that die in an @code{asm}, it is
8463 necessary to know which are implicitly popped by the @code{asm}, and
8464 which must be explicitly popped by GCC@.
8466 An input register that is implicitly popped by the @code{asm} must be
8467 explicitly clobbered, unless it is constrained to match an
8471 For any input register that is implicitly popped by an @code{asm}, it is
8472 necessary to know how to adjust the stack to compensate for the pop.
8473 If any non-popped input is closer to the top of the reg-stack than
8474 the implicitly popped register, it would not be possible to know what the
8475 stack looked like---it's not clear how the rest of the stack ``slides
8478 All implicitly popped input registers must be closer to the top of
8479 the reg-stack than any input that is not implicitly popped.
8481 It is possible that if an input dies in an @code{asm}, the compiler might
8482 use the input register for an output reload. Consider this example:
8485 asm ("foo" : "=t" (a) : "f" (b));
8489 This code says that input @code{b} is not popped by the @code{asm}, and that
8490 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8491 deeper after the @code{asm} than it was before. But, it is possible that
8492 reload may think that it can use the same register for both the input and
8495 To prevent this from happening,
8496 if any input operand uses the @samp{f} constraint, all output register
8497 constraints must use the @samp{&} early-clobber modifier.
8499 The example above is correctly written as:
8502 asm ("foo" : "=&t" (a) : "f" (b));
8506 Some operands need to be in particular places on the stack. All
8507 output operands fall in this category---GCC has no other way to
8508 know which registers the outputs appear in unless you indicate
8509 this in the constraints.
8511 Output operands must specifically indicate which register an output
8512 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8513 constraints must select a class with a single register.
8516 Output operands may not be ``inserted'' between existing stack registers.
8517 Since no 387 opcode uses a read/write operand, all output operands
8518 are dead before the @code{asm}, and are pushed by the @code{asm}.
8519 It makes no sense to push anywhere but the top of the reg-stack.
8521 Output operands must start at the top of the reg-stack: output
8522 operands may not ``skip'' a register.
8525 Some @code{asm} statements may need extra stack space for internal
8526 calculations. This can be guaranteed by clobbering stack registers
8527 unrelated to the inputs and outputs.
8532 takes one input, which is internally popped, and produces two outputs.
8535 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8539 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8540 and replaces them with one output. The @code{st(1)} clobber is necessary
8541 for the compiler to know that @code{fyl2xp1} pops both inputs.
8544 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8552 @subsection Controlling Names Used in Assembler Code
8553 @cindex assembler names for identifiers
8554 @cindex names used in assembler code
8555 @cindex identifiers, names in assembler code
8557 You can specify the name to be used in the assembler code for a C
8558 function or variable by writing the @code{asm} (or @code{__asm__})
8559 keyword after the declarator.
8560 It is up to you to make sure that the assembler names you choose do not
8561 conflict with any other assembler symbols, or reference registers.
8563 @subsubheading Assembler names for data:
8565 This sample shows how to specify the assembler name for data:
8568 int foo asm ("myfoo") = 2;
8572 This specifies that the name to be used for the variable @code{foo} in
8573 the assembler code should be @samp{myfoo} rather than the usual
8576 On systems where an underscore is normally prepended to the name of a C
8577 variable, this feature allows you to define names for the
8578 linker that do not start with an underscore.
8580 GCC does not support using this feature with a non-static local variable
8581 since such variables do not have assembler names. If you are
8582 trying to put the variable in a particular register, see
8583 @ref{Explicit Register Variables}.
8585 @subsubheading Assembler names for functions:
8587 To specify the assembler name for functions, write a declaration for the
8588 function before its definition and put @code{asm} there, like this:
8591 int func (int x, int y) asm ("MYFUNC");
8593 int func (int x, int y)
8599 This specifies that the name to be used for the function @code{func} in
8600 the assembler code should be @code{MYFUNC}.
8602 @node Explicit Register Variables
8603 @subsection Variables in Specified Registers
8604 @anchor{Explicit Reg Vars}
8605 @cindex explicit register variables
8606 @cindex variables in specified registers
8607 @cindex specified registers
8609 GNU C allows you to associate specific hardware registers with C
8610 variables. In almost all cases, allowing the compiler to assign
8611 registers produces the best code. However under certain unusual
8612 circumstances, more precise control over the variable storage is
8615 Both global and local variables can be associated with a register. The
8616 consequences of performing this association are very different between
8617 the two, as explained in the sections below.
8620 * Global Register Variables:: Variables declared at global scope.
8621 * Local Register Variables:: Variables declared within a function.
8624 @node Global Register Variables
8625 @subsubsection Defining Global Register Variables
8626 @anchor{Global Reg Vars}
8627 @cindex global register variables
8628 @cindex registers, global variables in
8629 @cindex registers, global allocation
8631 You can define a global register variable and associate it with a specified
8635 register int *foo asm ("r12");
8639 Here @code{r12} is the name of the register that should be used. Note that
8640 this is the same syntax used for defining local register variables, but for
8641 a global variable the declaration appears outside a function. The
8642 @code{register} keyword is required, and cannot be combined with
8643 @code{static}. The register name must be a valid register name for the
8646 Registers are a scarce resource on most systems and allowing the
8647 compiler to manage their usage usually results in the best code. However,
8648 under special circumstances it can make sense to reserve some globally.
8649 For example this may be useful in programs such as programming language
8650 interpreters that have a couple of global variables that are accessed
8653 After defining a global register variable, for the current compilation
8657 @item The register is reserved entirely for this use, and will not be
8658 allocated for any other purpose.
8659 @item The register is not saved and restored by any functions.
8660 @item Stores into this register are never deleted even if they appear to be
8661 dead, but references may be deleted, moved or simplified.
8664 Note that these points @emph{only} apply to code that is compiled with the
8665 definition. The behavior of code that is merely linked in (for example
8666 code from libraries) is not affected.
8668 If you want to recompile source files that do not actually use your global
8669 register variable so they do not use the specified register for any other
8670 purpose, you need not actually add the global register declaration to
8671 their source code. It suffices to specify the compiler option
8672 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8675 @subsubheading Declaring the variable
8677 Global register variables can not have initial values, because an
8678 executable file has no means to supply initial contents for a register.
8680 When selecting a register, choose one that is normally saved and
8681 restored by function calls on your machine. This ensures that code
8682 which is unaware of this reservation (such as library routines) will
8683 restore it before returning.
8685 On machines with register windows, be sure to choose a global
8686 register that is not affected magically by the function call mechanism.
8688 @subsubheading Using the variable
8690 @cindex @code{qsort}, and global register variables
8691 When calling routines that are not aware of the reservation, be
8692 cautious if those routines call back into code which uses them. As an
8693 example, if you call the system library version of @code{qsort}, it may
8694 clobber your registers during execution, but (if you have selected
8695 appropriate registers) it will restore them before returning. However
8696 it will @emph{not} restore them before calling @code{qsort}'s comparison
8697 function. As a result, global values will not reliably be available to
8698 the comparison function unless the @code{qsort} function itself is rebuilt.
8700 Similarly, it is not safe to access the global register variables from signal
8701 handlers or from more than one thread of control. Unless you recompile
8702 them specially for the task at hand, the system library routines may
8703 temporarily use the register for other things.
8705 @cindex register variable after @code{longjmp}
8706 @cindex global register after @code{longjmp}
8707 @cindex value after @code{longjmp}
8710 On most machines, @code{longjmp} restores to each global register
8711 variable the value it had at the time of the @code{setjmp}. On some
8712 machines, however, @code{longjmp} does not change the value of global
8713 register variables. To be portable, the function that called @code{setjmp}
8714 should make other arrangements to save the values of the global register
8715 variables, and to restore them in a @code{longjmp}. This way, the same
8716 thing happens regardless of what @code{longjmp} does.
8718 Eventually there may be a way of asking the compiler to choose a register
8719 automatically, but first we need to figure out how it should choose and
8720 how to enable you to guide the choice. No solution is evident.
8722 @node Local Register Variables
8723 @subsubsection Specifying Registers for Local Variables
8724 @anchor{Local Reg Vars}
8725 @cindex local variables, specifying registers
8726 @cindex specifying registers for local variables
8727 @cindex registers for local variables
8729 You can define a local register variable and associate it with a specified
8733 register int *foo asm ("r12");
8737 Here @code{r12} is the name of the register that should be used. Note
8738 that this is the same syntax used for defining global register variables,
8739 but for a local variable the declaration appears within a function. The
8740 @code{register} keyword is required, and cannot be combined with
8741 @code{static}. The register name must be a valid register name for the
8744 As with global register variables, it is recommended that you choose
8745 a register that is normally saved and restored by function calls on your
8746 machine, so that calls to library routines will not clobber it.
8748 The only supported use for this feature is to specify registers
8749 for input and output operands when calling Extended @code{asm}
8750 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8751 particular machine don't provide sufficient control to select the desired
8752 register. To force an operand into a register, create a local variable
8753 and specify the register name after the variable's declaration. Then use
8754 the local variable for the @code{asm} operand and specify any constraint
8755 letter that matches the register:
8758 register int *p1 asm ("r0") = @dots{};
8759 register int *p2 asm ("r1") = @dots{};
8760 register int *result asm ("r0");
8761 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8764 @emph{Warning:} In the above example, be aware that a register (for example
8765 @code{r0}) can be call-clobbered by subsequent code, including function
8766 calls and library calls for arithmetic operators on other variables (for
8767 example the initialization of @code{p2}). In this case, use temporary
8768 variables for expressions between the register assignments:
8772 register int *p1 asm ("r0") = @dots{};
8773 register int *p2 asm ("r1") = t1;
8774 register int *result asm ("r0");
8775 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8778 Defining a register variable does not reserve the register. Other than
8779 when invoking the Extended @code{asm}, the contents of the specified
8780 register are not guaranteed. For this reason, the following uses
8781 are explicitly @emph{not} supported. If they appear to work, it is only
8782 happenstance, and may stop working as intended due to (seemingly)
8783 unrelated changes in surrounding code, or even minor changes in the
8784 optimization of a future version of gcc:
8787 @item Passing parameters to or from Basic @code{asm}
8788 @item Passing parameters to or from Extended @code{asm} without using input
8790 @item Passing parameters to or from routines written in assembler (or
8791 other languages) using non-standard calling conventions.
8794 Some developers use Local Register Variables in an attempt to improve
8795 gcc's allocation of registers, especially in large functions. In this
8796 case the register name is essentially a hint to the register allocator.
8797 While in some instances this can generate better code, improvements are
8798 subject to the whims of the allocator/optimizers. Since there are no
8799 guarantees that your improvements won't be lost, this usage of Local
8800 Register Variables is discouraged.
8802 On the MIPS platform, there is related use for local register variables
8803 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8804 Defining coprocessor specifics for MIPS targets, gccint,
8805 GNU Compiler Collection (GCC) Internals}).
8807 @node Size of an asm
8808 @subsection Size of an @code{asm}
8810 Some targets require that GCC track the size of each instruction used
8811 in order to generate correct code. Because the final length of the
8812 code produced by an @code{asm} statement is only known by the
8813 assembler, GCC must make an estimate as to how big it will be. It
8814 does this by counting the number of instructions in the pattern of the
8815 @code{asm} and multiplying that by the length of the longest
8816 instruction supported by that processor. (When working out the number
8817 of instructions, it assumes that any occurrence of a newline or of
8818 whatever statement separator character is supported by the assembler --
8819 typically @samp{;} --- indicates the end of an instruction.)
8821 Normally, GCC's estimate is adequate to ensure that correct
8822 code is generated, but it is possible to confuse the compiler if you use
8823 pseudo instructions or assembler macros that expand into multiple real
8824 instructions, or if you use assembler directives that expand to more
8825 space in the object file than is needed for a single instruction.
8826 If this happens then the assembler may produce a diagnostic saying that
8827 a label is unreachable.
8829 @node Alternate Keywords
8830 @section Alternate Keywords
8831 @cindex alternate keywords
8832 @cindex keywords, alternate
8834 @option{-ansi} and the various @option{-std} options disable certain
8835 keywords. This causes trouble when you want to use GNU C extensions, or
8836 a general-purpose header file that should be usable by all programs,
8837 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8838 @code{inline} are not available in programs compiled with
8839 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8840 program compiled with @option{-std=c99} or @option{-std=c11}). The
8842 @code{restrict} is only available when @option{-std=gnu99} (which will
8843 eventually be the default) or @option{-std=c99} (or the equivalent
8844 @option{-std=iso9899:1999}), or an option for a later standard
8847 The way to solve these problems is to put @samp{__} at the beginning and
8848 end of each problematical keyword. For example, use @code{__asm__}
8849 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8851 Other C compilers won't accept these alternative keywords; if you want to
8852 compile with another compiler, you can define the alternate keywords as
8853 macros to replace them with the customary keywords. It looks like this:
8861 @findex __extension__
8863 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8865 prevent such warnings within one expression by writing
8866 @code{__extension__} before the expression. @code{__extension__} has no
8867 effect aside from this.
8869 @node Incomplete Enums
8870 @section Incomplete @code{enum} Types
8872 You can define an @code{enum} tag without specifying its possible values.
8873 This results in an incomplete type, much like what you get if you write
8874 @code{struct foo} without describing the elements. A later declaration
8875 that does specify the possible values completes the type.
8877 You can't allocate variables or storage using the type while it is
8878 incomplete. However, you can work with pointers to that type.
8880 This extension may not be very useful, but it makes the handling of
8881 @code{enum} more consistent with the way @code{struct} and @code{union}
8884 This extension is not supported by GNU C++.
8886 @node Function Names
8887 @section Function Names as Strings
8888 @cindex @code{__func__} identifier
8889 @cindex @code{__FUNCTION__} identifier
8890 @cindex @code{__PRETTY_FUNCTION__} identifier
8892 GCC provides three magic variables that hold the name of the current
8893 function, as a string. The first of these is @code{__func__}, which
8894 is part of the C99 standard:
8896 The identifier @code{__func__} is implicitly declared by the translator
8897 as if, immediately following the opening brace of each function
8898 definition, the declaration
8901 static const char __func__[] = "function-name";
8905 appeared, where function-name is the name of the lexically-enclosing
8906 function. This name is the unadorned name of the function.
8908 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8909 backward compatibility with old versions of GCC.
8911 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8912 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8913 the type signature of the function as well as its bare name. For
8914 example, this program:
8918 extern int printf (char *, ...);
8925 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8926 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8944 __PRETTY_FUNCTION__ = void a::sub(int)
8947 These identifiers are variables, not preprocessor macros, and may not
8948 be used to initialize @code{char} arrays or be concatenated with other string
8951 @node Return Address
8952 @section Getting the Return or Frame Address of a Function
8954 These functions may be used to get information about the callers of a
8957 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8958 This function returns the return address of the current function, or of
8959 one of its callers. The @var{level} argument is number of frames to
8960 scan up the call stack. A value of @code{0} yields the return address
8961 of the current function, a value of @code{1} yields the return address
8962 of the caller of the current function, and so forth. When inlining
8963 the expected behavior is that the function returns the address of
8964 the function that is returned to. To work around this behavior use
8965 the @code{noinline} function attribute.
8967 The @var{level} argument must be a constant integer.
8969 On some machines it may be impossible to determine the return address of
8970 any function other than the current one; in such cases, or when the top
8971 of the stack has been reached, this function returns @code{0} or a
8972 random value. In addition, @code{__builtin_frame_address} may be used
8973 to determine if the top of the stack has been reached.
8975 Additional post-processing of the returned value may be needed, see
8976 @code{__builtin_extract_return_addr}.
8978 Calling this function with a nonzero argument can have unpredictable
8979 effects, including crashing the calling program. As a result, calls
8980 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8981 option is in effect. Such calls should only be made in debugging
8985 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
8986 The address as returned by @code{__builtin_return_address} may have to be fed
8987 through this function to get the actual encoded address. For example, on the
8988 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
8989 platforms an offset has to be added for the true next instruction to be
8992 If no fixup is needed, this function simply passes through @var{addr}.
8995 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
8996 This function does the reverse of @code{__builtin_extract_return_addr}.
8999 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9000 This function is similar to @code{__builtin_return_address}, but it
9001 returns the address of the function frame rather than the return address
9002 of the function. Calling @code{__builtin_frame_address} with a value of
9003 @code{0} yields the frame address of the current function, a value of
9004 @code{1} yields the frame address of the caller of the current function,
9007 The frame is the area on the stack that holds local variables and saved
9008 registers. The frame address is normally the address of the first word
9009 pushed on to the stack by the function. However, the exact definition
9010 depends upon the processor and the calling convention. If the processor
9011 has a dedicated frame pointer register, and the function has a frame,
9012 then @code{__builtin_frame_address} returns the value of the frame
9015 On some machines it may be impossible to determine the frame address of
9016 any function other than the current one; in such cases, or when the top
9017 of the stack has been reached, this function returns @code{0} if
9018 the first frame pointer is properly initialized by the startup code.
9020 Calling this function with a nonzero argument can have unpredictable
9021 effects, including crashing the calling program. As a result, calls
9022 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9023 option is in effect. Such calls should only be made in debugging
9027 @node Vector Extensions
9028 @section Using Vector Instructions through Built-in Functions
9030 On some targets, the instruction set contains SIMD vector instructions which
9031 operate on multiple values contained in one large register at the same time.
9032 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9035 The first step in using these extensions is to provide the necessary data
9036 types. This should be done using an appropriate @code{typedef}:
9039 typedef int v4si __attribute__ ((vector_size (16)));
9043 The @code{int} type specifies the base type, while the attribute specifies
9044 the vector size for the variable, measured in bytes. For example, the
9045 declaration above causes the compiler to set the mode for the @code{v4si}
9046 type to be 16 bytes wide and divided into @code{int} sized units. For
9047 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9048 corresponding mode of @code{foo} is @acronym{V4SI}.
9050 The @code{vector_size} attribute is only applicable to integral and
9051 float scalars, although arrays, pointers, and function return values
9052 are allowed in conjunction with this construct. Only sizes that are
9053 a power of two are currently allowed.
9055 All the basic integer types can be used as base types, both as signed
9056 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9057 @code{long long}. In addition, @code{float} and @code{double} can be
9058 used to build floating-point vector types.
9060 Specifying a combination that is not valid for the current architecture
9061 causes GCC to synthesize the instructions using a narrower mode.
9062 For example, if you specify a variable of type @code{V4SI} and your
9063 architecture does not allow for this specific SIMD type, GCC
9064 produces code that uses 4 @code{SIs}.
9066 The types defined in this manner can be used with a subset of normal C
9067 operations. Currently, GCC allows using the following operators
9068 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9070 The operations behave like C++ @code{valarrays}. Addition is defined as
9071 the addition of the corresponding elements of the operands. For
9072 example, in the code below, each of the 4 elements in @var{a} is
9073 added to the corresponding 4 elements in @var{b} and the resulting
9074 vector is stored in @var{c}.
9077 typedef int v4si __attribute__ ((vector_size (16)));
9084 Subtraction, multiplication, division, and the logical operations
9085 operate in a similar manner. Likewise, the result of using the unary
9086 minus or complement operators on a vector type is a vector whose
9087 elements are the negative or complemented values of the corresponding
9088 elements in the operand.
9090 It is possible to use shifting operators @code{<<}, @code{>>} on
9091 integer-type vectors. The operation is defined as following: @code{@{a0,
9092 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9093 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9096 For convenience, it is allowed to use a binary vector operation
9097 where one operand is a scalar. In that case the compiler transforms
9098 the scalar operand into a vector where each element is the scalar from
9099 the operation. The transformation happens only if the scalar could be
9100 safely converted to the vector-element type.
9101 Consider the following code.
9104 typedef int v4si __attribute__ ((vector_size (16)));
9109 a = b + 1; /* a = b + @{1,1,1,1@}; */
9110 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9112 a = l + a; /* Error, cannot convert long to int. */
9115 Vectors can be subscripted as if the vector were an array with
9116 the same number of elements and base type. Out of bound accesses
9117 invoke undefined behavior at run time. Warnings for out of bound
9118 accesses for vector subscription can be enabled with
9119 @option{-Warray-bounds}.
9121 Vector comparison is supported with standard comparison
9122 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9123 vector expressions of integer-type or real-type. Comparison between
9124 integer-type vectors and real-type vectors are not supported. The
9125 result of the comparison is a vector of the same width and number of
9126 elements as the comparison operands with a signed integral element
9129 Vectors are compared element-wise producing 0 when comparison is false
9130 and -1 (constant of the appropriate type where all bits are set)
9131 otherwise. Consider the following example.
9134 typedef int v4si __attribute__ ((vector_size (16)));
9136 v4si a = @{1,2,3,4@};
9137 v4si b = @{3,2,1,4@};
9140 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9141 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9144 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9145 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9146 integer vector with the same number of elements of the same size as @code{b}
9147 and @code{c}, computes all three arguments and creates a vector
9148 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9149 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9150 As in the case of binary operations, this syntax is also accepted when
9151 one of @code{b} or @code{c} is a scalar that is then transformed into a
9152 vector. If both @code{b} and @code{c} are scalars and the type of
9153 @code{true?b:c} has the same size as the element type of @code{a}, then
9154 @code{b} and @code{c} are converted to a vector type whose elements have
9155 this type and with the same number of elements as @code{a}.
9157 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9158 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9159 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9160 For mixed operations between a scalar @code{s} and a vector @code{v},
9161 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9162 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9164 Vector shuffling is available using functions
9165 @code{__builtin_shuffle (vec, mask)} and
9166 @code{__builtin_shuffle (vec0, vec1, mask)}.
9167 Both functions construct a permutation of elements from one or two
9168 vectors and return a vector of the same type as the input vector(s).
9169 The @var{mask} is an integral vector with the same width (@var{W})
9170 and element count (@var{N}) as the output vector.
9172 The elements of the input vectors are numbered in memory ordering of
9173 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9174 elements of @var{mask} are considered modulo @var{N} in the single-operand
9175 case and modulo @math{2*@var{N}} in the two-operand case.
9177 Consider the following example,
9180 typedef int v4si __attribute__ ((vector_size (16)));
9182 v4si a = @{1,2,3,4@};
9183 v4si b = @{5,6,7,8@};
9184 v4si mask1 = @{0,1,1,3@};
9185 v4si mask2 = @{0,4,2,5@};
9188 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9189 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9192 Note that @code{__builtin_shuffle} is intentionally semantically
9193 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9195 You can declare variables and use them in function calls and returns, as
9196 well as in assignments and some casts. You can specify a vector type as
9197 a return type for a function. Vector types can also be used as function
9198 arguments. It is possible to cast from one vector type to another,
9199 provided they are of the same size (in fact, you can also cast vectors
9200 to and from other datatypes of the same size).
9202 You cannot operate between vectors of different lengths or different
9203 signedness without a cast.
9206 @section Support for @code{offsetof}
9207 @findex __builtin_offsetof
9209 GCC implements for both C and C++ a syntactic extension to implement
9210 the @code{offsetof} macro.
9214 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9216 offsetof_member_designator:
9218 | offsetof_member_designator "." @code{identifier}
9219 | offsetof_member_designator "[" @code{expr} "]"
9222 This extension is sufficient such that
9225 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9229 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9230 may be dependent. In either case, @var{member} may consist of a single
9231 identifier, or a sequence of member accesses and array references.
9233 @node __sync Builtins
9234 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9236 The following built-in functions
9237 are intended to be compatible with those described
9238 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9239 section 7.4. As such, they depart from normal GCC practice by not using
9240 the @samp{__builtin_} prefix and also by being overloaded so that they
9241 work on multiple types.
9243 The definition given in the Intel documentation allows only for the use of
9244 the types @code{int}, @code{long}, @code{long long} or their unsigned
9245 counterparts. GCC allows any integral scalar or pointer type that is
9246 1, 2, 4 or 8 bytes in length.
9248 These functions are implemented in terms of the @samp{__atomic}
9249 builtins (@pxref{__atomic Builtins}). They should not be used for new
9250 code which should use the @samp{__atomic} builtins instead.
9252 Not all operations are supported by all target processors. If a particular
9253 operation cannot be implemented on the target processor, a warning is
9254 generated and a call to an external function is generated. The external
9255 function carries the same name as the built-in version,
9256 with an additional suffix
9257 @samp{_@var{n}} where @var{n} is the size of the data type.
9259 @c ??? Should we have a mechanism to suppress this warning? This is almost
9260 @c useful for implementing the operation under the control of an external
9263 In most cases, these built-in functions are considered a @dfn{full barrier}.
9265 no memory operand is moved across the operation, either forward or
9266 backward. Further, instructions are issued as necessary to prevent the
9267 processor from speculating loads across the operation and from queuing stores
9268 after the operation.
9270 All of the routines are described in the Intel documentation to take
9271 ``an optional list of variables protected by the memory barrier''. It's
9272 not clear what is meant by that; it could mean that @emph{only} the
9273 listed variables are protected, or it could mean a list of additional
9274 variables to be protected. The list is ignored by GCC which treats it as
9275 empty. GCC interprets an empty list as meaning that all globally
9276 accessible variables should be protected.
9279 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9280 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9281 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9282 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9283 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9284 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9285 @findex __sync_fetch_and_add
9286 @findex __sync_fetch_and_sub
9287 @findex __sync_fetch_and_or
9288 @findex __sync_fetch_and_and
9289 @findex __sync_fetch_and_xor
9290 @findex __sync_fetch_and_nand
9291 These built-in functions perform the operation suggested by the name, and
9292 returns the value that had previously been in memory. That is,
9295 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9296 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9299 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9300 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9302 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9303 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9304 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9305 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9306 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9307 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9308 @findex __sync_add_and_fetch
9309 @findex __sync_sub_and_fetch
9310 @findex __sync_or_and_fetch
9311 @findex __sync_and_and_fetch
9312 @findex __sync_xor_and_fetch
9313 @findex __sync_nand_and_fetch
9314 These built-in functions perform the operation suggested by the name, and
9315 return the new value. That is,
9318 @{ *ptr @var{op}= value; return *ptr; @}
9319 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9322 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9323 as @code{*ptr = ~(*ptr & value)} instead of
9324 @code{*ptr = ~*ptr & value}.
9326 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9327 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9328 @findex __sync_bool_compare_and_swap
9329 @findex __sync_val_compare_and_swap
9330 These built-in functions perform an atomic compare and swap.
9331 That is, if the current
9332 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9335 The ``bool'' version returns true if the comparison is successful and
9336 @var{newval} is written. The ``val'' version returns the contents
9337 of @code{*@var{ptr}} before the operation.
9339 @item __sync_synchronize (...)
9340 @findex __sync_synchronize
9341 This built-in function issues a full memory barrier.
9343 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9344 @findex __sync_lock_test_and_set
9345 This built-in function, as described by Intel, is not a traditional test-and-set
9346 operation, but rather an atomic exchange operation. It writes @var{value}
9347 into @code{*@var{ptr}}, and returns the previous contents of
9350 Many targets have only minimal support for such locks, and do not support
9351 a full exchange operation. In this case, a target may support reduced
9352 functionality here by which the @emph{only} valid value to store is the
9353 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9354 is implementation defined.
9356 This built-in function is not a full barrier,
9357 but rather an @dfn{acquire barrier}.
9358 This means that references after the operation cannot move to (or be
9359 speculated to) before the operation, but previous memory stores may not
9360 be globally visible yet, and previous memory loads may not yet be
9363 @item void __sync_lock_release (@var{type} *ptr, ...)
9364 @findex __sync_lock_release
9365 This built-in function releases the lock acquired by
9366 @code{__sync_lock_test_and_set}.
9367 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9369 This built-in function is not a full barrier,
9370 but rather a @dfn{release barrier}.
9371 This means that all previous memory stores are globally visible, and all
9372 previous memory loads have been satisfied, but following memory reads
9373 are not prevented from being speculated to before the barrier.
9376 @node __atomic Builtins
9377 @section Built-in Functions for Memory Model Aware Atomic Operations
9379 The following built-in functions approximately match the requirements
9380 for the C++11 memory model. They are all
9381 identified by being prefixed with @samp{__atomic} and most are
9382 overloaded so that they work with multiple types.
9384 These functions are intended to replace the legacy @samp{__sync}
9385 builtins. The main difference is that the memory order that is requested
9386 is a parameter to the functions. New code should always use the
9387 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9389 Note that the @samp{__atomic} builtins assume that programs will
9390 conform to the C++11 memory model. In particular, they assume
9391 that programs are free of data races. See the C++11 standard for
9392 detailed requirements.
9394 The @samp{__atomic} builtins can be used with any integral scalar or
9395 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9396 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9397 supported by the architecture.
9399 The four non-arithmetic functions (load, store, exchange, and
9400 compare_exchange) all have a generic version as well. This generic
9401 version works on any data type. It uses the lock-free built-in function
9402 if the specific data type size makes that possible; otherwise, an
9403 external call is left to be resolved at run time. This external call is
9404 the same format with the addition of a @samp{size_t} parameter inserted
9405 as the first parameter indicating the size of the object being pointed to.
9406 All objects must be the same size.
9408 There are 6 different memory orders that can be specified. These map
9409 to the C++11 memory orders with the same names, see the C++11 standard
9410 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9411 on atomic synchronization} for detailed definitions. Individual
9412 targets may also support additional memory orders for use on specific
9413 architectures. Refer to the target documentation for details of
9416 An atomic operation can both constrain code motion and
9417 be mapped to hardware instructions for synchronization between threads
9418 (e.g., a fence). To which extent this happens is controlled by the
9419 memory orders, which are listed here in approximately ascending order of
9420 strength. The description of each memory order is only meant to roughly
9421 illustrate the effects and is not a specification; see the C++11
9422 memory model for precise semantics.
9425 @item __ATOMIC_RELAXED
9426 Implies no inter-thread ordering constraints.
9427 @item __ATOMIC_CONSUME
9428 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9429 memory order because of a deficiency in C++11's semantics for
9430 @code{memory_order_consume}.
9431 @item __ATOMIC_ACQUIRE
9432 Creates an inter-thread happens-before constraint from the release (or
9433 stronger) semantic store to this acquire load. Can prevent hoisting
9434 of code to before the operation.
9435 @item __ATOMIC_RELEASE
9436 Creates an inter-thread happens-before constraint to acquire (or stronger)
9437 semantic loads that read from this release store. Can prevent sinking
9438 of code to after the operation.
9439 @item __ATOMIC_ACQ_REL
9440 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9441 @code{__ATOMIC_RELEASE}.
9442 @item __ATOMIC_SEQ_CST
9443 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9446 Note that in the C++11 memory model, @emph{fences} (e.g.,
9447 @samp{__atomic_thread_fence}) take effect in combination with other
9448 atomic operations on specific memory locations (e.g., atomic loads);
9449 operations on specific memory locations do not necessarily affect other
9450 operations in the same way.
9452 Target architectures are encouraged to provide their own patterns for
9453 each of the atomic built-in functions. If no target is provided, the original
9454 non-memory model set of @samp{__sync} atomic built-in functions are
9455 used, along with any required synchronization fences surrounding it in
9456 order to achieve the proper behavior. Execution in this case is subject
9457 to the same restrictions as those built-in functions.
9459 If there is no pattern or mechanism to provide a lock-free instruction
9460 sequence, a call is made to an external routine with the same parameters
9461 to be resolved at run time.
9463 When implementing patterns for these built-in functions, the memory order
9464 parameter can be ignored as long as the pattern implements the most
9465 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9466 orders execute correctly with this memory order but they may not execute as
9467 efficiently as they could with a more appropriate implementation of the
9468 relaxed requirements.
9470 Note that the C++11 standard allows for the memory order parameter to be
9471 determined at run time rather than at compile time. These built-in
9472 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9473 than invoke a runtime library call or inline a switch statement. This is
9474 standard compliant, safe, and the simplest approach for now.
9476 The memory order parameter is a signed int, but only the lower 16 bits are
9477 reserved for the memory order. The remainder of the signed int is reserved
9478 for target use and should be 0. Use of the predefined atomic values
9479 ensures proper usage.
9481 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9482 This built-in function implements an atomic load operation. It returns the
9483 contents of @code{*@var{ptr}}.
9485 The valid memory order variants are
9486 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9487 and @code{__ATOMIC_CONSUME}.
9491 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9492 This is the generic version of an atomic load. It returns the
9493 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9497 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9498 This built-in function implements an atomic store operation. It writes
9499 @code{@var{val}} into @code{*@var{ptr}}.
9501 The valid memory order variants are
9502 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9506 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9507 This is the generic version of an atomic store. It stores the value
9508 of @code{*@var{val}} into @code{*@var{ptr}}.
9512 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9513 This built-in function implements an atomic exchange operation. It writes
9514 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9517 The valid memory order variants are
9518 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9519 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9523 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9524 This is the generic version of an atomic exchange. It stores the
9525 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9526 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9530 @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)
9531 This built-in function implements an atomic compare and exchange operation.
9532 This compares the contents of @code{*@var{ptr}} with the contents of
9533 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9534 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9535 equal, the operation is a @emph{read} and the current contents of
9536 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
9537 for weak compare_exchange, and false for the strong variation. Many targets
9538 only offer the strong variation and ignore the parameter. When in doubt, use
9539 the strong variation.
9541 True is returned if @var{desired} is written into
9542 @code{*@var{ptr}} and the operation is considered to conform to the
9543 memory order specified by @var{success_memorder}. There are no
9544 restrictions on what memory order can be used here.
9546 False is returned otherwise, and the operation is considered to conform
9547 to @var{failure_memorder}. This memory order cannot be
9548 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9549 stronger order than that specified by @var{success_memorder}.
9553 @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)
9554 This built-in function implements the generic version of
9555 @code{__atomic_compare_exchange}. The function is virtually identical to
9556 @code{__atomic_compare_exchange_n}, except the desired value is also a
9561 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9562 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9563 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9564 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9565 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9566 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9567 These built-in functions perform the operation suggested by the name, and
9568 return the result of the operation. That is,
9571 @{ *ptr @var{op}= val; return *ptr; @}
9574 All memory orders are valid.
9578 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9579 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9580 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9581 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9582 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9583 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9584 These built-in functions perform the operation suggested by the name, and
9585 return the value that had previously been in @code{*@var{ptr}}. That is,
9588 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9591 All memory orders are valid.
9595 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9597 This built-in function performs an atomic test-and-set operation on
9598 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9599 defined nonzero ``set'' value and the return value is @code{true} if and only
9600 if the previous contents were ``set''.
9601 It should be only used for operands of type @code{bool} or @code{char}. For
9602 other types only part of the value may be set.
9604 All memory orders are valid.
9608 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9610 This built-in function performs an atomic clear operation on
9611 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9612 It should be only used for operands of type @code{bool} or @code{char} and
9613 in conjunction with @code{__atomic_test_and_set}.
9614 For other types it may only clear partially. If the type is not @code{bool}
9615 prefer using @code{__atomic_store}.
9617 The valid memory order variants are
9618 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9619 @code{__ATOMIC_RELEASE}.
9623 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9625 This built-in function acts as a synchronization fence between threads
9626 based on the specified memory order.
9628 All memory orders are valid.
9632 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9634 This built-in function acts as a synchronization fence between a thread
9635 and signal handlers based in the same thread.
9637 All memory orders are valid.
9641 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9643 This built-in function returns true if objects of @var{size} bytes always
9644 generate lock-free atomic instructions for the target architecture.
9645 @var{size} must resolve to a compile-time constant and the result also
9646 resolves to a compile-time constant.
9648 @var{ptr} is an optional pointer to the object that may be used to determine
9649 alignment. A value of 0 indicates typical alignment should be used. The
9650 compiler may also ignore this parameter.
9653 if (_atomic_always_lock_free (sizeof (long long), 0))
9658 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9660 This built-in function returns true if objects of @var{size} bytes always
9661 generate lock-free atomic instructions for the target architecture. If
9662 the built-in function is not known to be lock-free, a call is made to a
9663 runtime routine named @code{__atomic_is_lock_free}.
9665 @var{ptr} is an optional pointer to the object that may be used to determine
9666 alignment. A value of 0 indicates typical alignment should be used. The
9667 compiler may also ignore this parameter.
9670 @node Integer Overflow Builtins
9671 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9673 The following built-in functions allow performing simple arithmetic operations
9674 together with checking whether the operations overflowed.
9676 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9677 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9678 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9679 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9680 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9681 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9682 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9684 These built-in functions promote the first two operands into infinite precision signed
9685 type and perform addition on those promoted operands. The result is then
9686 cast to the type the third pointer argument points to and stored there.
9687 If the stored result is equal to the infinite precision result, the built-in
9688 functions return false, otherwise they return true. As the addition is
9689 performed in infinite signed precision, these built-in functions have fully defined
9690 behavior for all argument values.
9692 The first built-in function allows arbitrary integral types for operands and
9693 the result type must be pointer to some integer type, the rest of the built-in
9694 functions have explicit integer types.
9696 The compiler will attempt to use hardware instructions to implement
9697 these built-in functions where possible, like conditional jump on overflow
9698 after addition, conditional jump on carry etc.
9702 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9703 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9704 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9705 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9706 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9707 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9708 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9710 These built-in functions are similar to the add overflow checking built-in
9711 functions above, except they perform subtraction, subtract the second argument
9712 from the first one, instead of addition.
9716 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9717 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9718 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9719 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9720 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9721 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9722 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9724 These built-in functions are similar to the add overflow checking built-in
9725 functions above, except they perform multiplication, instead of addition.
9729 @node x86 specific memory model extensions for transactional memory
9730 @section x86-Specific Memory Model Extensions for Transactional Memory
9732 The x86 architecture supports additional memory ordering flags
9733 to mark lock critical sections for hardware lock elision.
9734 These must be specified in addition to an existing memory order to
9738 @item __ATOMIC_HLE_ACQUIRE
9739 Start lock elision on a lock variable.
9740 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9741 @item __ATOMIC_HLE_RELEASE
9742 End lock elision on a lock variable.
9743 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9746 When a lock acquire fails, it is required for good performance to abort
9747 the transaction quickly. This can be done with a @code{_mm_pause}.
9750 #include <immintrin.h> // For _mm_pause
9754 /* Acquire lock with lock elision */
9755 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9756 _mm_pause(); /* Abort failed transaction */
9758 /* Free lock with lock elision */
9759 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9762 @node Object Size Checking
9763 @section Object Size Checking Built-in Functions
9764 @findex __builtin_object_size
9765 @findex __builtin___memcpy_chk
9766 @findex __builtin___mempcpy_chk
9767 @findex __builtin___memmove_chk
9768 @findex __builtin___memset_chk
9769 @findex __builtin___strcpy_chk
9770 @findex __builtin___stpcpy_chk
9771 @findex __builtin___strncpy_chk
9772 @findex __builtin___strcat_chk
9773 @findex __builtin___strncat_chk
9774 @findex __builtin___sprintf_chk
9775 @findex __builtin___snprintf_chk
9776 @findex __builtin___vsprintf_chk
9777 @findex __builtin___vsnprintf_chk
9778 @findex __builtin___printf_chk
9779 @findex __builtin___vprintf_chk
9780 @findex __builtin___fprintf_chk
9781 @findex __builtin___vfprintf_chk
9783 GCC implements a limited buffer overflow protection mechanism
9784 that can prevent some buffer overflow attacks.
9786 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9787 is a built-in construct that returns a constant number of bytes from
9788 @var{ptr} to the end of the object @var{ptr} pointer points to
9789 (if known at compile time). @code{__builtin_object_size} never evaluates
9790 its arguments for side-effects. If there are any side-effects in them, it
9791 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9792 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9793 point to and all of them are known at compile time, the returned number
9794 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9795 0 and minimum if nonzero. If it is not possible to determine which objects
9796 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9797 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9798 for @var{type} 2 or 3.
9800 @var{type} is an integer constant from 0 to 3. If the least significant
9801 bit is clear, objects are whole variables, if it is set, a closest
9802 surrounding subobject is considered the object a pointer points to.
9803 The second bit determines if maximum or minimum of remaining bytes
9807 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9808 char *p = &var.buf1[1], *q = &var.b;
9810 /* Here the object p points to is var. */
9811 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9812 /* The subobject p points to is var.buf1. */
9813 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9814 /* The object q points to is var. */
9815 assert (__builtin_object_size (q, 0)
9816 == (char *) (&var + 1) - (char *) &var.b);
9817 /* The subobject q points to is var.b. */
9818 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9822 There are built-in functions added for many common string operation
9823 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9824 built-in is provided. This built-in has an additional last argument,
9825 which is the number of bytes remaining in object the @var{dest}
9826 argument points to or @code{(size_t) -1} if the size is not known.
9828 The built-in functions are optimized into the normal string functions
9829 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9830 it is known at compile time that the destination object will not
9831 be overflown. If the compiler can determine at compile time the
9832 object will be always overflown, it issues a warning.
9834 The intended use can be e.g.@:
9838 #define bos0(dest) __builtin_object_size (dest, 0)
9839 #define memcpy(dest, src, n) \
9840 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9844 /* It is unknown what object p points to, so this is optimized
9845 into plain memcpy - no checking is possible. */
9846 memcpy (p, "abcde", n);
9847 /* Destination is known and length too. It is known at compile
9848 time there will be no overflow. */
9849 memcpy (&buf[5], "abcde", 5);
9850 /* Destination is known, but the length is not known at compile time.
9851 This will result in __memcpy_chk call that can check for overflow
9853 memcpy (&buf[5], "abcde", n);
9854 /* Destination is known and it is known at compile time there will
9855 be overflow. There will be a warning and __memcpy_chk call that
9856 will abort the program at run time. */
9857 memcpy (&buf[6], "abcde", 5);
9860 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9861 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9862 @code{strcat} and @code{strncat}.
9864 There are also checking built-in functions for formatted output functions.
9866 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9867 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9868 const char *fmt, ...);
9869 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9871 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9872 const char *fmt, va_list ap);
9875 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9876 etc.@: functions and can contain implementation specific flags on what
9877 additional security measures the checking function might take, such as
9878 handling @code{%n} differently.
9880 The @var{os} argument is the object size @var{s} points to, like in the
9881 other built-in functions. There is a small difference in the behavior
9882 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9883 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9884 the checking function is called with @var{os} argument set to
9887 In addition to this, there are checking built-in functions
9888 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9889 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9890 These have just one additional argument, @var{flag}, right before
9891 format string @var{fmt}. If the compiler is able to optimize them to
9892 @code{fputc} etc.@: functions, it does, otherwise the checking function
9893 is called and the @var{flag} argument passed to it.
9895 @node Pointer Bounds Checker builtins
9896 @section Pointer Bounds Checker Built-in Functions
9897 @cindex Pointer Bounds Checker builtins
9898 @findex __builtin___bnd_set_ptr_bounds
9899 @findex __builtin___bnd_narrow_ptr_bounds
9900 @findex __builtin___bnd_copy_ptr_bounds
9901 @findex __builtin___bnd_init_ptr_bounds
9902 @findex __builtin___bnd_null_ptr_bounds
9903 @findex __builtin___bnd_store_ptr_bounds
9904 @findex __builtin___bnd_chk_ptr_lbounds
9905 @findex __builtin___bnd_chk_ptr_ubounds
9906 @findex __builtin___bnd_chk_ptr_bounds
9907 @findex __builtin___bnd_get_ptr_lbound
9908 @findex __builtin___bnd_get_ptr_ubound
9910 GCC provides a set of built-in functions to control Pointer Bounds Checker
9911 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9912 even if you compile with Pointer Bounds Checker off
9913 (@option{-fno-check-pointer-bounds}).
9914 The behavior may differ in such case as documented below.
9916 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9918 This built-in function returns a new pointer with the value of @var{q}, and
9919 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9920 Bounds Checker off, the built-in function just returns the first argument.
9923 extern void *__wrap_malloc (size_t n)
9925 void *p = (void *)__real_malloc (n);
9926 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9927 return __builtin___bnd_set_ptr_bounds (p, n);
9933 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9935 This built-in function returns a new pointer with the value of @var{p}
9936 and associates it with the narrowed bounds formed by the intersection
9937 of bounds associated with @var{q} and the bounds
9938 [@var{p}, @var{p} + @var{size} - 1].
9939 With Pointer Bounds Checker off, the built-in function just returns the first
9943 void init_objects (object *objs, size_t size)
9946 /* Initialize objects one-by-one passing pointers with bounds of
9947 an object, not the full array of objects. */
9948 for (i = 0; i < size; i++)
9949 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
9956 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
9958 This built-in function returns a new pointer with the value of @var{q},
9959 and associates it with the bounds already associated with pointer @var{r}.
9960 With Pointer Bounds Checker off, the built-in function just returns the first
9964 /* Here is a way to get pointer to object's field but
9965 still with the full object's bounds. */
9966 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
9972 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
9974 This built-in function returns a new pointer with the value of @var{q}, and
9975 associates it with INIT (allowing full memory access) bounds. With Pointer
9976 Bounds Checker off, the built-in function just returns the first argument.
9980 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
9982 This built-in function returns a new pointer with the value of @var{q}, and
9983 associates it with NULL (allowing no memory access) bounds. With Pointer
9984 Bounds Checker off, the built-in function just returns the first argument.
9988 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
9990 This built-in function stores the bounds associated with pointer @var{ptr_val}
9991 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
9992 bounds from legacy code without touching the associated pointer's memory when
9993 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
9994 function call is ignored.
9998 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10000 This built-in function checks if the pointer @var{q} is within the lower
10001 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10002 function call is ignored.
10005 extern void *__wrap_memset (void *dst, int c, size_t len)
10009 __builtin___bnd_chk_ptr_lbounds (dst);
10010 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10011 __real_memset (dst, c, len);
10019 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10021 This built-in function checks if the pointer @var{q} is within the upper
10022 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10023 function call is ignored.
10027 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10029 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10030 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10031 off, the built-in function call is ignored.
10034 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10038 __bnd_chk_ptr_bounds (dst, n);
10039 __bnd_chk_ptr_bounds (src, n);
10040 __real_memcpy (dst, src, n);
10048 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10050 This built-in function returns the lower bound associated
10051 with the pointer @var{q}, as a pointer value.
10052 This is useful for debugging using @code{printf}.
10053 With Pointer Bounds Checker off, the built-in function returns 0.
10056 void *lb = __builtin___bnd_get_ptr_lbound (q);
10057 void *ub = __builtin___bnd_get_ptr_ubound (q);
10058 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10063 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10065 This built-in function returns the upper bound (which is a pointer) associated
10066 with the pointer @var{q}. With Pointer Bounds Checker off,
10067 the built-in function returns -1.
10071 @node Cilk Plus Builtins
10072 @section Cilk Plus C/C++ Language Extension Built-in Functions
10074 GCC provides support for the following built-in reduction functions if Cilk Plus
10075 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10078 @item @code{__sec_implicit_index}
10079 @item @code{__sec_reduce}
10080 @item @code{__sec_reduce_add}
10081 @item @code{__sec_reduce_all_nonzero}
10082 @item @code{__sec_reduce_all_zero}
10083 @item @code{__sec_reduce_any_nonzero}
10084 @item @code{__sec_reduce_any_zero}
10085 @item @code{__sec_reduce_max}
10086 @item @code{__sec_reduce_min}
10087 @item @code{__sec_reduce_max_ind}
10088 @item @code{__sec_reduce_min_ind}
10089 @item @code{__sec_reduce_mul}
10090 @item @code{__sec_reduce_mutating}
10093 Further details and examples about these built-in functions are described
10094 in the Cilk Plus language manual which can be found at
10095 @uref{http://www.cilkplus.org}.
10097 @node Other Builtins
10098 @section Other Built-in Functions Provided by GCC
10099 @cindex built-in functions
10100 @findex __builtin_call_with_static_chain
10101 @findex __builtin_fpclassify
10102 @findex __builtin_isfinite
10103 @findex __builtin_isnormal
10104 @findex __builtin_isgreater
10105 @findex __builtin_isgreaterequal
10106 @findex __builtin_isinf_sign
10107 @findex __builtin_isless
10108 @findex __builtin_islessequal
10109 @findex __builtin_islessgreater
10110 @findex __builtin_isunordered
10111 @findex __builtin_powi
10112 @findex __builtin_powif
10113 @findex __builtin_powil
10271 @findex fprintf_unlocked
10273 @findex fputs_unlocked
10381 @findex nexttowardf
10382 @findex nexttowardl
10390 @findex printf_unlocked
10420 @findex signbitd128
10421 @findex significand
10422 @findex significandf
10423 @findex significandl
10451 @findex strncasecmp
10494 GCC provides a large number of built-in functions other than the ones
10495 mentioned above. Some of these are for internal use in the processing
10496 of exceptions or variable-length argument lists and are not
10497 documented here because they may change from time to time; we do not
10498 recommend general use of these functions.
10500 The remaining functions are provided for optimization purposes.
10502 With the exception of built-ins that have library equivalents such as
10503 the standard C library functions discussed below, or that expand to
10504 library calls, GCC built-in functions are always expanded inline and
10505 thus do not have corresponding entry points and their address cannot
10506 be obtained. Attempting to use them in an expression other than
10507 a function call results in a compile-time error.
10509 @opindex fno-builtin
10510 GCC includes built-in versions of many of the functions in the standard
10511 C library. These functions come in two forms: one whose names start with
10512 the @code{__builtin_} prefix, and the other without. Both forms have the
10513 same type (including prototype), the same address (when their address is
10514 taken), and the same meaning as the C library functions even if you specify
10515 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10516 functions are only optimized in certain cases; if they are not optimized in
10517 a particular case, a call to the library function is emitted.
10521 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10522 @option{-std=c99} or @option{-std=c11}), the functions
10523 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10524 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10525 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10526 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10527 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10528 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10529 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10530 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10531 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10532 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10533 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10534 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10535 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10536 @code{significandl}, @code{significand}, @code{sincosf},
10537 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10538 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10539 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10540 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10542 may be handled as built-in functions.
10543 All these functions have corresponding versions
10544 prefixed with @code{__builtin_}, which may be used even in strict C90
10547 The ISO C99 functions
10548 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10549 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10550 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10551 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10552 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10553 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10554 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10555 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10556 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10557 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10558 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10559 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10560 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10561 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10562 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10563 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10564 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10565 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10566 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10567 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10568 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10569 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10570 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10571 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10572 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10573 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10574 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10575 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10576 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10577 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10578 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10579 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10580 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10581 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10582 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10583 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10584 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10585 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10586 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10587 are handled as built-in functions
10588 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10590 There are also built-in versions of the ISO C99 functions
10591 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10592 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10593 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10594 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10595 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10596 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10597 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10598 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10599 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10600 that are recognized in any mode since ISO C90 reserves these names for
10601 the purpose to which ISO C99 puts them. All these functions have
10602 corresponding versions prefixed with @code{__builtin_}.
10604 The ISO C94 functions
10605 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10606 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10607 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10609 are handled as built-in functions
10610 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10612 The ISO C90 functions
10613 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10614 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10615 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10616 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10617 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10618 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10619 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10620 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10621 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10622 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10623 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10624 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10625 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10626 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10627 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10628 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10629 are all recognized as built-in functions unless
10630 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10631 is specified for an individual function). All of these functions have
10632 corresponding versions prefixed with @code{__builtin_}.
10634 GCC provides built-in versions of the ISO C99 floating-point comparison
10635 macros that avoid raising exceptions for unordered operands. They have
10636 the same names as the standard macros ( @code{isgreater},
10637 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10638 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10639 prefixed. We intend for a library implementor to be able to simply
10640 @code{#define} each standard macro to its built-in equivalent.
10641 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10642 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10643 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10644 built-in functions appear both with and without the @code{__builtin_} prefix.
10646 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10648 You can use the built-in function @code{__builtin_types_compatible_p} to
10649 determine whether two types are the same.
10651 This built-in function returns 1 if the unqualified versions of the
10652 types @var{type1} and @var{type2} (which are types, not expressions) are
10653 compatible, 0 otherwise. The result of this built-in function can be
10654 used in integer constant expressions.
10656 This built-in function ignores top level qualifiers (e.g., @code{const},
10657 @code{volatile}). For example, @code{int} is equivalent to @code{const
10660 The type @code{int[]} and @code{int[5]} are compatible. On the other
10661 hand, @code{int} and @code{char *} are not compatible, even if the size
10662 of their types, on the particular architecture are the same. Also, the
10663 amount of pointer indirection is taken into account when determining
10664 similarity. Consequently, @code{short *} is not similar to
10665 @code{short **}. Furthermore, two types that are typedefed are
10666 considered compatible if their underlying types are compatible.
10668 An @code{enum} type is not considered to be compatible with another
10669 @code{enum} type even if both are compatible with the same integer
10670 type; this is what the C standard specifies.
10671 For example, @code{enum @{foo, bar@}} is not similar to
10672 @code{enum @{hot, dog@}}.
10674 You typically use this function in code whose execution varies
10675 depending on the arguments' types. For example:
10680 typeof (x) tmp = (x); \
10681 if (__builtin_types_compatible_p (typeof (x), long double)) \
10682 tmp = foo_long_double (tmp); \
10683 else if (__builtin_types_compatible_p (typeof (x), double)) \
10684 tmp = foo_double (tmp); \
10685 else if (__builtin_types_compatible_p (typeof (x), float)) \
10686 tmp = foo_float (tmp); \
10693 @emph{Note:} This construct is only available for C@.
10697 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10699 The @var{call_exp} expression must be a function call, and the
10700 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10701 is passed to the function call in the target's static chain location.
10702 The result of builtin is the result of the function call.
10704 @emph{Note:} This builtin is only available for C@.
10705 This builtin can be used to call Go closures from C.
10709 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10711 You can use the built-in function @code{__builtin_choose_expr} to
10712 evaluate code depending on the value of a constant expression. This
10713 built-in function returns @var{exp1} if @var{const_exp}, which is an
10714 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10716 This built-in function is analogous to the @samp{? :} operator in C,
10717 except that the expression returned has its type unaltered by promotion
10718 rules. Also, the built-in function does not evaluate the expression
10719 that is not chosen. For example, if @var{const_exp} evaluates to true,
10720 @var{exp2} is not evaluated even if it has side-effects.
10722 This built-in function can return an lvalue if the chosen argument is an
10725 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10726 type. Similarly, if @var{exp2} is returned, its return type is the same
10733 __builtin_choose_expr ( \
10734 __builtin_types_compatible_p (typeof (x), double), \
10736 __builtin_choose_expr ( \
10737 __builtin_types_compatible_p (typeof (x), float), \
10739 /* @r{The void expression results in a compile-time error} \
10740 @r{when assigning the result to something.} */ \
10744 @emph{Note:} This construct is only available for C@. Furthermore, the
10745 unused expression (@var{exp1} or @var{exp2} depending on the value of
10746 @var{const_exp}) may still generate syntax errors. This may change in
10751 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10753 The built-in function @code{__builtin_complex} is provided for use in
10754 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10755 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10756 real binary floating-point type, and the result has the corresponding
10757 complex type with real and imaginary parts @var{real} and @var{imag}.
10758 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10759 infinities, NaNs and negative zeros are involved.
10763 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10764 You can use the built-in function @code{__builtin_constant_p} to
10765 determine if a value is known to be constant at compile time and hence
10766 that GCC can perform constant-folding on expressions involving that
10767 value. The argument of the function is the value to test. The function
10768 returns the integer 1 if the argument is known to be a compile-time
10769 constant and 0 if it is not known to be a compile-time constant. A
10770 return of 0 does not indicate that the value is @emph{not} a constant,
10771 but merely that GCC cannot prove it is a constant with the specified
10772 value of the @option{-O} option.
10774 You typically use this function in an embedded application where
10775 memory is a critical resource. If you have some complex calculation,
10776 you may want it to be folded if it involves constants, but need to call
10777 a function if it does not. For example:
10780 #define Scale_Value(X) \
10781 (__builtin_constant_p (X) \
10782 ? ((X) * SCALE + OFFSET) : Scale (X))
10785 You may use this built-in function in either a macro or an inline
10786 function. However, if you use it in an inlined function and pass an
10787 argument of the function as the argument to the built-in, GCC
10788 never returns 1 when you call the inline function with a string constant
10789 or compound literal (@pxref{Compound Literals}) and does not return 1
10790 when you pass a constant numeric value to the inline function unless you
10791 specify the @option{-O} option.
10793 You may also use @code{__builtin_constant_p} in initializers for static
10794 data. For instance, you can write
10797 static const int table[] = @{
10798 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10804 This is an acceptable initializer even if @var{EXPRESSION} is not a
10805 constant expression, including the case where
10806 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10807 folded to a constant but @var{EXPRESSION} contains operands that are
10808 not otherwise permitted in a static initializer (for example,
10809 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10810 built-in in this case, because it has no opportunity to perform
10814 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10815 @opindex fprofile-arcs
10816 You may use @code{__builtin_expect} to provide the compiler with
10817 branch prediction information. In general, you should prefer to
10818 use actual profile feedback for this (@option{-fprofile-arcs}), as
10819 programmers are notoriously bad at predicting how their programs
10820 actually perform. However, there are applications in which this
10821 data is hard to collect.
10823 The return value is the value of @var{exp}, which should be an integral
10824 expression. The semantics of the built-in are that it is expected that
10825 @var{exp} == @var{c}. For example:
10828 if (__builtin_expect (x, 0))
10833 indicates that we do not expect to call @code{foo}, since
10834 we expect @code{x} to be zero. Since you are limited to integral
10835 expressions for @var{exp}, you should use constructions such as
10838 if (__builtin_expect (ptr != NULL, 1))
10843 when testing pointer or floating-point values.
10846 @deftypefn {Built-in Function} void __builtin_trap (void)
10847 This function causes the program to exit abnormally. GCC implements
10848 this function by using a target-dependent mechanism (such as
10849 intentionally executing an illegal instruction) or by calling
10850 @code{abort}. The mechanism used may vary from release to release so
10851 you should not rely on any particular implementation.
10854 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10855 If control flow reaches the point of the @code{__builtin_unreachable},
10856 the program is undefined. It is useful in situations where the
10857 compiler cannot deduce the unreachability of the code.
10859 One such case is immediately following an @code{asm} statement that
10860 either never terminates, or one that transfers control elsewhere
10861 and never returns. In this example, without the
10862 @code{__builtin_unreachable}, GCC issues a warning that control
10863 reaches the end of a non-void function. It also generates code
10864 to return after the @code{asm}.
10867 int f (int c, int v)
10875 asm("jmp error_handler");
10876 __builtin_unreachable ();
10882 Because the @code{asm} statement unconditionally transfers control out
10883 of the function, control never reaches the end of the function
10884 body. The @code{__builtin_unreachable} is in fact unreachable and
10885 communicates this fact to the compiler.
10887 Another use for @code{__builtin_unreachable} is following a call a
10888 function that never returns but that is not declared
10889 @code{__attribute__((noreturn))}, as in this example:
10892 void function_that_never_returns (void);
10902 function_that_never_returns ();
10903 __builtin_unreachable ();
10910 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
10911 This function returns its first argument, and allows the compiler
10912 to assume that the returned pointer is at least @var{align} bytes
10913 aligned. This built-in can have either two or three arguments,
10914 if it has three, the third argument should have integer type, and
10915 if it is nonzero means misalignment offset. For example:
10918 void *x = __builtin_assume_aligned (arg, 16);
10922 means that the compiler can assume @code{x}, set to @code{arg}, is at least
10923 16-byte aligned, while:
10926 void *x = __builtin_assume_aligned (arg, 32, 8);
10930 means that the compiler can assume for @code{x}, set to @code{arg}, that
10931 @code{(char *) x - 8} is 32-byte aligned.
10934 @deftypefn {Built-in Function} int __builtin_LINE ()
10935 This function is the equivalent to the preprocessor @code{__LINE__}
10936 macro and returns the line number of the invocation of the built-in.
10937 In a C++ default argument for a function @var{F}, it gets the line number of
10938 the call to @var{F}.
10941 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
10942 This function is the equivalent to the preprocessor @code{__FUNCTION__}
10943 macro and returns the function name the invocation of the built-in is in.
10946 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
10947 This function is the equivalent to the preprocessor @code{__FILE__}
10948 macro and returns the file name the invocation of the built-in is in.
10949 In a C++ default argument for a function @var{F}, it gets the file name of
10950 the call to @var{F}.
10953 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
10954 This function is used to flush the processor's instruction cache for
10955 the region of memory between @var{begin} inclusive and @var{end}
10956 exclusive. Some targets require that the instruction cache be
10957 flushed, after modifying memory containing code, in order to obtain
10958 deterministic behavior.
10960 If the target does not require instruction cache flushes,
10961 @code{__builtin___clear_cache} has no effect. Otherwise either
10962 instructions are emitted in-line to clear the instruction cache or a
10963 call to the @code{__clear_cache} function in libgcc is made.
10966 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
10967 This function is used to minimize cache-miss latency by moving data into
10968 a cache before it is accessed.
10969 You can insert calls to @code{__builtin_prefetch} into code for which
10970 you know addresses of data in memory that is likely to be accessed soon.
10971 If the target supports them, data prefetch instructions are generated.
10972 If the prefetch is done early enough before the access then the data will
10973 be in the cache by the time it is accessed.
10975 The value of @var{addr} is the address of the memory to prefetch.
10976 There are two optional arguments, @var{rw} and @var{locality}.
10977 The value of @var{rw} is a compile-time constant one or zero; one
10978 means that the prefetch is preparing for a write to the memory address
10979 and zero, the default, means that the prefetch is preparing for a read.
10980 The value @var{locality} must be a compile-time constant integer between
10981 zero and three. A value of zero means that the data has no temporal
10982 locality, so it need not be left in the cache after the access. A value
10983 of three means that the data has a high degree of temporal locality and
10984 should be left in all levels of cache possible. Values of one and two
10985 mean, respectively, a low or moderate degree of temporal locality. The
10989 for (i = 0; i < n; i++)
10991 a[i] = a[i] + b[i];
10992 __builtin_prefetch (&a[i+j], 1, 1);
10993 __builtin_prefetch (&b[i+j], 0, 1);
10998 Data prefetch does not generate faults if @var{addr} is invalid, but
10999 the address expression itself must be valid. For example, a prefetch
11000 of @code{p->next} does not fault if @code{p->next} is not a valid
11001 address, but evaluation faults if @code{p} is not a valid address.
11003 If the target does not support data prefetch, the address expression
11004 is evaluated if it includes side effects but no other code is generated
11005 and GCC does not issue a warning.
11008 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11009 Returns a positive infinity, if supported by the floating-point format,
11010 else @code{DBL_MAX}. This function is suitable for implementing the
11011 ISO C macro @code{HUGE_VAL}.
11014 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11015 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11018 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11019 Similar to @code{__builtin_huge_val}, except the return
11020 type is @code{long double}.
11023 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11024 This built-in implements the C99 fpclassify functionality. The first
11025 five int arguments should be the target library's notion of the
11026 possible FP classes and are used for return values. They must be
11027 constant values and they must appear in this order: @code{FP_NAN},
11028 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11029 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11030 to classify. GCC treats the last argument as type-generic, which
11031 means it does not do default promotion from float to double.
11034 @deftypefn {Built-in Function} double __builtin_inf (void)
11035 Similar to @code{__builtin_huge_val}, except a warning is generated
11036 if the target floating-point format does not support infinities.
11039 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11040 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11043 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11044 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11047 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11048 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11051 @deftypefn {Built-in Function} float __builtin_inff (void)
11052 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11053 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11056 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11057 Similar to @code{__builtin_inf}, except the return
11058 type is @code{long double}.
11061 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11062 Similar to @code{isinf}, except the return value is -1 for
11063 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11064 Note while the parameter list is an
11065 ellipsis, this function only accepts exactly one floating-point
11066 argument. GCC treats this parameter as type-generic, which means it
11067 does not do default promotion from float to double.
11070 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11071 This is an implementation of the ISO C99 function @code{nan}.
11073 Since ISO C99 defines this function in terms of @code{strtod}, which we
11074 do not implement, a description of the parsing is in order. The string
11075 is parsed as by @code{strtol}; that is, the base is recognized by
11076 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11077 in the significand such that the least significant bit of the number
11078 is at the least significant bit of the significand. The number is
11079 truncated to fit the significand field provided. The significand is
11080 forced to be a quiet NaN@.
11082 This function, if given a string literal all of which would have been
11083 consumed by @code{strtol}, is evaluated early enough that it is considered a
11084 compile-time constant.
11087 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11088 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11091 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11092 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11095 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11096 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11099 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11100 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11103 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11104 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11107 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11108 Similar to @code{__builtin_nan}, except the significand is forced
11109 to be a signaling NaN@. The @code{nans} function is proposed by
11110 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11113 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11114 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11117 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11118 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11121 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11122 Returns one plus the index of the least significant 1-bit of @var{x}, or
11123 if @var{x} is zero, returns zero.
11126 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11127 Returns the number of leading 0-bits in @var{x}, starting at the most
11128 significant bit position. If @var{x} is 0, the result is undefined.
11131 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11132 Returns the number of trailing 0-bits in @var{x}, starting at the least
11133 significant bit position. If @var{x} is 0, the result is undefined.
11136 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11137 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11138 number of bits following the most significant bit that are identical
11139 to it. There are no special cases for 0 or other values.
11142 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11143 Returns the number of 1-bits in @var{x}.
11146 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11147 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11151 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11152 Similar to @code{__builtin_ffs}, except the argument type is
11156 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11157 Similar to @code{__builtin_clz}, except the argument type is
11158 @code{unsigned long}.
11161 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11162 Similar to @code{__builtin_ctz}, except the argument type is
11163 @code{unsigned long}.
11166 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11167 Similar to @code{__builtin_clrsb}, except the argument type is
11171 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11172 Similar to @code{__builtin_popcount}, except the argument type is
11173 @code{unsigned long}.
11176 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11177 Similar to @code{__builtin_parity}, except the argument type is
11178 @code{unsigned long}.
11181 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11182 Similar to @code{__builtin_ffs}, except the argument type is
11186 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11187 Similar to @code{__builtin_clz}, except the argument type is
11188 @code{unsigned long long}.
11191 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11192 Similar to @code{__builtin_ctz}, except the argument type is
11193 @code{unsigned long long}.
11196 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11197 Similar to @code{__builtin_clrsb}, except the argument type is
11201 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11202 Similar to @code{__builtin_popcount}, except the argument type is
11203 @code{unsigned long long}.
11206 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11207 Similar to @code{__builtin_parity}, except the argument type is
11208 @code{unsigned long long}.
11211 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11212 Returns the first argument raised to the power of the second. Unlike the
11213 @code{pow} function no guarantees about precision and rounding are made.
11216 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11217 Similar to @code{__builtin_powi}, except the argument and return types
11221 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11222 Similar to @code{__builtin_powi}, except the argument and return types
11223 are @code{long double}.
11226 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11227 Returns @var{x} with the order of the bytes reversed; for example,
11228 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11232 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11233 Similar to @code{__builtin_bswap16}, except the argument and return types
11237 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11238 Similar to @code{__builtin_bswap32}, except the argument and return types
11242 @node Target Builtins
11243 @section Built-in Functions Specific to Particular Target Machines
11245 On some target machines, GCC supports many built-in functions specific
11246 to those machines. Generally these generate calls to specific machine
11247 instructions, but allow the compiler to schedule those calls.
11250 * AArch64 Built-in Functions::
11251 * Alpha Built-in Functions::
11252 * Altera Nios II Built-in Functions::
11253 * ARC Built-in Functions::
11254 * ARC SIMD Built-in Functions::
11255 * ARM iWMMXt Built-in Functions::
11256 * ARM C Language Extensions (ACLE)::
11257 * ARM Floating Point Status and Control Intrinsics::
11258 * AVR Built-in Functions::
11259 * Blackfin Built-in Functions::
11260 * FR-V Built-in Functions::
11261 * MIPS DSP Built-in Functions::
11262 * MIPS Paired-Single Support::
11263 * MIPS Loongson Built-in Functions::
11264 * Other MIPS Built-in Functions::
11265 * MSP430 Built-in Functions::
11266 * NDS32 Built-in Functions::
11267 * picoChip Built-in Functions::
11268 * PowerPC Built-in Functions::
11269 * PowerPC AltiVec/VSX Built-in Functions::
11270 * PowerPC Hardware Transactional Memory Built-in Functions::
11271 * RX Built-in Functions::
11272 * S/390 System z Built-in Functions::
11273 * SH Built-in Functions::
11274 * SPARC VIS Built-in Functions::
11275 * SPU Built-in Functions::
11276 * TI C6X Built-in Functions::
11277 * TILE-Gx Built-in Functions::
11278 * TILEPro Built-in Functions::
11279 * x86 Built-in Functions::
11280 * x86 transactional memory intrinsics::
11283 @node AArch64 Built-in Functions
11284 @subsection AArch64 Built-in Functions
11286 These built-in functions are available for the AArch64 family of
11289 unsigned int __builtin_aarch64_get_fpcr ()
11290 void __builtin_aarch64_set_fpcr (unsigned int)
11291 unsigned int __builtin_aarch64_get_fpsr ()
11292 void __builtin_aarch64_set_fpsr (unsigned int)
11295 @node Alpha Built-in Functions
11296 @subsection Alpha Built-in Functions
11298 These built-in functions are available for the Alpha family of
11299 processors, depending on the command-line switches used.
11301 The following built-in functions are always available. They
11302 all generate the machine instruction that is part of the name.
11305 long __builtin_alpha_implver (void)
11306 long __builtin_alpha_rpcc (void)
11307 long __builtin_alpha_amask (long)
11308 long __builtin_alpha_cmpbge (long, long)
11309 long __builtin_alpha_extbl (long, long)
11310 long __builtin_alpha_extwl (long, long)
11311 long __builtin_alpha_extll (long, long)
11312 long __builtin_alpha_extql (long, long)
11313 long __builtin_alpha_extwh (long, long)
11314 long __builtin_alpha_extlh (long, long)
11315 long __builtin_alpha_extqh (long, long)
11316 long __builtin_alpha_insbl (long, long)
11317 long __builtin_alpha_inswl (long, long)
11318 long __builtin_alpha_insll (long, long)
11319 long __builtin_alpha_insql (long, long)
11320 long __builtin_alpha_inswh (long, long)
11321 long __builtin_alpha_inslh (long, long)
11322 long __builtin_alpha_insqh (long, long)
11323 long __builtin_alpha_mskbl (long, long)
11324 long __builtin_alpha_mskwl (long, long)
11325 long __builtin_alpha_mskll (long, long)
11326 long __builtin_alpha_mskql (long, long)
11327 long __builtin_alpha_mskwh (long, long)
11328 long __builtin_alpha_msklh (long, long)
11329 long __builtin_alpha_mskqh (long, long)
11330 long __builtin_alpha_umulh (long, long)
11331 long __builtin_alpha_zap (long, long)
11332 long __builtin_alpha_zapnot (long, long)
11335 The following built-in functions are always with @option{-mmax}
11336 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11337 later. They all generate the machine instruction that is part
11341 long __builtin_alpha_pklb (long)
11342 long __builtin_alpha_pkwb (long)
11343 long __builtin_alpha_unpkbl (long)
11344 long __builtin_alpha_unpkbw (long)
11345 long __builtin_alpha_minub8 (long, long)
11346 long __builtin_alpha_minsb8 (long, long)
11347 long __builtin_alpha_minuw4 (long, long)
11348 long __builtin_alpha_minsw4 (long, long)
11349 long __builtin_alpha_maxub8 (long, long)
11350 long __builtin_alpha_maxsb8 (long, long)
11351 long __builtin_alpha_maxuw4 (long, long)
11352 long __builtin_alpha_maxsw4 (long, long)
11353 long __builtin_alpha_perr (long, long)
11356 The following built-in functions are always with @option{-mcix}
11357 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11358 later. They all generate the machine instruction that is part
11362 long __builtin_alpha_cttz (long)
11363 long __builtin_alpha_ctlz (long)
11364 long __builtin_alpha_ctpop (long)
11367 The following built-in functions are available on systems that use the OSF/1
11368 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11369 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11370 @code{rdval} and @code{wrval}.
11373 void *__builtin_thread_pointer (void)
11374 void __builtin_set_thread_pointer (void *)
11377 @node Altera Nios II Built-in Functions
11378 @subsection Altera Nios II Built-in Functions
11380 These built-in functions are available for the Altera Nios II
11381 family of processors.
11383 The following built-in functions are always available. They
11384 all generate the machine instruction that is part of the name.
11387 int __builtin_ldbio (volatile const void *)
11388 int __builtin_ldbuio (volatile const void *)
11389 int __builtin_ldhio (volatile const void *)
11390 int __builtin_ldhuio (volatile const void *)
11391 int __builtin_ldwio (volatile const void *)
11392 void __builtin_stbio (volatile void *, int)
11393 void __builtin_sthio (volatile void *, int)
11394 void __builtin_stwio (volatile void *, int)
11395 void __builtin_sync (void)
11396 int __builtin_rdctl (int)
11397 int __builtin_rdprs (int, int)
11398 void __builtin_wrctl (int, int)
11399 void __builtin_flushd (volatile void *)
11400 void __builtin_flushda (volatile void *)
11401 int __builtin_wrpie (int);
11402 void __builtin_eni (int);
11403 int __builtin_ldex (volatile const void *)
11404 int __builtin_stex (volatile void *, int)
11405 int __builtin_ldsex (volatile const void *)
11406 int __builtin_stsex (volatile void *, int)
11409 The following built-in functions are always available. They
11410 all generate a Nios II Custom Instruction. The name of the
11411 function represents the types that the function takes and
11412 returns. The letter before the @code{n} is the return type
11413 or void if absent. The @code{n} represents the first parameter
11414 to all the custom instructions, the custom instruction number.
11415 The two letters after the @code{n} represent the up to two
11416 parameters to the function.
11418 The letters represent the following data types:
11421 @code{void} for return type and no parameter for parameter types.
11424 @code{int} for return type and parameter type
11427 @code{float} for return type and parameter type
11430 @code{void *} for return type and parameter type
11434 And the function names are:
11436 void __builtin_custom_n (void)
11437 void __builtin_custom_ni (int)
11438 void __builtin_custom_nf (float)
11439 void __builtin_custom_np (void *)
11440 void __builtin_custom_nii (int, int)
11441 void __builtin_custom_nif (int, float)
11442 void __builtin_custom_nip (int, void *)
11443 void __builtin_custom_nfi (float, int)
11444 void __builtin_custom_nff (float, float)
11445 void __builtin_custom_nfp (float, void *)
11446 void __builtin_custom_npi (void *, int)
11447 void __builtin_custom_npf (void *, float)
11448 void __builtin_custom_npp (void *, void *)
11449 int __builtin_custom_in (void)
11450 int __builtin_custom_ini (int)
11451 int __builtin_custom_inf (float)
11452 int __builtin_custom_inp (void *)
11453 int __builtin_custom_inii (int, int)
11454 int __builtin_custom_inif (int, float)
11455 int __builtin_custom_inip (int, void *)
11456 int __builtin_custom_infi (float, int)
11457 int __builtin_custom_inff (float, float)
11458 int __builtin_custom_infp (float, void *)
11459 int __builtin_custom_inpi (void *, int)
11460 int __builtin_custom_inpf (void *, float)
11461 int __builtin_custom_inpp (void *, void *)
11462 float __builtin_custom_fn (void)
11463 float __builtin_custom_fni (int)
11464 float __builtin_custom_fnf (float)
11465 float __builtin_custom_fnp (void *)
11466 float __builtin_custom_fnii (int, int)
11467 float __builtin_custom_fnif (int, float)
11468 float __builtin_custom_fnip (int, void *)
11469 float __builtin_custom_fnfi (float, int)
11470 float __builtin_custom_fnff (float, float)
11471 float __builtin_custom_fnfp (float, void *)
11472 float __builtin_custom_fnpi (void *, int)
11473 float __builtin_custom_fnpf (void *, float)
11474 float __builtin_custom_fnpp (void *, void *)
11475 void * __builtin_custom_pn (void)
11476 void * __builtin_custom_pni (int)
11477 void * __builtin_custom_pnf (float)
11478 void * __builtin_custom_pnp (void *)
11479 void * __builtin_custom_pnii (int, int)
11480 void * __builtin_custom_pnif (int, float)
11481 void * __builtin_custom_pnip (int, void *)
11482 void * __builtin_custom_pnfi (float, int)
11483 void * __builtin_custom_pnff (float, float)
11484 void * __builtin_custom_pnfp (float, void *)
11485 void * __builtin_custom_pnpi (void *, int)
11486 void * __builtin_custom_pnpf (void *, float)
11487 void * __builtin_custom_pnpp (void *, void *)
11490 @node ARC Built-in Functions
11491 @subsection ARC Built-in Functions
11493 The following built-in functions are provided for ARC targets. The
11494 built-ins generate the corresponding assembly instructions. In the
11495 examples given below, the generated code often requires an operand or
11496 result to be in a register. Where necessary further code will be
11497 generated to ensure this is true, but for brevity this is not
11498 described in each case.
11500 @emph{Note:} Using a built-in to generate an instruction not supported
11501 by a target may cause problems. At present the compiler is not
11502 guaranteed to detect such misuse, and as a result an internal compiler
11503 error may be generated.
11505 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11506 Return 1 if @var{val} is known to have the byte alignment given
11507 by @var{alignval}, otherwise return 0.
11508 Note that this is different from
11510 __alignof__(*(char *)@var{val}) >= alignval
11512 because __alignof__ sees only the type of the dereference, whereas
11513 __builtin_arc_align uses alignment information from the pointer
11514 as well as from the pointed-to type.
11515 The information available will depend on optimization level.
11518 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11525 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11526 The operand is the number of a register to be read. Generates:
11528 mov @var{dest}, r@var{regno}
11530 where the value in @var{dest} will be the result returned from the
11534 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11535 The first operand is the number of a register to be written, the
11536 second operand is a compile time constant to write into that
11537 register. Generates:
11539 mov r@var{regno}, @var{val}
11543 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11544 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11547 divaw @var{dest}, @var{a}, @var{b}
11549 where the value in @var{dest} will be the result returned from the
11553 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11560 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11561 The operand, @var{auxv}, is the address of an auxiliary register and
11562 must be a compile time constant. Generates:
11564 lr @var{dest}, [@var{auxr}]
11566 Where the value in @var{dest} will be the result returned from the
11570 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11571 Only available with @option{-mmul64}. Generates:
11573 mul64 @var{a}, @var{b}
11577 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11578 Only available with @option{-mmul64}. Generates:
11580 mulu64 @var{a}, @var{b}
11584 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11591 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11592 Only valid if the @samp{norm} instruction is available through the
11593 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11596 norm @var{dest}, @var{src}
11598 Where the value in @var{dest} will be the result returned from the
11602 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11603 Only valid if the @samp{normw} instruction is available through the
11604 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11607 normw @var{dest}, @var{src}
11609 Where the value in @var{dest} will be the result returned from the
11613 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11620 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11627 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11628 The first argument, @var{auxv}, is the address of an auxiliary
11629 register, the second argument, @var{val}, is a compile time constant
11630 to be written to the register. Generates:
11632 sr @var{auxr}, [@var{val}]
11636 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11637 Only valid with @option{-mswap}. Generates:
11639 swap @var{dest}, @var{src}
11641 Where the value in @var{dest} will be the result returned from the
11645 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11652 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11653 Only available with @option{-mcpu=ARC700}. Generates:
11659 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11660 Only available with @option{-mcpu=ARC700}. Generates:
11666 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11667 Only available with @option{-mcpu=ARC700}. Generates:
11673 The instructions generated by the following builtins are not
11674 considered as candidates for scheduling. They are not moved around by
11675 the compiler during scheduling, and thus can be expected to appear
11676 where they are put in the C code:
11678 __builtin_arc_brk()
11679 __builtin_arc_core_read()
11680 __builtin_arc_core_write()
11681 __builtin_arc_flag()
11683 __builtin_arc_sleep()
11685 __builtin_arc_swi()
11688 @node ARC SIMD Built-in Functions
11689 @subsection ARC SIMD Built-in Functions
11691 SIMD builtins provided by the compiler can be used to generate the
11692 vector instructions. This section describes the available builtins
11693 and their usage in programs. With the @option{-msimd} option, the
11694 compiler provides 128-bit vector types, which can be specified using
11695 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11696 can be included to use the following predefined types:
11698 typedef int __v4si __attribute__((vector_size(16)));
11699 typedef short __v8hi __attribute__((vector_size(16)));
11702 These types can be used to define 128-bit variables. The built-in
11703 functions listed in the following section can be used on these
11704 variables to generate the vector operations.
11706 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11707 @file{arc-simd.h} also provides equivalent macros called
11708 @code{_@var{someinsn}} that can be used for programming ease and
11709 improved readability. The following macros for DMA control are also
11712 #define _setup_dma_in_channel_reg _vdiwr
11713 #define _setup_dma_out_channel_reg _vdowr
11716 The following is a complete list of all the SIMD built-ins provided
11717 for ARC, grouped by calling signature.
11719 The following take two @code{__v8hi} arguments and return a
11720 @code{__v8hi} result:
11722 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11723 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11724 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11725 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11726 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11727 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11728 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11729 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11730 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11731 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11732 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11733 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11734 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11735 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11736 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11737 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11738 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11739 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11740 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11741 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11742 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11743 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11744 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11745 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11746 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11747 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11748 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11749 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11750 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11751 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11752 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11753 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11754 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11755 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11756 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11757 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11758 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11759 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11760 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11761 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11762 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11763 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11764 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11765 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11766 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11767 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11768 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11769 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11772 The following take one @code{__v8hi} and one @code{int} argument and return a
11773 @code{__v8hi} result:
11776 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11777 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11778 __v8hi __builtin_arc_vbminw (__v8hi, int)
11779 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11780 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11781 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11782 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11783 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11786 The following take one @code{__v8hi} argument and one @code{int} argument which
11787 must be a 3-bit compile time constant indicating a register number
11788 I0-I7. They return a @code{__v8hi} result.
11790 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11791 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11792 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11795 The following take one @code{__v8hi} argument and one @code{int}
11796 argument which must be a 6-bit compile time constant. They return a
11797 @code{__v8hi} result.
11799 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11800 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11801 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11802 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11803 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11804 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11805 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11808 The following take one @code{__v8hi} argument and one @code{int} argument which
11809 must be a 8-bit compile time constant. They return a @code{__v8hi}
11812 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11813 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11814 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11815 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11818 The following take two @code{int} arguments, the second of which which
11819 must be a 8-bit compile time constant. They return a @code{__v8hi}
11822 __v8hi __builtin_arc_vmovaw (int, const int)
11823 __v8hi __builtin_arc_vmovw (int, const int)
11824 __v8hi __builtin_arc_vmovzw (int, const int)
11827 The following take a single @code{__v8hi} argument and return a
11828 @code{__v8hi} result:
11830 __v8hi __builtin_arc_vabsaw (__v8hi)
11831 __v8hi __builtin_arc_vabsw (__v8hi)
11832 __v8hi __builtin_arc_vaddsuw (__v8hi)
11833 __v8hi __builtin_arc_vexch1 (__v8hi)
11834 __v8hi __builtin_arc_vexch2 (__v8hi)
11835 __v8hi __builtin_arc_vexch4 (__v8hi)
11836 __v8hi __builtin_arc_vsignw (__v8hi)
11837 __v8hi __builtin_arc_vupbaw (__v8hi)
11838 __v8hi __builtin_arc_vupbw (__v8hi)
11839 __v8hi __builtin_arc_vupsbaw (__v8hi)
11840 __v8hi __builtin_arc_vupsbw (__v8hi)
11843 The following take two @code{int} arguments and return no result:
11845 void __builtin_arc_vdirun (int, int)
11846 void __builtin_arc_vdorun (int, int)
11849 The following take two @code{int} arguments and return no result. The
11850 first argument must a 3-bit compile time constant indicating one of
11851 the DR0-DR7 DMA setup channels:
11853 void __builtin_arc_vdiwr (const int, int)
11854 void __builtin_arc_vdowr (const int, int)
11857 The following take an @code{int} argument and return no result:
11859 void __builtin_arc_vendrec (int)
11860 void __builtin_arc_vrec (int)
11861 void __builtin_arc_vrecrun (int)
11862 void __builtin_arc_vrun (int)
11865 The following take a @code{__v8hi} argument and two @code{int}
11866 arguments and return a @code{__v8hi} result. The second argument must
11867 be a 3-bit compile time constants, indicating one the registers I0-I7,
11868 and the third argument must be an 8-bit compile time constant.
11870 @emph{Note:} Although the equivalent hardware instructions do not take
11871 an SIMD register as an operand, these builtins overwrite the relevant
11872 bits of the @code{__v8hi} register provided as the first argument with
11873 the value loaded from the @code{[Ib, u8]} location in the SDM.
11876 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
11877 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
11878 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
11879 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
11882 The following take two @code{int} arguments and return a @code{__v8hi}
11883 result. The first argument must be a 3-bit compile time constants,
11884 indicating one the registers I0-I7, and the second argument must be an
11885 8-bit compile time constant.
11888 __v8hi __builtin_arc_vld128 (const int, const int)
11889 __v8hi __builtin_arc_vld64w (const int, const int)
11892 The following take a @code{__v8hi} argument and two @code{int}
11893 arguments and return no result. The second argument must be a 3-bit
11894 compile time constants, indicating one the registers I0-I7, and the
11895 third argument must be an 8-bit compile time constant.
11898 void __builtin_arc_vst128 (__v8hi, const int, const int)
11899 void __builtin_arc_vst64 (__v8hi, const int, const int)
11902 The following take a @code{__v8hi} argument and three @code{int}
11903 arguments and return no result. The second argument must be a 3-bit
11904 compile-time constant, identifying the 16-bit sub-register to be
11905 stored, the third argument must be a 3-bit compile time constants,
11906 indicating one the registers I0-I7, and the fourth argument must be an
11907 8-bit compile time constant.
11910 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
11911 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
11914 @node ARM iWMMXt Built-in Functions
11915 @subsection ARM iWMMXt Built-in Functions
11917 These built-in functions are available for the ARM family of
11918 processors when the @option{-mcpu=iwmmxt} switch is used:
11921 typedef int v2si __attribute__ ((vector_size (8)));
11922 typedef short v4hi __attribute__ ((vector_size (8)));
11923 typedef char v8qi __attribute__ ((vector_size (8)));
11925 int __builtin_arm_getwcgr0 (void)
11926 void __builtin_arm_setwcgr0 (int)
11927 int __builtin_arm_getwcgr1 (void)
11928 void __builtin_arm_setwcgr1 (int)
11929 int __builtin_arm_getwcgr2 (void)
11930 void __builtin_arm_setwcgr2 (int)
11931 int __builtin_arm_getwcgr3 (void)
11932 void __builtin_arm_setwcgr3 (int)
11933 int __builtin_arm_textrmsb (v8qi, int)
11934 int __builtin_arm_textrmsh (v4hi, int)
11935 int __builtin_arm_textrmsw (v2si, int)
11936 int __builtin_arm_textrmub (v8qi, int)
11937 int __builtin_arm_textrmuh (v4hi, int)
11938 int __builtin_arm_textrmuw (v2si, int)
11939 v8qi __builtin_arm_tinsrb (v8qi, int, int)
11940 v4hi __builtin_arm_tinsrh (v4hi, int, int)
11941 v2si __builtin_arm_tinsrw (v2si, int, int)
11942 long long __builtin_arm_tmia (long long, int, int)
11943 long long __builtin_arm_tmiabb (long long, int, int)
11944 long long __builtin_arm_tmiabt (long long, int, int)
11945 long long __builtin_arm_tmiaph (long long, int, int)
11946 long long __builtin_arm_tmiatb (long long, int, int)
11947 long long __builtin_arm_tmiatt (long long, int, int)
11948 int __builtin_arm_tmovmskb (v8qi)
11949 int __builtin_arm_tmovmskh (v4hi)
11950 int __builtin_arm_tmovmskw (v2si)
11951 long long __builtin_arm_waccb (v8qi)
11952 long long __builtin_arm_wacch (v4hi)
11953 long long __builtin_arm_waccw (v2si)
11954 v8qi __builtin_arm_waddb (v8qi, v8qi)
11955 v8qi __builtin_arm_waddbss (v8qi, v8qi)
11956 v8qi __builtin_arm_waddbus (v8qi, v8qi)
11957 v4hi __builtin_arm_waddh (v4hi, v4hi)
11958 v4hi __builtin_arm_waddhss (v4hi, v4hi)
11959 v4hi __builtin_arm_waddhus (v4hi, v4hi)
11960 v2si __builtin_arm_waddw (v2si, v2si)
11961 v2si __builtin_arm_waddwss (v2si, v2si)
11962 v2si __builtin_arm_waddwus (v2si, v2si)
11963 v8qi __builtin_arm_walign (v8qi, v8qi, int)
11964 long long __builtin_arm_wand(long long, long long)
11965 long long __builtin_arm_wandn (long long, long long)
11966 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
11967 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
11968 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
11969 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
11970 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
11971 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
11972 v2si __builtin_arm_wcmpeqw (v2si, v2si)
11973 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
11974 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
11975 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
11976 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
11977 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
11978 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
11979 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
11980 long long __builtin_arm_wmacsz (v4hi, v4hi)
11981 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
11982 long long __builtin_arm_wmacuz (v4hi, v4hi)
11983 v4hi __builtin_arm_wmadds (v4hi, v4hi)
11984 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
11985 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
11986 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
11987 v2si __builtin_arm_wmaxsw (v2si, v2si)
11988 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
11989 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
11990 v2si __builtin_arm_wmaxuw (v2si, v2si)
11991 v8qi __builtin_arm_wminsb (v8qi, v8qi)
11992 v4hi __builtin_arm_wminsh (v4hi, v4hi)
11993 v2si __builtin_arm_wminsw (v2si, v2si)
11994 v8qi __builtin_arm_wminub (v8qi, v8qi)
11995 v4hi __builtin_arm_wminuh (v4hi, v4hi)
11996 v2si __builtin_arm_wminuw (v2si, v2si)
11997 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
11998 v4hi __builtin_arm_wmulul (v4hi, v4hi)
11999 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12000 long long __builtin_arm_wor (long long, long long)
12001 v2si __builtin_arm_wpackdss (long long, long long)
12002 v2si __builtin_arm_wpackdus (long long, long long)
12003 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12004 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12005 v4hi __builtin_arm_wpackwss (v2si, v2si)
12006 v4hi __builtin_arm_wpackwus (v2si, v2si)
12007 long long __builtin_arm_wrord (long long, long long)
12008 long long __builtin_arm_wrordi (long long, int)
12009 v4hi __builtin_arm_wrorh (v4hi, long long)
12010 v4hi __builtin_arm_wrorhi (v4hi, int)
12011 v2si __builtin_arm_wrorw (v2si, long long)
12012 v2si __builtin_arm_wrorwi (v2si, int)
12013 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12014 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12015 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12016 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12017 v4hi __builtin_arm_wshufh (v4hi, int)
12018 long long __builtin_arm_wslld (long long, long long)
12019 long long __builtin_arm_wslldi (long long, int)
12020 v4hi __builtin_arm_wsllh (v4hi, long long)
12021 v4hi __builtin_arm_wsllhi (v4hi, int)
12022 v2si __builtin_arm_wsllw (v2si, long long)
12023 v2si __builtin_arm_wsllwi (v2si, int)
12024 long long __builtin_arm_wsrad (long long, long long)
12025 long long __builtin_arm_wsradi (long long, int)
12026 v4hi __builtin_arm_wsrah (v4hi, long long)
12027 v4hi __builtin_arm_wsrahi (v4hi, int)
12028 v2si __builtin_arm_wsraw (v2si, long long)
12029 v2si __builtin_arm_wsrawi (v2si, int)
12030 long long __builtin_arm_wsrld (long long, long long)
12031 long long __builtin_arm_wsrldi (long long, int)
12032 v4hi __builtin_arm_wsrlh (v4hi, long long)
12033 v4hi __builtin_arm_wsrlhi (v4hi, int)
12034 v2si __builtin_arm_wsrlw (v2si, long long)
12035 v2si __builtin_arm_wsrlwi (v2si, int)
12036 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12037 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12038 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12039 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12040 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12041 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12042 v2si __builtin_arm_wsubw (v2si, v2si)
12043 v2si __builtin_arm_wsubwss (v2si, v2si)
12044 v2si __builtin_arm_wsubwus (v2si, v2si)
12045 v4hi __builtin_arm_wunpckehsb (v8qi)
12046 v2si __builtin_arm_wunpckehsh (v4hi)
12047 long long __builtin_arm_wunpckehsw (v2si)
12048 v4hi __builtin_arm_wunpckehub (v8qi)
12049 v2si __builtin_arm_wunpckehuh (v4hi)
12050 long long __builtin_arm_wunpckehuw (v2si)
12051 v4hi __builtin_arm_wunpckelsb (v8qi)
12052 v2si __builtin_arm_wunpckelsh (v4hi)
12053 long long __builtin_arm_wunpckelsw (v2si)
12054 v4hi __builtin_arm_wunpckelub (v8qi)
12055 v2si __builtin_arm_wunpckeluh (v4hi)
12056 long long __builtin_arm_wunpckeluw (v2si)
12057 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12058 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12059 v2si __builtin_arm_wunpckihw (v2si, v2si)
12060 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12061 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12062 v2si __builtin_arm_wunpckilw (v2si, v2si)
12063 long long __builtin_arm_wxor (long long, long long)
12064 long long __builtin_arm_wzero ()
12068 @node ARM C Language Extensions (ACLE)
12069 @subsection ARM C Language Extensions (ACLE)
12071 GCC implements extensions for C as described in the ARM C Language
12072 Extensions (ACLE) specification, which can be found at
12073 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12075 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12076 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12077 intrinsics can be found at
12078 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12079 The built-in intrinsics for the Advanced SIMD extension are available when
12082 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12083 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12084 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12085 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12088 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12089 availability of extensions.
12091 @node ARM Floating Point Status and Control Intrinsics
12092 @subsection ARM Floating Point Status and Control Intrinsics
12094 These built-in functions are available for the ARM family of
12095 processors with floating-point unit.
12098 unsigned int __builtin_arm_get_fpscr ()
12099 void __builtin_arm_set_fpscr (unsigned int)
12102 @node AVR Built-in Functions
12103 @subsection AVR Built-in Functions
12105 For each built-in function for AVR, there is an equally named,
12106 uppercase built-in macro defined. That way users can easily query if
12107 or if not a specific built-in is implemented or not. For example, if
12108 @code{__builtin_avr_nop} is available the macro
12109 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12111 The following built-in functions map to the respective machine
12112 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12113 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12114 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12115 as library call if no hardware multiplier is available.
12118 void __builtin_avr_nop (void)
12119 void __builtin_avr_sei (void)
12120 void __builtin_avr_cli (void)
12121 void __builtin_avr_sleep (void)
12122 void __builtin_avr_wdr (void)
12123 unsigned char __builtin_avr_swap (unsigned char)
12124 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12125 int __builtin_avr_fmuls (char, char)
12126 int __builtin_avr_fmulsu (char, unsigned char)
12129 In order to delay execution for a specific number of cycles, GCC
12132 void __builtin_avr_delay_cycles (unsigned long ticks)
12136 @code{ticks} is the number of ticks to delay execution. Note that this
12137 built-in does not take into account the effect of interrupts that
12138 might increase delay time. @code{ticks} must be a compile-time
12139 integer constant; delays with a variable number of cycles are not supported.
12142 char __builtin_avr_flash_segment (const __memx void*)
12146 This built-in takes a byte address to the 24-bit
12147 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12148 the number of the flash segment (the 64 KiB chunk) where the address
12149 points to. Counting starts at @code{0}.
12150 If the address does not point to flash memory, return @code{-1}.
12153 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12157 Insert bits from @var{bits} into @var{val} and return the resulting
12158 value. The nibbles of @var{map} determine how the insertion is
12159 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12161 @item If @var{X} is @code{0xf},
12162 then the @var{n}-th bit of @var{val} is returned unaltered.
12164 @item If X is in the range 0@dots{}7,
12165 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12167 @item If X is in the range 8@dots{}@code{0xe},
12168 then the @var{n}-th result bit is undefined.
12172 One typical use case for this built-in is adjusting input and
12173 output values to non-contiguous port layouts. Some examples:
12176 // same as val, bits is unused
12177 __builtin_avr_insert_bits (0xffffffff, bits, val)
12181 // same as bits, val is unused
12182 __builtin_avr_insert_bits (0x76543210, bits, val)
12186 // same as rotating bits by 4
12187 __builtin_avr_insert_bits (0x32107654, bits, 0)
12191 // high nibble of result is the high nibble of val
12192 // low nibble of result is the low nibble of bits
12193 __builtin_avr_insert_bits (0xffff3210, bits, val)
12197 // reverse the bit order of bits
12198 __builtin_avr_insert_bits (0x01234567, bits, 0)
12201 @node Blackfin Built-in Functions
12202 @subsection Blackfin Built-in Functions
12204 Currently, there are two Blackfin-specific built-in functions. These are
12205 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12206 using inline assembly; by using these built-in functions the compiler can
12207 automatically add workarounds for hardware errata involving these
12208 instructions. These functions are named as follows:
12211 void __builtin_bfin_csync (void)
12212 void __builtin_bfin_ssync (void)
12215 @node FR-V Built-in Functions
12216 @subsection FR-V Built-in Functions
12218 GCC provides many FR-V-specific built-in functions. In general,
12219 these functions are intended to be compatible with those described
12220 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12221 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12222 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12223 pointer rather than by value.
12225 Most of the functions are named after specific FR-V instructions.
12226 Such functions are said to be ``directly mapped'' and are summarized
12227 here in tabular form.
12231 * Directly-mapped Integer Functions::
12232 * Directly-mapped Media Functions::
12233 * Raw read/write Functions::
12234 * Other Built-in Functions::
12237 @node Argument Types
12238 @subsubsection Argument Types
12240 The arguments to the built-in functions can be divided into three groups:
12241 register numbers, compile-time constants and run-time values. In order
12242 to make this classification clear at a glance, the arguments and return
12243 values are given the following pseudo types:
12245 @multitable @columnfractions .20 .30 .15 .35
12246 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12247 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12248 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12249 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12250 @item @code{uw2} @tab @code{unsigned long long} @tab No
12251 @tab an unsigned doubleword
12252 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12253 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12254 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12255 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12258 These pseudo types are not defined by GCC, they are simply a notational
12259 convenience used in this manual.
12261 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12262 and @code{sw2} are evaluated at run time. They correspond to
12263 register operands in the underlying FR-V instructions.
12265 @code{const} arguments represent immediate operands in the underlying
12266 FR-V instructions. They must be compile-time constants.
12268 @code{acc} arguments are evaluated at compile time and specify the number
12269 of an accumulator register. For example, an @code{acc} argument of 2
12270 selects the ACC2 register.
12272 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12273 number of an IACC register. See @pxref{Other Built-in Functions}
12276 @node Directly-mapped Integer Functions
12277 @subsubsection Directly-Mapped Integer Functions
12279 The functions listed below map directly to FR-V I-type instructions.
12281 @multitable @columnfractions .45 .32 .23
12282 @item Function prototype @tab Example usage @tab Assembly output
12283 @item @code{sw1 __ADDSS (sw1, sw1)}
12284 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12285 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12286 @item @code{sw1 __SCAN (sw1, sw1)}
12287 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12288 @tab @code{SCAN @var{a},@var{b},@var{c}}
12289 @item @code{sw1 __SCUTSS (sw1)}
12290 @tab @code{@var{b} = __SCUTSS (@var{a})}
12291 @tab @code{SCUTSS @var{a},@var{b}}
12292 @item @code{sw1 __SLASS (sw1, sw1)}
12293 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12294 @tab @code{SLASS @var{a},@var{b},@var{c}}
12295 @item @code{void __SMASS (sw1, sw1)}
12296 @tab @code{__SMASS (@var{a}, @var{b})}
12297 @tab @code{SMASS @var{a},@var{b}}
12298 @item @code{void __SMSSS (sw1, sw1)}
12299 @tab @code{__SMSSS (@var{a}, @var{b})}
12300 @tab @code{SMSSS @var{a},@var{b}}
12301 @item @code{void __SMU (sw1, sw1)}
12302 @tab @code{__SMU (@var{a}, @var{b})}
12303 @tab @code{SMU @var{a},@var{b}}
12304 @item @code{sw2 __SMUL (sw1, sw1)}
12305 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12306 @tab @code{SMUL @var{a},@var{b},@var{c}}
12307 @item @code{sw1 __SUBSS (sw1, sw1)}
12308 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12309 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12310 @item @code{uw2 __UMUL (uw1, uw1)}
12311 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12312 @tab @code{UMUL @var{a},@var{b},@var{c}}
12315 @node Directly-mapped Media Functions
12316 @subsubsection Directly-Mapped Media Functions
12318 The functions listed below map directly to FR-V M-type instructions.
12320 @multitable @columnfractions .45 .32 .23
12321 @item Function prototype @tab Example usage @tab Assembly output
12322 @item @code{uw1 __MABSHS (sw1)}
12323 @tab @code{@var{b} = __MABSHS (@var{a})}
12324 @tab @code{MABSHS @var{a},@var{b}}
12325 @item @code{void __MADDACCS (acc, acc)}
12326 @tab @code{__MADDACCS (@var{b}, @var{a})}
12327 @tab @code{MADDACCS @var{a},@var{b}}
12328 @item @code{sw1 __MADDHSS (sw1, sw1)}
12329 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12330 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12331 @item @code{uw1 __MADDHUS (uw1, uw1)}
12332 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12333 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12334 @item @code{uw1 __MAND (uw1, uw1)}
12335 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12336 @tab @code{MAND @var{a},@var{b},@var{c}}
12337 @item @code{void __MASACCS (acc, acc)}
12338 @tab @code{__MASACCS (@var{b}, @var{a})}
12339 @tab @code{MASACCS @var{a},@var{b}}
12340 @item @code{uw1 __MAVEH (uw1, uw1)}
12341 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12342 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12343 @item @code{uw2 __MBTOH (uw1)}
12344 @tab @code{@var{b} = __MBTOH (@var{a})}
12345 @tab @code{MBTOH @var{a},@var{b}}
12346 @item @code{void __MBTOHE (uw1 *, uw1)}
12347 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12348 @tab @code{MBTOHE @var{a},@var{b}}
12349 @item @code{void __MCLRACC (acc)}
12350 @tab @code{__MCLRACC (@var{a})}
12351 @tab @code{MCLRACC @var{a}}
12352 @item @code{void __MCLRACCA (void)}
12353 @tab @code{__MCLRACCA ()}
12354 @tab @code{MCLRACCA}
12355 @item @code{uw1 __Mcop1 (uw1, uw1)}
12356 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12357 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12358 @item @code{uw1 __Mcop2 (uw1, uw1)}
12359 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12360 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12361 @item @code{uw1 __MCPLHI (uw2, const)}
12362 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12363 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12364 @item @code{uw1 __MCPLI (uw2, const)}
12365 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12366 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12367 @item @code{void __MCPXIS (acc, sw1, sw1)}
12368 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12369 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12370 @item @code{void __MCPXIU (acc, uw1, uw1)}
12371 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12372 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12373 @item @code{void __MCPXRS (acc, sw1, sw1)}
12374 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12375 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12376 @item @code{void __MCPXRU (acc, uw1, uw1)}
12377 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12378 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12379 @item @code{uw1 __MCUT (acc, uw1)}
12380 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12381 @tab @code{MCUT @var{a},@var{b},@var{c}}
12382 @item @code{uw1 __MCUTSS (acc, sw1)}
12383 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12384 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12385 @item @code{void __MDADDACCS (acc, acc)}
12386 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12387 @tab @code{MDADDACCS @var{a},@var{b}}
12388 @item @code{void __MDASACCS (acc, acc)}
12389 @tab @code{__MDASACCS (@var{b}, @var{a})}
12390 @tab @code{MDASACCS @var{a},@var{b}}
12391 @item @code{uw2 __MDCUTSSI (acc, const)}
12392 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12393 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12394 @item @code{uw2 __MDPACKH (uw2, uw2)}
12395 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12396 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12397 @item @code{uw2 __MDROTLI (uw2, const)}
12398 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12399 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12400 @item @code{void __MDSUBACCS (acc, acc)}
12401 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12402 @tab @code{MDSUBACCS @var{a},@var{b}}
12403 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12404 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12405 @tab @code{MDUNPACKH @var{a},@var{b}}
12406 @item @code{uw2 __MEXPDHD (uw1, const)}
12407 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12408 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12409 @item @code{uw1 __MEXPDHW (uw1, const)}
12410 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12411 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12412 @item @code{uw1 __MHDSETH (uw1, const)}
12413 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12414 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12415 @item @code{sw1 __MHDSETS (const)}
12416 @tab @code{@var{b} = __MHDSETS (@var{a})}
12417 @tab @code{MHDSETS #@var{a},@var{b}}
12418 @item @code{uw1 __MHSETHIH (uw1, const)}
12419 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12420 @tab @code{MHSETHIH #@var{a},@var{b}}
12421 @item @code{sw1 __MHSETHIS (sw1, const)}
12422 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12423 @tab @code{MHSETHIS #@var{a},@var{b}}
12424 @item @code{uw1 __MHSETLOH (uw1, const)}
12425 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12426 @tab @code{MHSETLOH #@var{a},@var{b}}
12427 @item @code{sw1 __MHSETLOS (sw1, const)}
12428 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12429 @tab @code{MHSETLOS #@var{a},@var{b}}
12430 @item @code{uw1 __MHTOB (uw2)}
12431 @tab @code{@var{b} = __MHTOB (@var{a})}
12432 @tab @code{MHTOB @var{a},@var{b}}
12433 @item @code{void __MMACHS (acc, sw1, sw1)}
12434 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12435 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12436 @item @code{void __MMACHU (acc, uw1, uw1)}
12437 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12438 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12439 @item @code{void __MMRDHS (acc, sw1, sw1)}
12440 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12441 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12442 @item @code{void __MMRDHU (acc, uw1, uw1)}
12443 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12444 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12445 @item @code{void __MMULHS (acc, sw1, sw1)}
12446 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12447 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12448 @item @code{void __MMULHU (acc, uw1, uw1)}
12449 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12450 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12451 @item @code{void __MMULXHS (acc, sw1, sw1)}
12452 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12453 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12454 @item @code{void __MMULXHU (acc, uw1, uw1)}
12455 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12456 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12457 @item @code{uw1 __MNOT (uw1)}
12458 @tab @code{@var{b} = __MNOT (@var{a})}
12459 @tab @code{MNOT @var{a},@var{b}}
12460 @item @code{uw1 __MOR (uw1, uw1)}
12461 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12462 @tab @code{MOR @var{a},@var{b},@var{c}}
12463 @item @code{uw1 __MPACKH (uh, uh)}
12464 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12465 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12466 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12467 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12468 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12469 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12470 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12471 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12472 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12473 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12474 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12475 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12476 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12477 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12478 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12479 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12480 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12481 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12482 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12483 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12484 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12485 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12486 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12487 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12488 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12489 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12490 @item @code{void __MQMACHS (acc, sw2, sw2)}
12491 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12492 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12493 @item @code{void __MQMACHU (acc, uw2, uw2)}
12494 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12495 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12496 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12497 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12498 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12499 @item @code{void __MQMULHS (acc, sw2, sw2)}
12500 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12501 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12502 @item @code{void __MQMULHU (acc, uw2, uw2)}
12503 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12504 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12505 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12506 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12507 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12508 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12509 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12510 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12511 @item @code{sw2 __MQSATHS (sw2, sw2)}
12512 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12513 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12514 @item @code{uw2 __MQSLLHI (uw2, int)}
12515 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12516 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12517 @item @code{sw2 __MQSRAHI (sw2, int)}
12518 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12519 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12520 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12521 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12522 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12523 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12524 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12525 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12526 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12527 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12528 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12529 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12530 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12531 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12532 @item @code{uw1 __MRDACC (acc)}
12533 @tab @code{@var{b} = __MRDACC (@var{a})}
12534 @tab @code{MRDACC @var{a},@var{b}}
12535 @item @code{uw1 __MRDACCG (acc)}
12536 @tab @code{@var{b} = __MRDACCG (@var{a})}
12537 @tab @code{MRDACCG @var{a},@var{b}}
12538 @item @code{uw1 __MROTLI (uw1, const)}
12539 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12540 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12541 @item @code{uw1 __MROTRI (uw1, const)}
12542 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12543 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12544 @item @code{sw1 __MSATHS (sw1, sw1)}
12545 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12546 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12547 @item @code{uw1 __MSATHU (uw1, uw1)}
12548 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12549 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12550 @item @code{uw1 __MSLLHI (uw1, const)}
12551 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12552 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12553 @item @code{sw1 __MSRAHI (sw1, const)}
12554 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12555 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12556 @item @code{uw1 __MSRLHI (uw1, const)}
12557 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12558 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12559 @item @code{void __MSUBACCS (acc, acc)}
12560 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12561 @tab @code{MSUBACCS @var{a},@var{b}}
12562 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12563 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12564 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12565 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12566 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12567 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12568 @item @code{void __MTRAP (void)}
12569 @tab @code{__MTRAP ()}
12571 @item @code{uw2 __MUNPACKH (uw1)}
12572 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12573 @tab @code{MUNPACKH @var{a},@var{b}}
12574 @item @code{uw1 __MWCUT (uw2, uw1)}
12575 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12576 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12577 @item @code{void __MWTACC (acc, uw1)}
12578 @tab @code{__MWTACC (@var{b}, @var{a})}
12579 @tab @code{MWTACC @var{a},@var{b}}
12580 @item @code{void __MWTACCG (acc, uw1)}
12581 @tab @code{__MWTACCG (@var{b}, @var{a})}
12582 @tab @code{MWTACCG @var{a},@var{b}}
12583 @item @code{uw1 __MXOR (uw1, uw1)}
12584 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12585 @tab @code{MXOR @var{a},@var{b},@var{c}}
12588 @node Raw read/write Functions
12589 @subsubsection Raw Read/Write Functions
12591 This sections describes built-in functions related to read and write
12592 instructions to access memory. These functions generate
12593 @code{membar} instructions to flush the I/O load and stores where
12594 appropriate, as described in Fujitsu's manual described above.
12598 @item unsigned char __builtin_read8 (void *@var{data})
12599 @item unsigned short __builtin_read16 (void *@var{data})
12600 @item unsigned long __builtin_read32 (void *@var{data})
12601 @item unsigned long long __builtin_read64 (void *@var{data})
12603 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12604 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12605 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12606 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12609 @node Other Built-in Functions
12610 @subsubsection Other Built-in Functions
12612 This section describes built-in functions that are not named after
12613 a specific FR-V instruction.
12616 @item sw2 __IACCreadll (iacc @var{reg})
12617 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12618 for future expansion and must be 0.
12620 @item sw1 __IACCreadl (iacc @var{reg})
12621 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12622 Other values of @var{reg} are rejected as invalid.
12624 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12625 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12626 is reserved for future expansion and must be 0.
12628 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12629 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12630 is 1. Other values of @var{reg} are rejected as invalid.
12632 @item void __data_prefetch0 (const void *@var{x})
12633 Use the @code{dcpl} instruction to load the contents of address @var{x}
12634 into the data cache.
12636 @item void __data_prefetch (const void *@var{x})
12637 Use the @code{nldub} instruction to load the contents of address @var{x}
12638 into the data cache. The instruction is issued in slot I1@.
12641 @node MIPS DSP Built-in Functions
12642 @subsection MIPS DSP Built-in Functions
12644 The MIPS DSP Application-Specific Extension (ASE) includes new
12645 instructions that are designed to improve the performance of DSP and
12646 media applications. It provides instructions that operate on packed
12647 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12649 GCC supports MIPS DSP operations using both the generic
12650 vector extensions (@pxref{Vector Extensions}) and a collection of
12651 MIPS-specific built-in functions. Both kinds of support are
12652 enabled by the @option{-mdsp} command-line option.
12654 Revision 2 of the ASE was introduced in the second half of 2006.
12655 This revision adds extra instructions to the original ASE, but is
12656 otherwise backwards-compatible with it. You can select revision 2
12657 using the command-line option @option{-mdspr2}; this option implies
12660 The SCOUNT and POS bits of the DSP control register are global. The
12661 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12662 POS bits. During optimization, the compiler does not delete these
12663 instructions and it does not delete calls to functions containing
12664 these instructions.
12666 At present, GCC only provides support for operations on 32-bit
12667 vectors. The vector type associated with 8-bit integer data is
12668 usually called @code{v4i8}, the vector type associated with Q7
12669 is usually called @code{v4q7}, the vector type associated with 16-bit
12670 integer data is usually called @code{v2i16}, and the vector type
12671 associated with Q15 is usually called @code{v2q15}. They can be
12672 defined in C as follows:
12675 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12676 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12677 typedef short v2i16 __attribute__ ((vector_size(4)));
12678 typedef short v2q15 __attribute__ ((vector_size(4)));
12681 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12682 initialized in the same way as aggregates. For example:
12685 v4i8 a = @{1, 2, 3, 4@};
12687 b = (v4i8) @{5, 6, 7, 8@};
12689 v2q15 c = @{0x0fcb, 0x3a75@};
12691 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12694 @emph{Note:} The CPU's endianness determines the order in which values
12695 are packed. On little-endian targets, the first value is the least
12696 significant and the last value is the most significant. The opposite
12697 order applies to big-endian targets. For example, the code above
12698 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12699 and @code{4} on big-endian targets.
12701 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12702 representation. As shown in this example, the integer representation
12703 of a Q7 value can be obtained by multiplying the fractional value by
12704 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12705 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12708 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12709 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12710 and @code{c} and @code{d} are @code{v2q15} values.
12712 @multitable @columnfractions .50 .50
12713 @item C code @tab MIPS instruction
12714 @item @code{a + b} @tab @code{addu.qb}
12715 @item @code{c + d} @tab @code{addq.ph}
12716 @item @code{a - b} @tab @code{subu.qb}
12717 @item @code{c - d} @tab @code{subq.ph}
12720 The table below lists the @code{v2i16} operation for which
12721 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12722 @code{v2i16} values.
12724 @multitable @columnfractions .50 .50
12725 @item C code @tab MIPS instruction
12726 @item @code{e * f} @tab @code{mul.ph}
12729 It is easier to describe the DSP built-in functions if we first define
12730 the following types:
12735 typedef unsigned int ui32;
12736 typedef long long a64;
12739 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12740 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12741 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12742 @code{long long}, but we use @code{a64} to indicate values that are
12743 placed in one of the four DSP accumulators (@code{$ac0},
12744 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12746 Also, some built-in functions prefer or require immediate numbers as
12747 parameters, because the corresponding DSP instructions accept both immediate
12748 numbers and register operands, or accept immediate numbers only. The
12749 immediate parameters are listed as follows.
12757 imm0_255: 0 to 255.
12758 imm_n32_31: -32 to 31.
12759 imm_n512_511: -512 to 511.
12762 The following built-in functions map directly to a particular MIPS DSP
12763 instruction. Please refer to the architecture specification
12764 for details on what each instruction does.
12767 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12768 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12769 q31 __builtin_mips_addq_s_w (q31, q31)
12770 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12771 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12772 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12773 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12774 q31 __builtin_mips_subq_s_w (q31, q31)
12775 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12776 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12777 i32 __builtin_mips_addsc (i32, i32)
12778 i32 __builtin_mips_addwc (i32, i32)
12779 i32 __builtin_mips_modsub (i32, i32)
12780 i32 __builtin_mips_raddu_w_qb (v4i8)
12781 v2q15 __builtin_mips_absq_s_ph (v2q15)
12782 q31 __builtin_mips_absq_s_w (q31)
12783 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12784 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12785 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12786 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12787 q31 __builtin_mips_preceq_w_phl (v2q15)
12788 q31 __builtin_mips_preceq_w_phr (v2q15)
12789 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12790 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12791 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12792 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12793 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12794 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12795 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12796 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12797 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12798 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12799 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12800 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12801 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12802 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12803 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12804 q31 __builtin_mips_shll_s_w (q31, i32)
12805 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12806 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12807 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12808 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12809 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12810 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12811 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12812 q31 __builtin_mips_shra_r_w (q31, i32)
12813 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12814 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12815 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12816 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12817 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12818 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12819 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12820 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12821 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12822 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12823 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12824 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12825 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12826 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12827 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12828 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12829 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12830 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12831 i32 __builtin_mips_bitrev (i32)
12832 i32 __builtin_mips_insv (i32, i32)
12833 v4i8 __builtin_mips_repl_qb (imm0_255)
12834 v4i8 __builtin_mips_repl_qb (i32)
12835 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12836 v2q15 __builtin_mips_repl_ph (i32)
12837 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12838 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12839 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12840 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12841 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12842 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12843 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12844 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12845 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12846 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12847 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12848 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12849 i32 __builtin_mips_extr_w (a64, imm0_31)
12850 i32 __builtin_mips_extr_w (a64, i32)
12851 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12852 i32 __builtin_mips_extr_s_h (a64, i32)
12853 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12854 i32 __builtin_mips_extr_rs_w (a64, i32)
12855 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12856 i32 __builtin_mips_extr_r_w (a64, i32)
12857 i32 __builtin_mips_extp (a64, imm0_31)
12858 i32 __builtin_mips_extp (a64, i32)
12859 i32 __builtin_mips_extpdp (a64, imm0_31)
12860 i32 __builtin_mips_extpdp (a64, i32)
12861 a64 __builtin_mips_shilo (a64, imm_n32_31)
12862 a64 __builtin_mips_shilo (a64, i32)
12863 a64 __builtin_mips_mthlip (a64, i32)
12864 void __builtin_mips_wrdsp (i32, imm0_63)
12865 i32 __builtin_mips_rddsp (imm0_63)
12866 i32 __builtin_mips_lbux (void *, i32)
12867 i32 __builtin_mips_lhx (void *, i32)
12868 i32 __builtin_mips_lwx (void *, i32)
12869 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12870 i32 __builtin_mips_bposge32 (void)
12871 a64 __builtin_mips_madd (a64, i32, i32);
12872 a64 __builtin_mips_maddu (a64, ui32, ui32);
12873 a64 __builtin_mips_msub (a64, i32, i32);
12874 a64 __builtin_mips_msubu (a64, ui32, ui32);
12875 a64 __builtin_mips_mult (i32, i32);
12876 a64 __builtin_mips_multu (ui32, ui32);
12879 The following built-in functions map directly to a particular MIPS DSP REV 2
12880 instruction. Please refer to the architecture specification
12881 for details on what each instruction does.
12884 v4q7 __builtin_mips_absq_s_qb (v4q7);
12885 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12886 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12887 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12888 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12889 i32 __builtin_mips_append (i32, i32, imm0_31);
12890 i32 __builtin_mips_balign (i32, i32, imm0_3);
12891 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12892 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12893 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12894 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12895 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12896 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12897 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12898 q31 __builtin_mips_mulq_rs_w (q31, q31);
12899 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12900 q31 __builtin_mips_mulq_s_w (q31, q31);
12901 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12902 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12903 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12904 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12905 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12906 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12907 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12908 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12909 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12910 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12911 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12912 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12913 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12914 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12915 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12916 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12917 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12918 q31 __builtin_mips_addqh_w (q31, q31);
12919 q31 __builtin_mips_addqh_r_w (q31, q31);
12920 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12921 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12922 q31 __builtin_mips_subqh_w (q31, q31);
12923 q31 __builtin_mips_subqh_r_w (q31, q31);
12924 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12925 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12926 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12927 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12928 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12929 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12933 @node MIPS Paired-Single Support
12934 @subsection MIPS Paired-Single Support
12936 The MIPS64 architecture includes a number of instructions that
12937 operate on pairs of single-precision floating-point values.
12938 Each pair is packed into a 64-bit floating-point register,
12939 with one element being designated the ``upper half'' and
12940 the other being designated the ``lower half''.
12942 GCC supports paired-single operations using both the generic
12943 vector extensions (@pxref{Vector Extensions}) and a collection of
12944 MIPS-specific built-in functions. Both kinds of support are
12945 enabled by the @option{-mpaired-single} command-line option.
12947 The vector type associated with paired-single values is usually
12948 called @code{v2sf}. It can be defined in C as follows:
12951 typedef float v2sf __attribute__ ((vector_size (8)));
12954 @code{v2sf} values are initialized in the same way as aggregates.
12958 v2sf a = @{1.5, 9.1@};
12961 b = (v2sf) @{e, f@};
12964 @emph{Note:} The CPU's endianness determines which value is stored in
12965 the upper half of a register and which value is stored in the lower half.
12966 On little-endian targets, the first value is the lower one and the second
12967 value is the upper one. The opposite order applies to big-endian targets.
12968 For example, the code above sets the lower half of @code{a} to
12969 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
12971 @node MIPS Loongson Built-in Functions
12972 @subsection MIPS Loongson Built-in Functions
12974 GCC provides intrinsics to access the SIMD instructions provided by the
12975 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
12976 available after inclusion of the @code{loongson.h} header file,
12977 operate on the following 64-bit vector types:
12980 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
12981 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
12982 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
12983 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
12984 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
12985 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
12988 The intrinsics provided are listed below; each is named after the
12989 machine instruction to which it corresponds, with suffixes added as
12990 appropriate to distinguish intrinsics that expand to the same machine
12991 instruction yet have different argument types. Refer to the architecture
12992 documentation for a description of the functionality of each
12996 int16x4_t packsswh (int32x2_t s, int32x2_t t);
12997 int8x8_t packsshb (int16x4_t s, int16x4_t t);
12998 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
12999 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13000 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13001 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13002 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13003 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13004 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13005 uint64_t paddd_u (uint64_t s, uint64_t t);
13006 int64_t paddd_s (int64_t s, int64_t t);
13007 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13008 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13009 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13010 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13011 uint64_t pandn_ud (uint64_t s, uint64_t t);
13012 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13013 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13014 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13015 int64_t pandn_sd (int64_t s, int64_t t);
13016 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13017 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13018 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13019 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13020 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13021 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13022 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13023 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13024 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13025 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13026 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13027 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13028 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13029 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13030 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13031 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13032 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13033 uint16x4_t pextrh_u (uint16x4_t s, int field);
13034 int16x4_t pextrh_s (int16x4_t s, int field);
13035 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13036 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13037 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13038 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13039 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13040 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13041 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13042 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13043 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13044 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13045 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13046 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13047 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13048 uint8x8_t pmovmskb_u (uint8x8_t s);
13049 int8x8_t pmovmskb_s (int8x8_t s);
13050 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13051 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13052 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13053 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13054 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13055 uint16x4_t biadd (uint8x8_t s);
13056 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13057 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13058 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13059 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13060 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13061 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13062 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13063 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13064 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13065 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13066 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13067 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13068 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13069 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13070 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13071 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13072 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13073 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13074 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13075 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13076 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13077 uint64_t psubd_u (uint64_t s, uint64_t t);
13078 int64_t psubd_s (int64_t s, int64_t t);
13079 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13080 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13081 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13082 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13083 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13084 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13085 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13086 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13087 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13088 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13089 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13090 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13091 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13092 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13093 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13094 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13098 * Paired-Single Arithmetic::
13099 * Paired-Single Built-in Functions::
13100 * MIPS-3D Built-in Functions::
13103 @node Paired-Single Arithmetic
13104 @subsubsection Paired-Single Arithmetic
13106 The table below lists the @code{v2sf} operations for which hardware
13107 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13108 values and @code{x} is an integral value.
13110 @multitable @columnfractions .50 .50
13111 @item C code @tab MIPS instruction
13112 @item @code{a + b} @tab @code{add.ps}
13113 @item @code{a - b} @tab @code{sub.ps}
13114 @item @code{-a} @tab @code{neg.ps}
13115 @item @code{a * b} @tab @code{mul.ps}
13116 @item @code{a * b + c} @tab @code{madd.ps}
13117 @item @code{a * b - c} @tab @code{msub.ps}
13118 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13119 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13120 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13123 Note that the multiply-accumulate instructions can be disabled
13124 using the command-line option @code{-mno-fused-madd}.
13126 @node Paired-Single Built-in Functions
13127 @subsubsection Paired-Single Built-in Functions
13129 The following paired-single functions map directly to a particular
13130 MIPS instruction. Please refer to the architecture specification
13131 for details on what each instruction does.
13134 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13135 Pair lower lower (@code{pll.ps}).
13137 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13138 Pair upper lower (@code{pul.ps}).
13140 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13141 Pair lower upper (@code{plu.ps}).
13143 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13144 Pair upper upper (@code{puu.ps}).
13146 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13147 Convert pair to paired single (@code{cvt.ps.s}).
13149 @item float __builtin_mips_cvt_s_pl (v2sf)
13150 Convert pair lower to single (@code{cvt.s.pl}).
13152 @item float __builtin_mips_cvt_s_pu (v2sf)
13153 Convert pair upper to single (@code{cvt.s.pu}).
13155 @item v2sf __builtin_mips_abs_ps (v2sf)
13156 Absolute value (@code{abs.ps}).
13158 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13159 Align variable (@code{alnv.ps}).
13161 @emph{Note:} The value of the third parameter must be 0 or 4
13162 modulo 8, otherwise the result is unpredictable. Please read the
13163 instruction description for details.
13166 The following multi-instruction functions are also available.
13167 In each case, @var{cond} can be any of the 16 floating-point conditions:
13168 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13169 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13170 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13173 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13174 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13175 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13176 @code{movt.ps}/@code{movf.ps}).
13178 The @code{movt} functions return the value @var{x} computed by:
13181 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13182 mov.ps @var{x},@var{c}
13183 movt.ps @var{x},@var{d},@var{cc}
13186 The @code{movf} functions are similar but use @code{movf.ps} instead
13189 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13190 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13191 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13192 @code{bc1t}/@code{bc1f}).
13194 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13195 and return either the upper or lower half of the result. For example:
13199 if (__builtin_mips_upper_c_eq_ps (a, b))
13200 upper_halves_are_equal ();
13202 upper_halves_are_unequal ();
13204 if (__builtin_mips_lower_c_eq_ps (a, b))
13205 lower_halves_are_equal ();
13207 lower_halves_are_unequal ();
13211 @node MIPS-3D Built-in Functions
13212 @subsubsection MIPS-3D Built-in Functions
13214 The MIPS-3D Application-Specific Extension (ASE) includes additional
13215 paired-single instructions that are designed to improve the performance
13216 of 3D graphics operations. Support for these instructions is controlled
13217 by the @option{-mips3d} command-line option.
13219 The functions listed below map directly to a particular MIPS-3D
13220 instruction. Please refer to the architecture specification for
13221 more details on what each instruction does.
13224 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13225 Reduction add (@code{addr.ps}).
13227 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13228 Reduction multiply (@code{mulr.ps}).
13230 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13231 Convert paired single to paired word (@code{cvt.pw.ps}).
13233 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13234 Convert paired word to paired single (@code{cvt.ps.pw}).
13236 @item float __builtin_mips_recip1_s (float)
13237 @itemx double __builtin_mips_recip1_d (double)
13238 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13239 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13241 @item float __builtin_mips_recip2_s (float, float)
13242 @itemx double __builtin_mips_recip2_d (double, double)
13243 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13244 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13246 @item float __builtin_mips_rsqrt1_s (float)
13247 @itemx double __builtin_mips_rsqrt1_d (double)
13248 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13249 Reduced-precision reciprocal square root (sequence step 1)
13250 (@code{rsqrt1.@var{fmt}}).
13252 @item float __builtin_mips_rsqrt2_s (float, float)
13253 @itemx double __builtin_mips_rsqrt2_d (double, double)
13254 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13255 Reduced-precision reciprocal square root (sequence step 2)
13256 (@code{rsqrt2.@var{fmt}}).
13259 The following multi-instruction functions are also available.
13260 In each case, @var{cond} can be any of the 16 floating-point conditions:
13261 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13262 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13263 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13266 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13267 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13268 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13269 @code{bc1t}/@code{bc1f}).
13271 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13272 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13277 if (__builtin_mips_cabs_eq_s (a, b))
13283 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13284 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13285 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13286 @code{bc1t}/@code{bc1f}).
13288 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13289 and return either the upper or lower half of the result. For example:
13293 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13294 upper_halves_are_equal ();
13296 upper_halves_are_unequal ();
13298 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13299 lower_halves_are_equal ();
13301 lower_halves_are_unequal ();
13304 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13305 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13306 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13307 @code{movt.ps}/@code{movf.ps}).
13309 The @code{movt} functions return the value @var{x} computed by:
13312 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13313 mov.ps @var{x},@var{c}
13314 movt.ps @var{x},@var{d},@var{cc}
13317 The @code{movf} functions are similar but use @code{movf.ps} instead
13320 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13321 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13322 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13323 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13324 Comparison of two paired-single values
13325 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13326 @code{bc1any2t}/@code{bc1any2f}).
13328 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13329 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13330 result is true and the @code{all} forms return true if both results are true.
13335 if (__builtin_mips_any_c_eq_ps (a, b))
13340 if (__builtin_mips_all_c_eq_ps (a, b))
13346 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13347 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13348 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13349 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13350 Comparison of four paired-single values
13351 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13352 @code{bc1any4t}/@code{bc1any4f}).
13354 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13355 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13356 The @code{any} forms return true if any of the four results are true
13357 and the @code{all} forms return true if all four results are true.
13362 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13367 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13374 @node Other MIPS Built-in Functions
13375 @subsection Other MIPS Built-in Functions
13377 GCC provides other MIPS-specific built-in functions:
13380 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13381 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13382 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13383 when this function is available.
13385 @item unsigned int __builtin_mips_get_fcsr (void)
13386 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13387 Get and set the contents of the floating-point control and status register
13388 (FPU control register 31). These functions are only available in hard-float
13389 code but can be called in both MIPS16 and non-MIPS16 contexts.
13391 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13392 register except the condition codes, which GCC assumes are preserved.
13395 @node MSP430 Built-in Functions
13396 @subsection MSP430 Built-in Functions
13398 GCC provides a couple of special builtin functions to aid in the
13399 writing of interrupt handlers in C.
13402 @item __bic_SR_register_on_exit (int @var{mask})
13403 This clears the indicated bits in the saved copy of the status register
13404 currently residing on the stack. This only works inside interrupt
13405 handlers and the changes to the status register will only take affect
13406 once the handler returns.
13408 @item __bis_SR_register_on_exit (int @var{mask})
13409 This sets the indicated bits in the saved copy of the status register
13410 currently residing on the stack. This only works inside interrupt
13411 handlers and the changes to the status register will only take affect
13412 once the handler returns.
13414 @item __delay_cycles (long long @var{cycles})
13415 This inserts an instruction sequence that takes exactly @var{cycles}
13416 cycles (between 0 and about 17E9) to complete. The inserted sequence
13417 may use jumps, loops, or no-ops, and does not interfere with any other
13418 instructions. Note that @var{cycles} must be a compile-time constant
13419 integer - that is, you must pass a number, not a variable that may be
13420 optimized to a constant later. The number of cycles delayed by this
13424 @node NDS32 Built-in Functions
13425 @subsection NDS32 Built-in Functions
13427 These built-in functions are available for the NDS32 target:
13429 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13430 Insert an ISYNC instruction into the instruction stream where
13431 @var{addr} is an instruction address for serialization.
13434 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13435 Insert an ISB instruction into the instruction stream.
13438 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13439 Return the content of a system register which is mapped by @var{sr}.
13442 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13443 Return the content of a user space register which is mapped by @var{usr}.
13446 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13447 Move the @var{value} to a system register which is mapped by @var{sr}.
13450 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13451 Move the @var{value} to a user space register which is mapped by @var{usr}.
13454 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13455 Enable global interrupt.
13458 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13459 Disable global interrupt.
13462 @node picoChip Built-in Functions
13463 @subsection picoChip Built-in Functions
13465 GCC provides an interface to selected machine instructions from the
13466 picoChip instruction set.
13469 @item int __builtin_sbc (int @var{value})
13470 Sign bit count. Return the number of consecutive bits in @var{value}
13471 that have the same value as the sign bit. The result is the number of
13472 leading sign bits minus one, giving the number of redundant sign bits in
13475 @item int __builtin_byteswap (int @var{value})
13476 Byte swap. Return the result of swapping the upper and lower bytes of
13479 @item int __builtin_brev (int @var{value})
13480 Bit reversal. Return the result of reversing the bits in
13481 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13484 @item int __builtin_adds (int @var{x}, int @var{y})
13485 Saturating addition. Return the result of adding @var{x} and @var{y},
13486 storing the value 32767 if the result overflows.
13488 @item int __builtin_subs (int @var{x}, int @var{y})
13489 Saturating subtraction. Return the result of subtracting @var{y} from
13490 @var{x}, storing the value @minus{}32768 if the result overflows.
13492 @item void __builtin_halt (void)
13493 Halt. The processor stops execution. This built-in is useful for
13494 implementing assertions.
13498 @node PowerPC Built-in Functions
13499 @subsection PowerPC Built-in Functions
13501 These built-in functions are available for the PowerPC family of
13504 float __builtin_recipdivf (float, float);
13505 float __builtin_rsqrtf (float);
13506 double __builtin_recipdiv (double, double);
13507 double __builtin_rsqrt (double);
13508 uint64_t __builtin_ppc_get_timebase ();
13509 unsigned long __builtin_ppc_mftb ();
13510 double __builtin_unpack_longdouble (long double, int);
13511 long double __builtin_pack_longdouble (double, double);
13514 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13515 @code{__builtin_rsqrtf} functions generate multiple instructions to
13516 implement the reciprocal sqrt functionality using reciprocal sqrt
13517 estimate instructions.
13519 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13520 functions generate multiple instructions to implement division using
13521 the reciprocal estimate instructions.
13523 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13524 functions generate instructions to read the Time Base Register. The
13525 @code{__builtin_ppc_get_timebase} function may generate multiple
13526 instructions and always returns the 64 bits of the Time Base Register.
13527 The @code{__builtin_ppc_mftb} function always generates one instruction and
13528 returns the Time Base Register value as an unsigned long, throwing away
13529 the most significant word on 32-bit environments.
13531 The following built-in functions are available for the PowerPC family
13532 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13533 or @option{-mpopcntd}):
13535 long __builtin_bpermd (long, long);
13536 int __builtin_divwe (int, int);
13537 int __builtin_divweo (int, int);
13538 unsigned int __builtin_divweu (unsigned int, unsigned int);
13539 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13540 long __builtin_divde (long, long);
13541 long __builtin_divdeo (long, long);
13542 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13543 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13544 unsigned int cdtbcd (unsigned int);
13545 unsigned int cbcdtd (unsigned int);
13546 unsigned int addg6s (unsigned int, unsigned int);
13549 The @code{__builtin_divde}, @code{__builtin_divdeo},
13550 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13551 64-bit environment support ISA 2.06 or later.
13553 The following built-in functions are available for the PowerPC family
13554 of processors when hardware decimal floating point
13555 (@option{-mhard-dfp}) is available:
13557 _Decimal64 __builtin_dxex (_Decimal64);
13558 _Decimal128 __builtin_dxexq (_Decimal128);
13559 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13560 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13561 _Decimal64 __builtin_denbcd (int, _Decimal64);
13562 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13563 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13564 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13565 _Decimal64 __builtin_dscli (_Decimal64, int);
13566 _Decimal128 __builtin_dscliq (_Decimal128, int);
13567 _Decimal64 __builtin_dscri (_Decimal64, int);
13568 _Decimal128 __builtin_dscriq (_Decimal128, int);
13569 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13570 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13573 The following built-in functions are available for the PowerPC family
13574 of processors when the Vector Scalar (vsx) instruction set is
13577 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13578 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13579 unsigned long long);
13582 @node PowerPC AltiVec/VSX Built-in Functions
13583 @subsection PowerPC AltiVec Built-in Functions
13585 GCC provides an interface for the PowerPC family of processors to access
13586 the AltiVec operations described in Motorola's AltiVec Programming
13587 Interface Manual. The interface is made available by including
13588 @code{<altivec.h>} and using @option{-maltivec} and
13589 @option{-mabi=altivec}. The interface supports the following vector
13593 vector unsigned char
13597 vector unsigned short
13598 vector signed short
13602 vector unsigned int
13608 If @option{-mvsx} is used the following additional vector types are
13612 vector unsigned long
13617 The long types are only implemented for 64-bit code generation, and
13618 the long type is only used in the floating point/integer conversion
13621 GCC's implementation of the high-level language interface available from
13622 C and C++ code differs from Motorola's documentation in several ways.
13627 A vector constant is a list of constant expressions within curly braces.
13630 A vector initializer requires no cast if the vector constant is of the
13631 same type as the variable it is initializing.
13634 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13635 vector type is the default signedness of the base type. The default
13636 varies depending on the operating system, so a portable program should
13637 always specify the signedness.
13640 Compiling with @option{-maltivec} adds keywords @code{__vector},
13641 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13642 @code{bool}. When compiling ISO C, the context-sensitive substitution
13643 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13644 disabled. To use them, you must include @code{<altivec.h>} instead.
13647 GCC allows using a @code{typedef} name as the type specifier for a
13651 For C, overloaded functions are implemented with macros so the following
13655 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13659 Since @code{vec_add} is a macro, the vector constant in the example
13660 is treated as four separate arguments. Wrap the entire argument in
13661 parentheses for this to work.
13664 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13665 Internally, GCC uses built-in functions to achieve the functionality in
13666 the aforementioned header file, but they are not supported and are
13667 subject to change without notice.
13669 The following interfaces are supported for the generic and specific
13670 AltiVec operations and the AltiVec predicates. In cases where there
13671 is a direct mapping between generic and specific operations, only the
13672 generic names are shown here, although the specific operations can also
13675 Arguments that are documented as @code{const int} require literal
13676 integral values within the range required for that operation.
13679 vector signed char vec_abs (vector signed char);
13680 vector signed short vec_abs (vector signed short);
13681 vector signed int vec_abs (vector signed int);
13682 vector float vec_abs (vector float);
13684 vector signed char vec_abss (vector signed char);
13685 vector signed short vec_abss (vector signed short);
13686 vector signed int vec_abss (vector signed int);
13688 vector signed char vec_add (vector bool char, vector signed char);
13689 vector signed char vec_add (vector signed char, vector bool char);
13690 vector signed char vec_add (vector signed char, vector signed char);
13691 vector unsigned char vec_add (vector bool char, vector unsigned char);
13692 vector unsigned char vec_add (vector unsigned char, vector bool char);
13693 vector unsigned char vec_add (vector unsigned char,
13694 vector unsigned char);
13695 vector signed short vec_add (vector bool short, vector signed short);
13696 vector signed short vec_add (vector signed short, vector bool short);
13697 vector signed short vec_add (vector signed short, vector signed short);
13698 vector unsigned short vec_add (vector bool short,
13699 vector unsigned short);
13700 vector unsigned short vec_add (vector unsigned short,
13701 vector bool short);
13702 vector unsigned short vec_add (vector unsigned short,
13703 vector unsigned short);
13704 vector signed int vec_add (vector bool int, vector signed int);
13705 vector signed int vec_add (vector signed int, vector bool int);
13706 vector signed int vec_add (vector signed int, vector signed int);
13707 vector unsigned int vec_add (vector bool int, vector unsigned int);
13708 vector unsigned int vec_add (vector unsigned int, vector bool int);
13709 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
13710 vector float vec_add (vector float, vector float);
13712 vector float vec_vaddfp (vector float, vector float);
13714 vector signed int vec_vadduwm (vector bool int, vector signed int);
13715 vector signed int vec_vadduwm (vector signed int, vector bool int);
13716 vector signed int vec_vadduwm (vector signed int, vector signed int);
13717 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
13718 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
13719 vector unsigned int vec_vadduwm (vector unsigned int,
13720 vector unsigned int);
13722 vector signed short vec_vadduhm (vector bool short,
13723 vector signed short);
13724 vector signed short vec_vadduhm (vector signed short,
13725 vector bool short);
13726 vector signed short vec_vadduhm (vector signed short,
13727 vector signed short);
13728 vector unsigned short vec_vadduhm (vector bool short,
13729 vector unsigned short);
13730 vector unsigned short vec_vadduhm (vector unsigned short,
13731 vector bool short);
13732 vector unsigned short vec_vadduhm (vector unsigned short,
13733 vector unsigned short);
13735 vector signed char vec_vaddubm (vector bool char, vector signed char);
13736 vector signed char vec_vaddubm (vector signed char, vector bool char);
13737 vector signed char vec_vaddubm (vector signed char, vector signed char);
13738 vector unsigned char vec_vaddubm (vector bool char,
13739 vector unsigned char);
13740 vector unsigned char vec_vaddubm (vector unsigned char,
13742 vector unsigned char vec_vaddubm (vector unsigned char,
13743 vector unsigned char);
13745 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
13747 vector unsigned char vec_adds (vector bool char, vector unsigned char);
13748 vector unsigned char vec_adds (vector unsigned char, vector bool char);
13749 vector unsigned char vec_adds (vector unsigned char,
13750 vector unsigned char);
13751 vector signed char vec_adds (vector bool char, vector signed char);
13752 vector signed char vec_adds (vector signed char, vector bool char);
13753 vector signed char vec_adds (vector signed char, vector signed char);
13754 vector unsigned short vec_adds (vector bool short,
13755 vector unsigned short);
13756 vector unsigned short vec_adds (vector unsigned short,
13757 vector bool short);
13758 vector unsigned short vec_adds (vector unsigned short,
13759 vector unsigned short);
13760 vector signed short vec_adds (vector bool short, vector signed short);
13761 vector signed short vec_adds (vector signed short, vector bool short);
13762 vector signed short vec_adds (vector signed short, vector signed short);
13763 vector unsigned int vec_adds (vector bool int, vector unsigned int);
13764 vector unsigned int vec_adds (vector unsigned int, vector bool int);
13765 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
13766 vector signed int vec_adds (vector bool int, vector signed int);
13767 vector signed int vec_adds (vector signed int, vector bool int);
13768 vector signed int vec_adds (vector signed int, vector signed int);
13770 vector signed int vec_vaddsws (vector bool int, vector signed int);
13771 vector signed int vec_vaddsws (vector signed int, vector bool int);
13772 vector signed int vec_vaddsws (vector signed int, vector signed int);
13774 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
13775 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
13776 vector unsigned int vec_vadduws (vector unsigned int,
13777 vector unsigned int);
13779 vector signed short vec_vaddshs (vector bool short,
13780 vector signed short);
13781 vector signed short vec_vaddshs (vector signed short,
13782 vector bool short);
13783 vector signed short vec_vaddshs (vector signed short,
13784 vector signed short);
13786 vector unsigned short vec_vadduhs (vector bool short,
13787 vector unsigned short);
13788 vector unsigned short vec_vadduhs (vector unsigned short,
13789 vector bool short);
13790 vector unsigned short vec_vadduhs (vector unsigned short,
13791 vector unsigned short);
13793 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13794 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13795 vector signed char vec_vaddsbs (vector signed char, vector signed char);
13797 vector unsigned char vec_vaddubs (vector bool char,
13798 vector unsigned char);
13799 vector unsigned char vec_vaddubs (vector unsigned char,
13801 vector unsigned char vec_vaddubs (vector unsigned char,
13802 vector unsigned char);
13804 vector float vec_and (vector float, vector float);
13805 vector float vec_and (vector float, vector bool int);
13806 vector float vec_and (vector bool int, vector float);
13807 vector bool int vec_and (vector bool int, vector bool int);
13808 vector signed int vec_and (vector bool int, vector signed int);
13809 vector signed int vec_and (vector signed int, vector bool int);
13810 vector signed int vec_and (vector signed int, vector signed int);
13811 vector unsigned int vec_and (vector bool int, vector unsigned int);
13812 vector unsigned int vec_and (vector unsigned int, vector bool int);
13813 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
13814 vector bool short vec_and (vector bool short, vector bool short);
13815 vector signed short vec_and (vector bool short, vector signed short);
13816 vector signed short vec_and (vector signed short, vector bool short);
13817 vector signed short vec_and (vector signed short, vector signed short);
13818 vector unsigned short vec_and (vector bool short,
13819 vector unsigned short);
13820 vector unsigned short vec_and (vector unsigned short,
13821 vector bool short);
13822 vector unsigned short vec_and (vector unsigned short,
13823 vector unsigned short);
13824 vector signed char vec_and (vector bool char, vector signed char);
13825 vector bool char vec_and (vector bool char, vector bool char);
13826 vector signed char vec_and (vector signed char, vector bool char);
13827 vector signed char vec_and (vector signed char, vector signed char);
13828 vector unsigned char vec_and (vector bool char, vector unsigned char);
13829 vector unsigned char vec_and (vector unsigned char, vector bool char);
13830 vector unsigned char vec_and (vector unsigned char,
13831 vector unsigned char);
13833 vector float vec_andc (vector float, vector float);
13834 vector float vec_andc (vector float, vector bool int);
13835 vector float vec_andc (vector bool int, vector float);
13836 vector bool int vec_andc (vector bool int, vector bool int);
13837 vector signed int vec_andc (vector bool int, vector signed int);
13838 vector signed int vec_andc (vector signed int, vector bool int);
13839 vector signed int vec_andc (vector signed int, vector signed int);
13840 vector unsigned int vec_andc (vector bool int, vector unsigned int);
13841 vector unsigned int vec_andc (vector unsigned int, vector bool int);
13842 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
13843 vector bool short vec_andc (vector bool short, vector bool short);
13844 vector signed short vec_andc (vector bool short, vector signed short);
13845 vector signed short vec_andc (vector signed short, vector bool short);
13846 vector signed short vec_andc (vector signed short, vector signed short);
13847 vector unsigned short vec_andc (vector bool short,
13848 vector unsigned short);
13849 vector unsigned short vec_andc (vector unsigned short,
13850 vector bool short);
13851 vector unsigned short vec_andc (vector unsigned short,
13852 vector unsigned short);
13853 vector signed char vec_andc (vector bool char, vector signed char);
13854 vector bool char vec_andc (vector bool char, vector bool char);
13855 vector signed char vec_andc (vector signed char, vector bool char);
13856 vector signed char vec_andc (vector signed char, vector signed char);
13857 vector unsigned char vec_andc (vector bool char, vector unsigned char);
13858 vector unsigned char vec_andc (vector unsigned char, vector bool char);
13859 vector unsigned char vec_andc (vector unsigned char,
13860 vector unsigned char);
13862 vector unsigned char vec_avg (vector unsigned char,
13863 vector unsigned char);
13864 vector signed char vec_avg (vector signed char, vector signed char);
13865 vector unsigned short vec_avg (vector unsigned short,
13866 vector unsigned short);
13867 vector signed short vec_avg (vector signed short, vector signed short);
13868 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
13869 vector signed int vec_avg (vector signed int, vector signed int);
13871 vector signed int vec_vavgsw (vector signed int, vector signed int);
13873 vector unsigned int vec_vavguw (vector unsigned int,
13874 vector unsigned int);
13876 vector signed short vec_vavgsh (vector signed short,
13877 vector signed short);
13879 vector unsigned short vec_vavguh (vector unsigned short,
13880 vector unsigned short);
13882 vector signed char vec_vavgsb (vector signed char, vector signed char);
13884 vector unsigned char vec_vavgub (vector unsigned char,
13885 vector unsigned char);
13887 vector float vec_copysign (vector float);
13889 vector float vec_ceil (vector float);
13891 vector signed int vec_cmpb (vector float, vector float);
13893 vector bool char vec_cmpeq (vector signed char, vector signed char);
13894 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
13895 vector bool short vec_cmpeq (vector signed short, vector signed short);
13896 vector bool short vec_cmpeq (vector unsigned short,
13897 vector unsigned short);
13898 vector bool int vec_cmpeq (vector signed int, vector signed int);
13899 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
13900 vector bool int vec_cmpeq (vector float, vector float);
13902 vector bool int vec_vcmpeqfp (vector float, vector float);
13904 vector bool int vec_vcmpequw (vector signed int, vector signed int);
13905 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
13907 vector bool short vec_vcmpequh (vector signed short,
13908 vector signed short);
13909 vector bool short vec_vcmpequh (vector unsigned short,
13910 vector unsigned short);
13912 vector bool char vec_vcmpequb (vector signed char, vector signed char);
13913 vector bool char vec_vcmpequb (vector unsigned char,
13914 vector unsigned char);
13916 vector bool int vec_cmpge (vector float, vector float);
13918 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
13919 vector bool char vec_cmpgt (vector signed char, vector signed char);
13920 vector bool short vec_cmpgt (vector unsigned short,
13921 vector unsigned short);
13922 vector bool short vec_cmpgt (vector signed short, vector signed short);
13923 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
13924 vector bool int vec_cmpgt (vector signed int, vector signed int);
13925 vector bool int vec_cmpgt (vector float, vector float);
13927 vector bool int vec_vcmpgtfp (vector float, vector float);
13929 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
13931 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
13933 vector bool short vec_vcmpgtsh (vector signed short,
13934 vector signed short);
13936 vector bool short vec_vcmpgtuh (vector unsigned short,
13937 vector unsigned short);
13939 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
13941 vector bool char vec_vcmpgtub (vector unsigned char,
13942 vector unsigned char);
13944 vector bool int vec_cmple (vector float, vector float);
13946 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
13947 vector bool char vec_cmplt (vector signed char, vector signed char);
13948 vector bool short vec_cmplt (vector unsigned short,
13949 vector unsigned short);
13950 vector bool short vec_cmplt (vector signed short, vector signed short);
13951 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
13952 vector bool int vec_cmplt (vector signed int, vector signed int);
13953 vector bool int vec_cmplt (vector float, vector float);
13955 vector float vec_cpsgn (vector float, vector float);
13957 vector float vec_ctf (vector unsigned int, const int);
13958 vector float vec_ctf (vector signed int, const int);
13959 vector double vec_ctf (vector unsigned long, const int);
13960 vector double vec_ctf (vector signed long, const int);
13962 vector float vec_vcfsx (vector signed int, const int);
13964 vector float vec_vcfux (vector unsigned int, const int);
13966 vector signed int vec_cts (vector float, const int);
13967 vector signed long vec_cts (vector double, const int);
13969 vector unsigned int vec_ctu (vector float, const int);
13970 vector unsigned long vec_ctu (vector double, const int);
13972 void vec_dss (const int);
13974 void vec_dssall (void);
13976 void vec_dst (const vector unsigned char *, int, const int);
13977 void vec_dst (const vector signed char *, int, const int);
13978 void vec_dst (const vector bool char *, int, const int);
13979 void vec_dst (const vector unsigned short *, int, const int);
13980 void vec_dst (const vector signed short *, int, const int);
13981 void vec_dst (const vector bool short *, int, const int);
13982 void vec_dst (const vector pixel *, int, const int);
13983 void vec_dst (const vector unsigned int *, int, const int);
13984 void vec_dst (const vector signed int *, int, const int);
13985 void vec_dst (const vector bool int *, int, const int);
13986 void vec_dst (const vector float *, int, const int);
13987 void vec_dst (const unsigned char *, int, const int);
13988 void vec_dst (const signed char *, int, const int);
13989 void vec_dst (const unsigned short *, int, const int);
13990 void vec_dst (const short *, int, const int);
13991 void vec_dst (const unsigned int *, int, const int);
13992 void vec_dst (const int *, int, const int);
13993 void vec_dst (const unsigned long *, int, const int);
13994 void vec_dst (const long *, int, const int);
13995 void vec_dst (const float *, int, const int);
13997 void vec_dstst (const vector unsigned char *, int, const int);
13998 void vec_dstst (const vector signed char *, int, const int);
13999 void vec_dstst (const vector bool char *, int, const int);
14000 void vec_dstst (const vector unsigned short *, int, const int);
14001 void vec_dstst (const vector signed short *, int, const int);
14002 void vec_dstst (const vector bool short *, int, const int);
14003 void vec_dstst (const vector pixel *, int, const int);
14004 void vec_dstst (const vector unsigned int *, int, const int);
14005 void vec_dstst (const vector signed int *, int, const int);
14006 void vec_dstst (const vector bool int *, int, const int);
14007 void vec_dstst (const vector float *, int, const int);
14008 void vec_dstst (const unsigned char *, int, const int);
14009 void vec_dstst (const signed char *, int, const int);
14010 void vec_dstst (const unsigned short *, int, const int);
14011 void vec_dstst (const short *, int, const int);
14012 void vec_dstst (const unsigned int *, int, const int);
14013 void vec_dstst (const int *, int, const int);
14014 void vec_dstst (const unsigned long *, int, const int);
14015 void vec_dstst (const long *, int, const int);
14016 void vec_dstst (const float *, int, const int);
14018 void vec_dststt (const vector unsigned char *, int, const int);
14019 void vec_dststt (const vector signed char *, int, const int);
14020 void vec_dststt (const vector bool char *, int, const int);
14021 void vec_dststt (const vector unsigned short *, int, const int);
14022 void vec_dststt (const vector signed short *, int, const int);
14023 void vec_dststt (const vector bool short *, int, const int);
14024 void vec_dststt (const vector pixel *, int, const int);
14025 void vec_dststt (const vector unsigned int *, int, const int);
14026 void vec_dststt (const vector signed int *, int, const int);
14027 void vec_dststt (const vector bool int *, int, const int);
14028 void vec_dststt (const vector float *, int, const int);
14029 void vec_dststt (const unsigned char *, int, const int);
14030 void vec_dststt (const signed char *, int, const int);
14031 void vec_dststt (const unsigned short *, int, const int);
14032 void vec_dststt (const short *, int, const int);
14033 void vec_dststt (const unsigned int *, int, const int);
14034 void vec_dststt (const int *, int, const int);
14035 void vec_dststt (const unsigned long *, int, const int);
14036 void vec_dststt (const long *, int, const int);
14037 void vec_dststt (const float *, int, const int);
14039 void vec_dstt (const vector unsigned char *, int, const int);
14040 void vec_dstt (const vector signed char *, int, const int);
14041 void vec_dstt (const vector bool char *, int, const int);
14042 void vec_dstt (const vector unsigned short *, int, const int);
14043 void vec_dstt (const vector signed short *, int, const int);
14044 void vec_dstt (const vector bool short *, int, const int);
14045 void vec_dstt (const vector pixel *, int, const int);
14046 void vec_dstt (const vector unsigned int *, int, const int);
14047 void vec_dstt (const vector signed int *, int, const int);
14048 void vec_dstt (const vector bool int *, int, const int);
14049 void vec_dstt (const vector float *, int, const int);
14050 void vec_dstt (const unsigned char *, int, const int);
14051 void vec_dstt (const signed char *, int, const int);
14052 void vec_dstt (const unsigned short *, int, const int);
14053 void vec_dstt (const short *, int, const int);
14054 void vec_dstt (const unsigned int *, int, const int);
14055 void vec_dstt (const int *, int, const int);
14056 void vec_dstt (const unsigned long *, int, const int);
14057 void vec_dstt (const long *, int, const int);
14058 void vec_dstt (const float *, int, const int);
14060 vector float vec_expte (vector float);
14062 vector float vec_floor (vector float);
14064 vector float vec_ld (int, const vector float *);
14065 vector float vec_ld (int, const float *);
14066 vector bool int vec_ld (int, const vector bool int *);
14067 vector signed int vec_ld (int, const vector signed int *);
14068 vector signed int vec_ld (int, const int *);
14069 vector signed int vec_ld (int, const long *);
14070 vector unsigned int vec_ld (int, const vector unsigned int *);
14071 vector unsigned int vec_ld (int, const unsigned int *);
14072 vector unsigned int vec_ld (int, const unsigned long *);
14073 vector bool short vec_ld (int, const vector bool short *);
14074 vector pixel vec_ld (int, const vector pixel *);
14075 vector signed short vec_ld (int, const vector signed short *);
14076 vector signed short vec_ld (int, const short *);
14077 vector unsigned short vec_ld (int, const vector unsigned short *);
14078 vector unsigned short vec_ld (int, const unsigned short *);
14079 vector bool char vec_ld (int, const vector bool char *);
14080 vector signed char vec_ld (int, const vector signed char *);
14081 vector signed char vec_ld (int, const signed char *);
14082 vector unsigned char vec_ld (int, const vector unsigned char *);
14083 vector unsigned char vec_ld (int, const unsigned char *);
14085 vector signed char vec_lde (int, const signed char *);
14086 vector unsigned char vec_lde (int, const unsigned char *);
14087 vector signed short vec_lde (int, const short *);
14088 vector unsigned short vec_lde (int, const unsigned short *);
14089 vector float vec_lde (int, const float *);
14090 vector signed int vec_lde (int, const int *);
14091 vector unsigned int vec_lde (int, const unsigned int *);
14092 vector signed int vec_lde (int, const long *);
14093 vector unsigned int vec_lde (int, const unsigned long *);
14095 vector float vec_lvewx (int, float *);
14096 vector signed int vec_lvewx (int, int *);
14097 vector unsigned int vec_lvewx (int, unsigned int *);
14098 vector signed int vec_lvewx (int, long *);
14099 vector unsigned int vec_lvewx (int, unsigned long *);
14101 vector signed short vec_lvehx (int, short *);
14102 vector unsigned short vec_lvehx (int, unsigned short *);
14104 vector signed char vec_lvebx (int, char *);
14105 vector unsigned char vec_lvebx (int, unsigned char *);
14107 vector float vec_ldl (int, const vector float *);
14108 vector float vec_ldl (int, const float *);
14109 vector bool int vec_ldl (int, const vector bool int *);
14110 vector signed int vec_ldl (int, const vector signed int *);
14111 vector signed int vec_ldl (int, const int *);
14112 vector signed int vec_ldl (int, const long *);
14113 vector unsigned int vec_ldl (int, const vector unsigned int *);
14114 vector unsigned int vec_ldl (int, const unsigned int *);
14115 vector unsigned int vec_ldl (int, const unsigned long *);
14116 vector bool short vec_ldl (int, const vector bool short *);
14117 vector pixel vec_ldl (int, const vector pixel *);
14118 vector signed short vec_ldl (int, const vector signed short *);
14119 vector signed short vec_ldl (int, const short *);
14120 vector unsigned short vec_ldl (int, const vector unsigned short *);
14121 vector unsigned short vec_ldl (int, const unsigned short *);
14122 vector bool char vec_ldl (int, const vector bool char *);
14123 vector signed char vec_ldl (int, const vector signed char *);
14124 vector signed char vec_ldl (int, const signed char *);
14125 vector unsigned char vec_ldl (int, const vector unsigned char *);
14126 vector unsigned char vec_ldl (int, const unsigned char *);
14128 vector float vec_loge (vector float);
14130 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
14131 vector unsigned char vec_lvsl (int, const volatile signed char *);
14132 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
14133 vector unsigned char vec_lvsl (int, const volatile short *);
14134 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
14135 vector unsigned char vec_lvsl (int, const volatile int *);
14136 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
14137 vector unsigned char vec_lvsl (int, const volatile long *);
14138 vector unsigned char vec_lvsl (int, const volatile float *);
14140 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
14141 vector unsigned char vec_lvsr (int, const volatile signed char *);
14142 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
14143 vector unsigned char vec_lvsr (int, const volatile short *);
14144 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
14145 vector unsigned char vec_lvsr (int, const volatile int *);
14146 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
14147 vector unsigned char vec_lvsr (int, const volatile long *);
14148 vector unsigned char vec_lvsr (int, const volatile float *);
14150 vector float vec_madd (vector float, vector float, vector float);
14152 vector signed short vec_madds (vector signed short,
14153 vector signed short,
14154 vector signed short);
14156 vector unsigned char vec_max (vector bool char, vector unsigned char);
14157 vector unsigned char vec_max (vector unsigned char, vector bool char);
14158 vector unsigned char vec_max (vector unsigned char,
14159 vector unsigned char);
14160 vector signed char vec_max (vector bool char, vector signed char);
14161 vector signed char vec_max (vector signed char, vector bool char);
14162 vector signed char vec_max (vector signed char, vector signed char);
14163 vector unsigned short vec_max (vector bool short,
14164 vector unsigned short);
14165 vector unsigned short vec_max (vector unsigned short,
14166 vector bool short);
14167 vector unsigned short vec_max (vector unsigned short,
14168 vector unsigned short);
14169 vector signed short vec_max (vector bool short, vector signed short);
14170 vector signed short vec_max (vector signed short, vector bool short);
14171 vector signed short vec_max (vector signed short, vector signed short);
14172 vector unsigned int vec_max (vector bool int, vector unsigned int);
14173 vector unsigned int vec_max (vector unsigned int, vector bool int);
14174 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
14175 vector signed int vec_max (vector bool int, vector signed int);
14176 vector signed int vec_max (vector signed int, vector bool int);
14177 vector signed int vec_max (vector signed int, vector signed int);
14178 vector float vec_max (vector float, vector float);
14180 vector float vec_vmaxfp (vector float, vector float);
14182 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14183 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14184 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14186 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14187 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14188 vector unsigned int vec_vmaxuw (vector unsigned int,
14189 vector unsigned int);
14191 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14192 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14193 vector signed short vec_vmaxsh (vector signed short,
14194 vector signed short);
14196 vector unsigned short vec_vmaxuh (vector bool short,
14197 vector unsigned short);
14198 vector unsigned short vec_vmaxuh (vector unsigned short,
14199 vector bool short);
14200 vector unsigned short vec_vmaxuh (vector unsigned short,
14201 vector unsigned short);
14203 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14204 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14205 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14207 vector unsigned char vec_vmaxub (vector bool char,
14208 vector unsigned char);
14209 vector unsigned char vec_vmaxub (vector unsigned char,
14211 vector unsigned char vec_vmaxub (vector unsigned char,
14212 vector unsigned char);
14214 vector bool char vec_mergeh (vector bool char, vector bool char);
14215 vector signed char vec_mergeh (vector signed char, vector signed char);
14216 vector unsigned char vec_mergeh (vector unsigned char,
14217 vector unsigned char);
14218 vector bool short vec_mergeh (vector bool short, vector bool short);
14219 vector pixel vec_mergeh (vector pixel, vector pixel);
14220 vector signed short vec_mergeh (vector signed short,
14221 vector signed short);
14222 vector unsigned short vec_mergeh (vector unsigned short,
14223 vector unsigned short);
14224 vector float vec_mergeh (vector float, vector float);
14225 vector bool int vec_mergeh (vector bool int, vector bool int);
14226 vector signed int vec_mergeh (vector signed int, vector signed int);
14227 vector unsigned int vec_mergeh (vector unsigned int,
14228 vector unsigned int);
14230 vector float vec_vmrghw (vector float, vector float);
14231 vector bool int vec_vmrghw (vector bool int, vector bool int);
14232 vector signed int vec_vmrghw (vector signed int, vector signed int);
14233 vector unsigned int vec_vmrghw (vector unsigned int,
14234 vector unsigned int);
14236 vector bool short vec_vmrghh (vector bool short, vector bool short);
14237 vector signed short vec_vmrghh (vector signed short,
14238 vector signed short);
14239 vector unsigned short vec_vmrghh (vector unsigned short,
14240 vector unsigned short);
14241 vector pixel vec_vmrghh (vector pixel, vector pixel);
14243 vector bool char vec_vmrghb (vector bool char, vector bool char);
14244 vector signed char vec_vmrghb (vector signed char, vector signed char);
14245 vector unsigned char vec_vmrghb (vector unsigned char,
14246 vector unsigned char);
14248 vector bool char vec_mergel (vector bool char, vector bool char);
14249 vector signed char vec_mergel (vector signed char, vector signed char);
14250 vector unsigned char vec_mergel (vector unsigned char,
14251 vector unsigned char);
14252 vector bool short vec_mergel (vector bool short, vector bool short);
14253 vector pixel vec_mergel (vector pixel, vector pixel);
14254 vector signed short vec_mergel (vector signed short,
14255 vector signed short);
14256 vector unsigned short vec_mergel (vector unsigned short,
14257 vector unsigned short);
14258 vector float vec_mergel (vector float, vector float);
14259 vector bool int vec_mergel (vector bool int, vector bool int);
14260 vector signed int vec_mergel (vector signed int, vector signed int);
14261 vector unsigned int vec_mergel (vector unsigned int,
14262 vector unsigned int);
14264 vector float vec_vmrglw (vector float, vector float);
14265 vector signed int vec_vmrglw (vector signed int, vector signed int);
14266 vector unsigned int vec_vmrglw (vector unsigned int,
14267 vector unsigned int);
14268 vector bool int vec_vmrglw (vector bool int, vector bool int);
14270 vector bool short vec_vmrglh (vector bool short, vector bool short);
14271 vector signed short vec_vmrglh (vector signed short,
14272 vector signed short);
14273 vector unsigned short vec_vmrglh (vector unsigned short,
14274 vector unsigned short);
14275 vector pixel vec_vmrglh (vector pixel, vector pixel);
14277 vector bool char vec_vmrglb (vector bool char, vector bool char);
14278 vector signed char vec_vmrglb (vector signed char, vector signed char);
14279 vector unsigned char vec_vmrglb (vector unsigned char,
14280 vector unsigned char);
14282 vector unsigned short vec_mfvscr (void);
14284 vector unsigned char vec_min (vector bool char, vector unsigned char);
14285 vector unsigned char vec_min (vector unsigned char, vector bool char);
14286 vector unsigned char vec_min (vector unsigned char,
14287 vector unsigned char);
14288 vector signed char vec_min (vector bool char, vector signed char);
14289 vector signed char vec_min (vector signed char, vector bool char);
14290 vector signed char vec_min (vector signed char, vector signed char);
14291 vector unsigned short vec_min (vector bool short,
14292 vector unsigned short);
14293 vector unsigned short vec_min (vector unsigned short,
14294 vector bool short);
14295 vector unsigned short vec_min (vector unsigned short,
14296 vector unsigned short);
14297 vector signed short vec_min (vector bool short, vector signed short);
14298 vector signed short vec_min (vector signed short, vector bool short);
14299 vector signed short vec_min (vector signed short, vector signed short);
14300 vector unsigned int vec_min (vector bool int, vector unsigned int);
14301 vector unsigned int vec_min (vector unsigned int, vector bool int);
14302 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14303 vector signed int vec_min (vector bool int, vector signed int);
14304 vector signed int vec_min (vector signed int, vector bool int);
14305 vector signed int vec_min (vector signed int, vector signed int);
14306 vector float vec_min (vector float, vector float);
14308 vector float vec_vminfp (vector float, vector float);
14310 vector signed int vec_vminsw (vector bool int, vector signed int);
14311 vector signed int vec_vminsw (vector signed int, vector bool int);
14312 vector signed int vec_vminsw (vector signed int, vector signed int);
14314 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14315 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14316 vector unsigned int vec_vminuw (vector unsigned int,
14317 vector unsigned int);
14319 vector signed short vec_vminsh (vector bool short, vector signed short);
14320 vector signed short vec_vminsh (vector signed short, vector bool short);
14321 vector signed short vec_vminsh (vector signed short,
14322 vector signed short);
14324 vector unsigned short vec_vminuh (vector bool short,
14325 vector unsigned short);
14326 vector unsigned short vec_vminuh (vector unsigned short,
14327 vector bool short);
14328 vector unsigned short vec_vminuh (vector unsigned short,
14329 vector unsigned short);
14331 vector signed char vec_vminsb (vector bool char, vector signed char);
14332 vector signed char vec_vminsb (vector signed char, vector bool char);
14333 vector signed char vec_vminsb (vector signed char, vector signed char);
14335 vector unsigned char vec_vminub (vector bool char,
14336 vector unsigned char);
14337 vector unsigned char vec_vminub (vector unsigned char,
14339 vector unsigned char vec_vminub (vector unsigned char,
14340 vector unsigned char);
14342 vector signed short vec_mladd (vector signed short,
14343 vector signed short,
14344 vector signed short);
14345 vector signed short vec_mladd (vector signed short,
14346 vector unsigned short,
14347 vector unsigned short);
14348 vector signed short vec_mladd (vector unsigned short,
14349 vector signed short,
14350 vector signed short);
14351 vector unsigned short vec_mladd (vector unsigned short,
14352 vector unsigned short,
14353 vector unsigned short);
14355 vector signed short vec_mradds (vector signed short,
14356 vector signed short,
14357 vector signed short);
14359 vector unsigned int vec_msum (vector unsigned char,
14360 vector unsigned char,
14361 vector unsigned int);
14362 vector signed int vec_msum (vector signed char,
14363 vector unsigned char,
14364 vector signed int);
14365 vector unsigned int vec_msum (vector unsigned short,
14366 vector unsigned short,
14367 vector unsigned int);
14368 vector signed int vec_msum (vector signed short,
14369 vector signed short,
14370 vector signed int);
14372 vector signed int vec_vmsumshm (vector signed short,
14373 vector signed short,
14374 vector signed int);
14376 vector unsigned int vec_vmsumuhm (vector unsigned short,
14377 vector unsigned short,
14378 vector unsigned int);
14380 vector signed int vec_vmsummbm (vector signed char,
14381 vector unsigned char,
14382 vector signed int);
14384 vector unsigned int vec_vmsumubm (vector unsigned char,
14385 vector unsigned char,
14386 vector unsigned int);
14388 vector unsigned int vec_msums (vector unsigned short,
14389 vector unsigned short,
14390 vector unsigned int);
14391 vector signed int vec_msums (vector signed short,
14392 vector signed short,
14393 vector signed int);
14395 vector signed int vec_vmsumshs (vector signed short,
14396 vector signed short,
14397 vector signed int);
14399 vector unsigned int vec_vmsumuhs (vector unsigned short,
14400 vector unsigned short,
14401 vector unsigned int);
14403 void vec_mtvscr (vector signed int);
14404 void vec_mtvscr (vector unsigned int);
14405 void vec_mtvscr (vector bool int);
14406 void vec_mtvscr (vector signed short);
14407 void vec_mtvscr (vector unsigned short);
14408 void vec_mtvscr (vector bool short);
14409 void vec_mtvscr (vector pixel);
14410 void vec_mtvscr (vector signed char);
14411 void vec_mtvscr (vector unsigned char);
14412 void vec_mtvscr (vector bool char);
14414 vector unsigned short vec_mule (vector unsigned char,
14415 vector unsigned char);
14416 vector signed short vec_mule (vector signed char,
14417 vector signed char);
14418 vector unsigned int vec_mule (vector unsigned short,
14419 vector unsigned short);
14420 vector signed int vec_mule (vector signed short, vector signed short);
14422 vector signed int vec_vmulesh (vector signed short,
14423 vector signed short);
14425 vector unsigned int vec_vmuleuh (vector unsigned short,
14426 vector unsigned short);
14428 vector signed short vec_vmulesb (vector signed char,
14429 vector signed char);
14431 vector unsigned short vec_vmuleub (vector unsigned char,
14432 vector unsigned char);
14434 vector unsigned short vec_mulo (vector unsigned char,
14435 vector unsigned char);
14436 vector signed short vec_mulo (vector signed char, vector signed char);
14437 vector unsigned int vec_mulo (vector unsigned short,
14438 vector unsigned short);
14439 vector signed int vec_mulo (vector signed short, vector signed short);
14441 vector signed int vec_vmulosh (vector signed short,
14442 vector signed short);
14444 vector unsigned int vec_vmulouh (vector unsigned short,
14445 vector unsigned short);
14447 vector signed short vec_vmulosb (vector signed char,
14448 vector signed char);
14450 vector unsigned short vec_vmuloub (vector unsigned char,
14451 vector unsigned char);
14453 vector float vec_nmsub (vector float, vector float, vector float);
14455 vector float vec_nor (vector float, vector float);
14456 vector signed int vec_nor (vector signed int, vector signed int);
14457 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14458 vector bool int vec_nor (vector bool int, vector bool int);
14459 vector signed short vec_nor (vector signed short, vector signed short);
14460 vector unsigned short vec_nor (vector unsigned short,
14461 vector unsigned short);
14462 vector bool short vec_nor (vector bool short, vector bool short);
14463 vector signed char vec_nor (vector signed char, vector signed char);
14464 vector unsigned char vec_nor (vector unsigned char,
14465 vector unsigned char);
14466 vector bool char vec_nor (vector bool char, vector bool char);
14468 vector float vec_or (vector float, vector float);
14469 vector float vec_or (vector float, vector bool int);
14470 vector float vec_or (vector bool int, vector float);
14471 vector bool int vec_or (vector bool int, vector bool int);
14472 vector signed int vec_or (vector bool int, vector signed int);
14473 vector signed int vec_or (vector signed int, vector bool int);
14474 vector signed int vec_or (vector signed int, vector signed int);
14475 vector unsigned int vec_or (vector bool int, vector unsigned int);
14476 vector unsigned int vec_or (vector unsigned int, vector bool int);
14477 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14478 vector bool short vec_or (vector bool short, vector bool short);
14479 vector signed short vec_or (vector bool short, vector signed short);
14480 vector signed short vec_or (vector signed short, vector bool short);
14481 vector signed short vec_or (vector signed short, vector signed short);
14482 vector unsigned short vec_or (vector bool short, vector unsigned short);
14483 vector unsigned short vec_or (vector unsigned short, vector bool short);
14484 vector unsigned short vec_or (vector unsigned short,
14485 vector unsigned short);
14486 vector signed char vec_or (vector bool char, vector signed char);
14487 vector bool char vec_or (vector bool char, vector bool char);
14488 vector signed char vec_or (vector signed char, vector bool char);
14489 vector signed char vec_or (vector signed char, vector signed char);
14490 vector unsigned char vec_or (vector bool char, vector unsigned char);
14491 vector unsigned char vec_or (vector unsigned char, vector bool char);
14492 vector unsigned char vec_or (vector unsigned char,
14493 vector unsigned char);
14495 vector signed char vec_pack (vector signed short, vector signed short);
14496 vector unsigned char vec_pack (vector unsigned short,
14497 vector unsigned short);
14498 vector bool char vec_pack (vector bool short, vector bool short);
14499 vector signed short vec_pack (vector signed int, vector signed int);
14500 vector unsigned short vec_pack (vector unsigned int,
14501 vector unsigned int);
14502 vector bool short vec_pack (vector bool int, vector bool int);
14504 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14505 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14506 vector unsigned short vec_vpkuwum (vector unsigned int,
14507 vector unsigned int);
14509 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14510 vector signed char vec_vpkuhum (vector signed short,
14511 vector signed short);
14512 vector unsigned char vec_vpkuhum (vector unsigned short,
14513 vector unsigned short);
14515 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14517 vector unsigned char vec_packs (vector unsigned short,
14518 vector unsigned short);
14519 vector signed char vec_packs (vector signed short, vector signed short);
14520 vector unsigned short vec_packs (vector unsigned int,
14521 vector unsigned int);
14522 vector signed short vec_packs (vector signed int, vector signed int);
14524 vector signed short vec_vpkswss (vector signed int, vector signed int);
14526 vector unsigned short vec_vpkuwus (vector unsigned int,
14527 vector unsigned int);
14529 vector signed char vec_vpkshss (vector signed short,
14530 vector signed short);
14532 vector unsigned char vec_vpkuhus (vector unsigned short,
14533 vector unsigned short);
14535 vector unsigned char vec_packsu (vector unsigned short,
14536 vector unsigned short);
14537 vector unsigned char vec_packsu (vector signed short,
14538 vector signed short);
14539 vector unsigned short vec_packsu (vector unsigned int,
14540 vector unsigned int);
14541 vector unsigned short vec_packsu (vector signed int, vector signed int);
14543 vector unsigned short vec_vpkswus (vector signed int,
14544 vector signed int);
14546 vector unsigned char vec_vpkshus (vector signed short,
14547 vector signed short);
14549 vector float vec_perm (vector float,
14551 vector unsigned char);
14552 vector signed int vec_perm (vector signed int,
14554 vector unsigned char);
14555 vector unsigned int vec_perm (vector unsigned int,
14556 vector unsigned int,
14557 vector unsigned char);
14558 vector bool int vec_perm (vector bool int,
14560 vector unsigned char);
14561 vector signed short vec_perm (vector signed short,
14562 vector signed short,
14563 vector unsigned char);
14564 vector unsigned short vec_perm (vector unsigned short,
14565 vector unsigned short,
14566 vector unsigned char);
14567 vector bool short vec_perm (vector bool short,
14569 vector unsigned char);
14570 vector pixel vec_perm (vector pixel,
14572 vector unsigned char);
14573 vector signed char vec_perm (vector signed char,
14574 vector signed char,
14575 vector unsigned char);
14576 vector unsigned char vec_perm (vector unsigned char,
14577 vector unsigned char,
14578 vector unsigned char);
14579 vector bool char vec_perm (vector bool char,
14581 vector unsigned char);
14583 vector float vec_re (vector float);
14585 vector signed char vec_rl (vector signed char,
14586 vector unsigned char);
14587 vector unsigned char vec_rl (vector unsigned char,
14588 vector unsigned char);
14589 vector signed short vec_rl (vector signed short, vector unsigned short);
14590 vector unsigned short vec_rl (vector unsigned short,
14591 vector unsigned short);
14592 vector signed int vec_rl (vector signed int, vector unsigned int);
14593 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14595 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14596 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14598 vector signed short vec_vrlh (vector signed short,
14599 vector unsigned short);
14600 vector unsigned short vec_vrlh (vector unsigned short,
14601 vector unsigned short);
14603 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14604 vector unsigned char vec_vrlb (vector unsigned char,
14605 vector unsigned char);
14607 vector float vec_round (vector float);
14609 vector float vec_recip (vector float, vector float);
14611 vector float vec_rsqrt (vector float);
14613 vector float vec_rsqrte (vector float);
14615 vector float vec_sel (vector float, vector float, vector bool int);
14616 vector float vec_sel (vector float, vector float, vector unsigned int);
14617 vector signed int vec_sel (vector signed int,
14620 vector signed int vec_sel (vector signed int,
14622 vector unsigned int);
14623 vector unsigned int vec_sel (vector unsigned int,
14624 vector unsigned int,
14626 vector unsigned int vec_sel (vector unsigned int,
14627 vector unsigned int,
14628 vector unsigned int);
14629 vector bool int vec_sel (vector bool int,
14632 vector bool int vec_sel (vector bool int,
14634 vector unsigned int);
14635 vector signed short vec_sel (vector signed short,
14636 vector signed short,
14637 vector bool short);
14638 vector signed short vec_sel (vector signed short,
14639 vector signed short,
14640 vector unsigned short);
14641 vector unsigned short vec_sel (vector unsigned short,
14642 vector unsigned short,
14643 vector bool short);
14644 vector unsigned short vec_sel (vector unsigned short,
14645 vector unsigned short,
14646 vector unsigned short);
14647 vector bool short vec_sel (vector bool short,
14649 vector bool short);
14650 vector bool short vec_sel (vector bool short,
14652 vector unsigned short);
14653 vector signed char vec_sel (vector signed char,
14654 vector signed char,
14656 vector signed char vec_sel (vector signed char,
14657 vector signed char,
14658 vector unsigned char);
14659 vector unsigned char vec_sel (vector unsigned char,
14660 vector unsigned char,
14662 vector unsigned char vec_sel (vector unsigned char,
14663 vector unsigned char,
14664 vector unsigned char);
14665 vector bool char vec_sel (vector bool char,
14668 vector bool char vec_sel (vector bool char,
14670 vector unsigned char);
14672 vector signed char vec_sl (vector signed char,
14673 vector unsigned char);
14674 vector unsigned char vec_sl (vector unsigned char,
14675 vector unsigned char);
14676 vector signed short vec_sl (vector signed short, vector unsigned short);
14677 vector unsigned short vec_sl (vector unsigned short,
14678 vector unsigned short);
14679 vector signed int vec_sl (vector signed int, vector unsigned int);
14680 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14682 vector signed int vec_vslw (vector signed int, vector unsigned int);
14683 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14685 vector signed short vec_vslh (vector signed short,
14686 vector unsigned short);
14687 vector unsigned short vec_vslh (vector unsigned short,
14688 vector unsigned short);
14690 vector signed char vec_vslb (vector signed char, vector unsigned char);
14691 vector unsigned char vec_vslb (vector unsigned char,
14692 vector unsigned char);
14694 vector float vec_sld (vector float, vector float, const int);
14695 vector signed int vec_sld (vector signed int,
14698 vector unsigned int vec_sld (vector unsigned int,
14699 vector unsigned int,
14701 vector bool int vec_sld (vector bool int,
14704 vector signed short vec_sld (vector signed short,
14705 vector signed short,
14707 vector unsigned short vec_sld (vector unsigned short,
14708 vector unsigned short,
14710 vector bool short vec_sld (vector bool short,
14713 vector pixel vec_sld (vector pixel,
14716 vector signed char vec_sld (vector signed char,
14717 vector signed char,
14719 vector unsigned char vec_sld (vector unsigned char,
14720 vector unsigned char,
14722 vector bool char vec_sld (vector bool char,
14726 vector signed int vec_sll (vector signed int,
14727 vector unsigned int);
14728 vector signed int vec_sll (vector signed int,
14729 vector unsigned short);
14730 vector signed int vec_sll (vector signed int,
14731 vector unsigned char);
14732 vector unsigned int vec_sll (vector unsigned int,
14733 vector unsigned int);
14734 vector unsigned int vec_sll (vector unsigned int,
14735 vector unsigned short);
14736 vector unsigned int vec_sll (vector unsigned int,
14737 vector unsigned char);
14738 vector bool int vec_sll (vector bool int,
14739 vector unsigned int);
14740 vector bool int vec_sll (vector bool int,
14741 vector unsigned short);
14742 vector bool int vec_sll (vector bool int,
14743 vector unsigned char);
14744 vector signed short vec_sll (vector signed short,
14745 vector unsigned int);
14746 vector signed short vec_sll (vector signed short,
14747 vector unsigned short);
14748 vector signed short vec_sll (vector signed short,
14749 vector unsigned char);
14750 vector unsigned short vec_sll (vector unsigned short,
14751 vector unsigned int);
14752 vector unsigned short vec_sll (vector unsigned short,
14753 vector unsigned short);
14754 vector unsigned short vec_sll (vector unsigned short,
14755 vector unsigned char);
14756 vector bool short vec_sll (vector bool short, vector unsigned int);
14757 vector bool short vec_sll (vector bool short, vector unsigned short);
14758 vector bool short vec_sll (vector bool short, vector unsigned char);
14759 vector pixel vec_sll (vector pixel, vector unsigned int);
14760 vector pixel vec_sll (vector pixel, vector unsigned short);
14761 vector pixel vec_sll (vector pixel, vector unsigned char);
14762 vector signed char vec_sll (vector signed char, vector unsigned int);
14763 vector signed char vec_sll (vector signed char, vector unsigned short);
14764 vector signed char vec_sll (vector signed char, vector unsigned char);
14765 vector unsigned char vec_sll (vector unsigned char,
14766 vector unsigned int);
14767 vector unsigned char vec_sll (vector unsigned char,
14768 vector unsigned short);
14769 vector unsigned char vec_sll (vector unsigned char,
14770 vector unsigned char);
14771 vector bool char vec_sll (vector bool char, vector unsigned int);
14772 vector bool char vec_sll (vector bool char, vector unsigned short);
14773 vector bool char vec_sll (vector bool char, vector unsigned char);
14775 vector float vec_slo (vector float, vector signed char);
14776 vector float vec_slo (vector float, vector unsigned char);
14777 vector signed int vec_slo (vector signed int, vector signed char);
14778 vector signed int vec_slo (vector signed int, vector unsigned char);
14779 vector unsigned int vec_slo (vector unsigned int, vector signed char);
14780 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
14781 vector signed short vec_slo (vector signed short, vector signed char);
14782 vector signed short vec_slo (vector signed short, vector unsigned char);
14783 vector unsigned short vec_slo (vector unsigned short,
14784 vector signed char);
14785 vector unsigned short vec_slo (vector unsigned short,
14786 vector unsigned char);
14787 vector pixel vec_slo (vector pixel, vector signed char);
14788 vector pixel vec_slo (vector pixel, vector unsigned char);
14789 vector signed char vec_slo (vector signed char, vector signed char);
14790 vector signed char vec_slo (vector signed char, vector unsigned char);
14791 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14792 vector unsigned char vec_slo (vector unsigned char,
14793 vector unsigned char);
14795 vector signed char vec_splat (vector signed char, const int);
14796 vector unsigned char vec_splat (vector unsigned char, const int);
14797 vector bool char vec_splat (vector bool char, const int);
14798 vector signed short vec_splat (vector signed short, const int);
14799 vector unsigned short vec_splat (vector unsigned short, const int);
14800 vector bool short vec_splat (vector bool short, const int);
14801 vector pixel vec_splat (vector pixel, const int);
14802 vector float vec_splat (vector float, const int);
14803 vector signed int vec_splat (vector signed int, const int);
14804 vector unsigned int vec_splat (vector unsigned int, const int);
14805 vector bool int vec_splat (vector bool int, const int);
14806 vector signed long vec_splat (vector signed long, const int);
14807 vector unsigned long vec_splat (vector unsigned long, const int);
14809 vector signed char vec_splats (signed char);
14810 vector unsigned char vec_splats (unsigned char);
14811 vector signed short vec_splats (signed short);
14812 vector unsigned short vec_splats (unsigned short);
14813 vector signed int vec_splats (signed int);
14814 vector unsigned int vec_splats (unsigned int);
14815 vector float vec_splats (float);
14817 vector float vec_vspltw (vector float, const int);
14818 vector signed int vec_vspltw (vector signed int, const int);
14819 vector unsigned int vec_vspltw (vector unsigned int, const int);
14820 vector bool int vec_vspltw (vector bool int, const int);
14822 vector bool short vec_vsplth (vector bool short, const int);
14823 vector signed short vec_vsplth (vector signed short, const int);
14824 vector unsigned short vec_vsplth (vector unsigned short, const int);
14825 vector pixel vec_vsplth (vector pixel, const int);
14827 vector signed char vec_vspltb (vector signed char, const int);
14828 vector unsigned char vec_vspltb (vector unsigned char, const int);
14829 vector bool char vec_vspltb (vector bool char, const int);
14831 vector signed char vec_splat_s8 (const int);
14833 vector signed short vec_splat_s16 (const int);
14835 vector signed int vec_splat_s32 (const int);
14837 vector unsigned char vec_splat_u8 (const int);
14839 vector unsigned short vec_splat_u16 (const int);
14841 vector unsigned int vec_splat_u32 (const int);
14843 vector signed char vec_sr (vector signed char, vector unsigned char);
14844 vector unsigned char vec_sr (vector unsigned char,
14845 vector unsigned char);
14846 vector signed short vec_sr (vector signed short,
14847 vector unsigned short);
14848 vector unsigned short vec_sr (vector unsigned short,
14849 vector unsigned short);
14850 vector signed int vec_sr (vector signed int, vector unsigned int);
14851 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
14853 vector signed int vec_vsrw (vector signed int, vector unsigned int);
14854 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
14856 vector signed short vec_vsrh (vector signed short,
14857 vector unsigned short);
14858 vector unsigned short vec_vsrh (vector unsigned short,
14859 vector unsigned short);
14861 vector signed char vec_vsrb (vector signed char, vector unsigned char);
14862 vector unsigned char vec_vsrb (vector unsigned char,
14863 vector unsigned char);
14865 vector signed char vec_sra (vector signed char, vector unsigned char);
14866 vector unsigned char vec_sra (vector unsigned char,
14867 vector unsigned char);
14868 vector signed short vec_sra (vector signed short,
14869 vector unsigned short);
14870 vector unsigned short vec_sra (vector unsigned short,
14871 vector unsigned short);
14872 vector signed int vec_sra (vector signed int, vector unsigned int);
14873 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
14875 vector signed int vec_vsraw (vector signed int, vector unsigned int);
14876 vector unsigned int vec_vsraw (vector unsigned int,
14877 vector unsigned int);
14879 vector signed short vec_vsrah (vector signed short,
14880 vector unsigned short);
14881 vector unsigned short vec_vsrah (vector unsigned short,
14882 vector unsigned short);
14884 vector signed char vec_vsrab (vector signed char, vector unsigned char);
14885 vector unsigned char vec_vsrab (vector unsigned char,
14886 vector unsigned char);
14888 vector signed int vec_srl (vector signed int, vector unsigned int);
14889 vector signed int vec_srl (vector signed int, vector unsigned short);
14890 vector signed int vec_srl (vector signed int, vector unsigned char);
14891 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
14892 vector unsigned int vec_srl (vector unsigned int,
14893 vector unsigned short);
14894 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
14895 vector bool int vec_srl (vector bool int, vector unsigned int);
14896 vector bool int vec_srl (vector bool int, vector unsigned short);
14897 vector bool int vec_srl (vector bool int, vector unsigned char);
14898 vector signed short vec_srl (vector signed short, vector unsigned int);
14899 vector signed short vec_srl (vector signed short,
14900 vector unsigned short);
14901 vector signed short vec_srl (vector signed short, vector unsigned char);
14902 vector unsigned short vec_srl (vector unsigned short,
14903 vector unsigned int);
14904 vector unsigned short vec_srl (vector unsigned short,
14905 vector unsigned short);
14906 vector unsigned short vec_srl (vector unsigned short,
14907 vector unsigned char);
14908 vector bool short vec_srl (vector bool short, vector unsigned int);
14909 vector bool short vec_srl (vector bool short, vector unsigned short);
14910 vector bool short vec_srl (vector bool short, vector unsigned char);
14911 vector pixel vec_srl (vector pixel, vector unsigned int);
14912 vector pixel vec_srl (vector pixel, vector unsigned short);
14913 vector pixel vec_srl (vector pixel, vector unsigned char);
14914 vector signed char vec_srl (vector signed char, vector unsigned int);
14915 vector signed char vec_srl (vector signed char, vector unsigned short);
14916 vector signed char vec_srl (vector signed char, vector unsigned char);
14917 vector unsigned char vec_srl (vector unsigned char,
14918 vector unsigned int);
14919 vector unsigned char vec_srl (vector unsigned char,
14920 vector unsigned short);
14921 vector unsigned char vec_srl (vector unsigned char,
14922 vector unsigned char);
14923 vector bool char vec_srl (vector bool char, vector unsigned int);
14924 vector bool char vec_srl (vector bool char, vector unsigned short);
14925 vector bool char vec_srl (vector bool char, vector unsigned char);
14927 vector float vec_sro (vector float, vector signed char);
14928 vector float vec_sro (vector float, vector unsigned char);
14929 vector signed int vec_sro (vector signed int, vector signed char);
14930 vector signed int vec_sro (vector signed int, vector unsigned char);
14931 vector unsigned int vec_sro (vector unsigned int, vector signed char);
14932 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
14933 vector signed short vec_sro (vector signed short, vector signed char);
14934 vector signed short vec_sro (vector signed short, vector unsigned char);
14935 vector unsigned short vec_sro (vector unsigned short,
14936 vector signed char);
14937 vector unsigned short vec_sro (vector unsigned short,
14938 vector unsigned char);
14939 vector pixel vec_sro (vector pixel, vector signed char);
14940 vector pixel vec_sro (vector pixel, vector unsigned char);
14941 vector signed char vec_sro (vector signed char, vector signed char);
14942 vector signed char vec_sro (vector signed char, vector unsigned char);
14943 vector unsigned char vec_sro (vector unsigned char, vector signed char);
14944 vector unsigned char vec_sro (vector unsigned char,
14945 vector unsigned char);
14947 void vec_st (vector float, int, vector float *);
14948 void vec_st (vector float, int, float *);
14949 void vec_st (vector signed int, int, vector signed int *);
14950 void vec_st (vector signed int, int, int *);
14951 void vec_st (vector unsigned int, int, vector unsigned int *);
14952 void vec_st (vector unsigned int, int, unsigned int *);
14953 void vec_st (vector bool int, int, vector bool int *);
14954 void vec_st (vector bool int, int, unsigned int *);
14955 void vec_st (vector bool int, int, int *);
14956 void vec_st (vector signed short, int, vector signed short *);
14957 void vec_st (vector signed short, int, short *);
14958 void vec_st (vector unsigned short, int, vector unsigned short *);
14959 void vec_st (vector unsigned short, int, unsigned short *);
14960 void vec_st (vector bool short, int, vector bool short *);
14961 void vec_st (vector bool short, int, unsigned short *);
14962 void vec_st (vector pixel, int, vector pixel *);
14963 void vec_st (vector pixel, int, unsigned short *);
14964 void vec_st (vector pixel, int, short *);
14965 void vec_st (vector bool short, int, short *);
14966 void vec_st (vector signed char, int, vector signed char *);
14967 void vec_st (vector signed char, int, signed char *);
14968 void vec_st (vector unsigned char, int, vector unsigned char *);
14969 void vec_st (vector unsigned char, int, unsigned char *);
14970 void vec_st (vector bool char, int, vector bool char *);
14971 void vec_st (vector bool char, int, unsigned char *);
14972 void vec_st (vector bool char, int, signed char *);
14974 void vec_ste (vector signed char, int, signed char *);
14975 void vec_ste (vector unsigned char, int, unsigned char *);
14976 void vec_ste (vector bool char, int, signed char *);
14977 void vec_ste (vector bool char, int, unsigned char *);
14978 void vec_ste (vector signed short, int, short *);
14979 void vec_ste (vector unsigned short, int, unsigned short *);
14980 void vec_ste (vector bool short, int, short *);
14981 void vec_ste (vector bool short, int, unsigned short *);
14982 void vec_ste (vector pixel, int, short *);
14983 void vec_ste (vector pixel, int, unsigned short *);
14984 void vec_ste (vector float, int, float *);
14985 void vec_ste (vector signed int, int, int *);
14986 void vec_ste (vector unsigned int, int, unsigned int *);
14987 void vec_ste (vector bool int, int, int *);
14988 void vec_ste (vector bool int, int, unsigned int *);
14990 void vec_stvewx (vector float, int, float *);
14991 void vec_stvewx (vector signed int, int, int *);
14992 void vec_stvewx (vector unsigned int, int, unsigned int *);
14993 void vec_stvewx (vector bool int, int, int *);
14994 void vec_stvewx (vector bool int, int, unsigned int *);
14996 void vec_stvehx (vector signed short, int, short *);
14997 void vec_stvehx (vector unsigned short, int, unsigned short *);
14998 void vec_stvehx (vector bool short, int, short *);
14999 void vec_stvehx (vector bool short, int, unsigned short *);
15000 void vec_stvehx (vector pixel, int, short *);
15001 void vec_stvehx (vector pixel, int, unsigned short *);
15003 void vec_stvebx (vector signed char, int, signed char *);
15004 void vec_stvebx (vector unsigned char, int, unsigned char *);
15005 void vec_stvebx (vector bool char, int, signed char *);
15006 void vec_stvebx (vector bool char, int, unsigned char *);
15008 void vec_stl (vector float, int, vector float *);
15009 void vec_stl (vector float, int, float *);
15010 void vec_stl (vector signed int, int, vector signed int *);
15011 void vec_stl (vector signed int, int, int *);
15012 void vec_stl (vector unsigned int, int, vector unsigned int *);
15013 void vec_stl (vector unsigned int, int, unsigned int *);
15014 void vec_stl (vector bool int, int, vector bool int *);
15015 void vec_stl (vector bool int, int, unsigned int *);
15016 void vec_stl (vector bool int, int, int *);
15017 void vec_stl (vector signed short, int, vector signed short *);
15018 void vec_stl (vector signed short, int, short *);
15019 void vec_stl (vector unsigned short, int, vector unsigned short *);
15020 void vec_stl (vector unsigned short, int, unsigned short *);
15021 void vec_stl (vector bool short, int, vector bool short *);
15022 void vec_stl (vector bool short, int, unsigned short *);
15023 void vec_stl (vector bool short, int, short *);
15024 void vec_stl (vector pixel, int, vector pixel *);
15025 void vec_stl (vector pixel, int, unsigned short *);
15026 void vec_stl (vector pixel, int, short *);
15027 void vec_stl (vector signed char, int, vector signed char *);
15028 void vec_stl (vector signed char, int, signed char *);
15029 void vec_stl (vector unsigned char, int, vector unsigned char *);
15030 void vec_stl (vector unsigned char, int, unsigned char *);
15031 void vec_stl (vector bool char, int, vector bool char *);
15032 void vec_stl (vector bool char, int, unsigned char *);
15033 void vec_stl (vector bool char, int, signed char *);
15035 vector signed char vec_sub (vector bool char, vector signed char);
15036 vector signed char vec_sub (vector signed char, vector bool char);
15037 vector signed char vec_sub (vector signed char, vector signed char);
15038 vector unsigned char vec_sub (vector bool char, vector unsigned char);
15039 vector unsigned char vec_sub (vector unsigned char, vector bool char);
15040 vector unsigned char vec_sub (vector unsigned char,
15041 vector unsigned char);
15042 vector signed short vec_sub (vector bool short, vector signed short);
15043 vector signed short vec_sub (vector signed short, vector bool short);
15044 vector signed short vec_sub (vector signed short, vector signed short);
15045 vector unsigned short vec_sub (vector bool short,
15046 vector unsigned short);
15047 vector unsigned short vec_sub (vector unsigned short,
15048 vector bool short);
15049 vector unsigned short vec_sub (vector unsigned short,
15050 vector unsigned short);
15051 vector signed int vec_sub (vector bool int, vector signed int);
15052 vector signed int vec_sub (vector signed int, vector bool int);
15053 vector signed int vec_sub (vector signed int, vector signed int);
15054 vector unsigned int vec_sub (vector bool int, vector unsigned int);
15055 vector unsigned int vec_sub (vector unsigned int, vector bool int);
15056 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
15057 vector float vec_sub (vector float, vector float);
15059 vector float vec_vsubfp (vector float, vector float);
15061 vector signed int vec_vsubuwm (vector bool int, vector signed int);
15062 vector signed int vec_vsubuwm (vector signed int, vector bool int);
15063 vector signed int vec_vsubuwm (vector signed int, vector signed int);
15064 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
15065 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
15066 vector unsigned int vec_vsubuwm (vector unsigned int,
15067 vector unsigned int);
15069 vector signed short vec_vsubuhm (vector bool short,
15070 vector signed short);
15071 vector signed short vec_vsubuhm (vector signed short,
15072 vector bool short);
15073 vector signed short vec_vsubuhm (vector signed short,
15074 vector signed short);
15075 vector unsigned short vec_vsubuhm (vector bool short,
15076 vector unsigned short);
15077 vector unsigned short vec_vsubuhm (vector unsigned short,
15078 vector bool short);
15079 vector unsigned short vec_vsubuhm (vector unsigned short,
15080 vector unsigned short);
15082 vector signed char vec_vsububm (vector bool char, vector signed char);
15083 vector signed char vec_vsububm (vector signed char, vector bool char);
15084 vector signed char vec_vsububm (vector signed char, vector signed char);
15085 vector unsigned char vec_vsububm (vector bool char,
15086 vector unsigned char);
15087 vector unsigned char vec_vsububm (vector unsigned char,
15089 vector unsigned char vec_vsububm (vector unsigned char,
15090 vector unsigned char);
15092 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
15094 vector unsigned char vec_subs (vector bool char, vector unsigned char);
15095 vector unsigned char vec_subs (vector unsigned char, vector bool char);
15096 vector unsigned char vec_subs (vector unsigned char,
15097 vector unsigned char);
15098 vector signed char vec_subs (vector bool char, vector signed char);
15099 vector signed char vec_subs (vector signed char, vector bool char);
15100 vector signed char vec_subs (vector signed char, vector signed char);
15101 vector unsigned short vec_subs (vector bool short,
15102 vector unsigned short);
15103 vector unsigned short vec_subs (vector unsigned short,
15104 vector bool short);
15105 vector unsigned short vec_subs (vector unsigned short,
15106 vector unsigned short);
15107 vector signed short vec_subs (vector bool short, vector signed short);
15108 vector signed short vec_subs (vector signed short, vector bool short);
15109 vector signed short vec_subs (vector signed short, vector signed short);
15110 vector unsigned int vec_subs (vector bool int, vector unsigned int);
15111 vector unsigned int vec_subs (vector unsigned int, vector bool int);
15112 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
15113 vector signed int vec_subs (vector bool int, vector signed int);
15114 vector signed int vec_subs (vector signed int, vector bool int);
15115 vector signed int vec_subs (vector signed int, vector signed int);
15117 vector signed int vec_vsubsws (vector bool int, vector signed int);
15118 vector signed int vec_vsubsws (vector signed int, vector bool int);
15119 vector signed int vec_vsubsws (vector signed int, vector signed int);
15121 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
15122 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
15123 vector unsigned int vec_vsubuws (vector unsigned int,
15124 vector unsigned int);
15126 vector signed short vec_vsubshs (vector bool short,
15127 vector signed short);
15128 vector signed short vec_vsubshs (vector signed short,
15129 vector bool short);
15130 vector signed short vec_vsubshs (vector signed short,
15131 vector signed short);
15133 vector unsigned short vec_vsubuhs (vector bool short,
15134 vector unsigned short);
15135 vector unsigned short vec_vsubuhs (vector unsigned short,
15136 vector bool short);
15137 vector unsigned short vec_vsubuhs (vector unsigned short,
15138 vector unsigned short);
15140 vector signed char vec_vsubsbs (vector bool char, vector signed char);
15141 vector signed char vec_vsubsbs (vector signed char, vector bool char);
15142 vector signed char vec_vsubsbs (vector signed char, vector signed char);
15144 vector unsigned char vec_vsububs (vector bool char,
15145 vector unsigned char);
15146 vector unsigned char vec_vsububs (vector unsigned char,
15148 vector unsigned char vec_vsububs (vector unsigned char,
15149 vector unsigned char);
15151 vector unsigned int vec_sum4s (vector unsigned char,
15152 vector unsigned int);
15153 vector signed int vec_sum4s (vector signed char, vector signed int);
15154 vector signed int vec_sum4s (vector signed short, vector signed int);
15156 vector signed int vec_vsum4shs (vector signed short, vector signed int);
15158 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
15160 vector unsigned int vec_vsum4ubs (vector unsigned char,
15161 vector unsigned int);
15163 vector signed int vec_sum2s (vector signed int, vector signed int);
15165 vector signed int vec_sums (vector signed int, vector signed int);
15167 vector float vec_trunc (vector float);
15169 vector signed short vec_unpackh (vector signed char);
15170 vector bool short vec_unpackh (vector bool char);
15171 vector signed int vec_unpackh (vector signed short);
15172 vector bool int vec_unpackh (vector bool short);
15173 vector unsigned int vec_unpackh (vector pixel);
15175 vector bool int vec_vupkhsh (vector bool short);
15176 vector signed int vec_vupkhsh (vector signed short);
15178 vector unsigned int vec_vupkhpx (vector pixel);
15180 vector bool short vec_vupkhsb (vector bool char);
15181 vector signed short vec_vupkhsb (vector signed char);
15183 vector signed short vec_unpackl (vector signed char);
15184 vector bool short vec_unpackl (vector bool char);
15185 vector unsigned int vec_unpackl (vector pixel);
15186 vector signed int vec_unpackl (vector signed short);
15187 vector bool int vec_unpackl (vector bool short);
15189 vector unsigned int vec_vupklpx (vector pixel);
15191 vector bool int vec_vupklsh (vector bool short);
15192 vector signed int vec_vupklsh (vector signed short);
15194 vector bool short vec_vupklsb (vector bool char);
15195 vector signed short vec_vupklsb (vector signed char);
15197 vector float vec_xor (vector float, vector float);
15198 vector float vec_xor (vector float, vector bool int);
15199 vector float vec_xor (vector bool int, vector float);
15200 vector bool int vec_xor (vector bool int, vector bool int);
15201 vector signed int vec_xor (vector bool int, vector signed int);
15202 vector signed int vec_xor (vector signed int, vector bool int);
15203 vector signed int vec_xor (vector signed int, vector signed int);
15204 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15205 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15206 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15207 vector bool short vec_xor (vector bool short, vector bool short);
15208 vector signed short vec_xor (vector bool short, vector signed short);
15209 vector signed short vec_xor (vector signed short, vector bool short);
15210 vector signed short vec_xor (vector signed short, vector signed short);
15211 vector unsigned short vec_xor (vector bool short,
15212 vector unsigned short);
15213 vector unsigned short vec_xor (vector unsigned short,
15214 vector bool short);
15215 vector unsigned short vec_xor (vector unsigned short,
15216 vector unsigned short);
15217 vector signed char vec_xor (vector bool char, vector signed char);
15218 vector bool char vec_xor (vector bool char, vector bool char);
15219 vector signed char vec_xor (vector signed char, vector bool char);
15220 vector signed char vec_xor (vector signed char, vector signed char);
15221 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15222 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15223 vector unsigned char vec_xor (vector unsigned char,
15224 vector unsigned char);
15226 int vec_all_eq (vector signed char, vector bool char);
15227 int vec_all_eq (vector signed char, vector signed char);
15228 int vec_all_eq (vector unsigned char, vector bool char);
15229 int vec_all_eq (vector unsigned char, vector unsigned char);
15230 int vec_all_eq (vector bool char, vector bool char);
15231 int vec_all_eq (vector bool char, vector unsigned char);
15232 int vec_all_eq (vector bool char, vector signed char);
15233 int vec_all_eq (vector signed short, vector bool short);
15234 int vec_all_eq (vector signed short, vector signed short);
15235 int vec_all_eq (vector unsigned short, vector bool short);
15236 int vec_all_eq (vector unsigned short, vector unsigned short);
15237 int vec_all_eq (vector bool short, vector bool short);
15238 int vec_all_eq (vector bool short, vector unsigned short);
15239 int vec_all_eq (vector bool short, vector signed short);
15240 int vec_all_eq (vector pixel, vector pixel);
15241 int vec_all_eq (vector signed int, vector bool int);
15242 int vec_all_eq (vector signed int, vector signed int);
15243 int vec_all_eq (vector unsigned int, vector bool int);
15244 int vec_all_eq (vector unsigned int, vector unsigned int);
15245 int vec_all_eq (vector bool int, vector bool int);
15246 int vec_all_eq (vector bool int, vector unsigned int);
15247 int vec_all_eq (vector bool int, vector signed int);
15248 int vec_all_eq (vector float, vector float);
15250 int vec_all_ge (vector bool char, vector unsigned char);
15251 int vec_all_ge (vector unsigned char, vector bool char);
15252 int vec_all_ge (vector unsigned char, vector unsigned char);
15253 int vec_all_ge (vector bool char, vector signed char);
15254 int vec_all_ge (vector signed char, vector bool char);
15255 int vec_all_ge (vector signed char, vector signed char);
15256 int vec_all_ge (vector bool short, vector unsigned short);
15257 int vec_all_ge (vector unsigned short, vector bool short);
15258 int vec_all_ge (vector unsigned short, vector unsigned short);
15259 int vec_all_ge (vector signed short, vector signed short);
15260 int vec_all_ge (vector bool short, vector signed short);
15261 int vec_all_ge (vector signed short, vector bool short);
15262 int vec_all_ge (vector bool int, vector unsigned int);
15263 int vec_all_ge (vector unsigned int, vector bool int);
15264 int vec_all_ge (vector unsigned int, vector unsigned int);
15265 int vec_all_ge (vector bool int, vector signed int);
15266 int vec_all_ge (vector signed int, vector bool int);
15267 int vec_all_ge (vector signed int, vector signed int);
15268 int vec_all_ge (vector float, vector float);
15270 int vec_all_gt (vector bool char, vector unsigned char);
15271 int vec_all_gt (vector unsigned char, vector bool char);
15272 int vec_all_gt (vector unsigned char, vector unsigned char);
15273 int vec_all_gt (vector bool char, vector signed char);
15274 int vec_all_gt (vector signed char, vector bool char);
15275 int vec_all_gt (vector signed char, vector signed char);
15276 int vec_all_gt (vector bool short, vector unsigned short);
15277 int vec_all_gt (vector unsigned short, vector bool short);
15278 int vec_all_gt (vector unsigned short, vector unsigned short);
15279 int vec_all_gt (vector bool short, vector signed short);
15280 int vec_all_gt (vector signed short, vector bool short);
15281 int vec_all_gt (vector signed short, vector signed short);
15282 int vec_all_gt (vector bool int, vector unsigned int);
15283 int vec_all_gt (vector unsigned int, vector bool int);
15284 int vec_all_gt (vector unsigned int, vector unsigned int);
15285 int vec_all_gt (vector bool int, vector signed int);
15286 int vec_all_gt (vector signed int, vector bool int);
15287 int vec_all_gt (vector signed int, vector signed int);
15288 int vec_all_gt (vector float, vector float);
15290 int vec_all_in (vector float, vector float);
15292 int vec_all_le (vector bool char, vector unsigned char);
15293 int vec_all_le (vector unsigned char, vector bool char);
15294 int vec_all_le (vector unsigned char, vector unsigned char);
15295 int vec_all_le (vector bool char, vector signed char);
15296 int vec_all_le (vector signed char, vector bool char);
15297 int vec_all_le (vector signed char, vector signed char);
15298 int vec_all_le (vector bool short, vector unsigned short);
15299 int vec_all_le (vector unsigned short, vector bool short);
15300 int vec_all_le (vector unsigned short, vector unsigned short);
15301 int vec_all_le (vector bool short, vector signed short);
15302 int vec_all_le (vector signed short, vector bool short);
15303 int vec_all_le (vector signed short, vector signed short);
15304 int vec_all_le (vector bool int, vector unsigned int);
15305 int vec_all_le (vector unsigned int, vector bool int);
15306 int vec_all_le (vector unsigned int, vector unsigned int);
15307 int vec_all_le (vector bool int, vector signed int);
15308 int vec_all_le (vector signed int, vector bool int);
15309 int vec_all_le (vector signed int, vector signed int);
15310 int vec_all_le (vector float, vector float);
15312 int vec_all_lt (vector bool char, vector unsigned char);
15313 int vec_all_lt (vector unsigned char, vector bool char);
15314 int vec_all_lt (vector unsigned char, vector unsigned char);
15315 int vec_all_lt (vector bool char, vector signed char);
15316 int vec_all_lt (vector signed char, vector bool char);
15317 int vec_all_lt (vector signed char, vector signed char);
15318 int vec_all_lt (vector bool short, vector unsigned short);
15319 int vec_all_lt (vector unsigned short, vector bool short);
15320 int vec_all_lt (vector unsigned short, vector unsigned short);
15321 int vec_all_lt (vector bool short, vector signed short);
15322 int vec_all_lt (vector signed short, vector bool short);
15323 int vec_all_lt (vector signed short, vector signed short);
15324 int vec_all_lt (vector bool int, vector unsigned int);
15325 int vec_all_lt (vector unsigned int, vector bool int);
15326 int vec_all_lt (vector unsigned int, vector unsigned int);
15327 int vec_all_lt (vector bool int, vector signed int);
15328 int vec_all_lt (vector signed int, vector bool int);
15329 int vec_all_lt (vector signed int, vector signed int);
15330 int vec_all_lt (vector float, vector float);
15332 int vec_all_nan (vector float);
15334 int vec_all_ne (vector signed char, vector bool char);
15335 int vec_all_ne (vector signed char, vector signed char);
15336 int vec_all_ne (vector unsigned char, vector bool char);
15337 int vec_all_ne (vector unsigned char, vector unsigned char);
15338 int vec_all_ne (vector bool char, vector bool char);
15339 int vec_all_ne (vector bool char, vector unsigned char);
15340 int vec_all_ne (vector bool char, vector signed char);
15341 int vec_all_ne (vector signed short, vector bool short);
15342 int vec_all_ne (vector signed short, vector signed short);
15343 int vec_all_ne (vector unsigned short, vector bool short);
15344 int vec_all_ne (vector unsigned short, vector unsigned short);
15345 int vec_all_ne (vector bool short, vector bool short);
15346 int vec_all_ne (vector bool short, vector unsigned short);
15347 int vec_all_ne (vector bool short, vector signed short);
15348 int vec_all_ne (vector pixel, vector pixel);
15349 int vec_all_ne (vector signed int, vector bool int);
15350 int vec_all_ne (vector signed int, vector signed int);
15351 int vec_all_ne (vector unsigned int, vector bool int);
15352 int vec_all_ne (vector unsigned int, vector unsigned int);
15353 int vec_all_ne (vector bool int, vector bool int);
15354 int vec_all_ne (vector bool int, vector unsigned int);
15355 int vec_all_ne (vector bool int, vector signed int);
15356 int vec_all_ne (vector float, vector float);
15358 int vec_all_nge (vector float, vector float);
15360 int vec_all_ngt (vector float, vector float);
15362 int vec_all_nle (vector float, vector float);
15364 int vec_all_nlt (vector float, vector float);
15366 int vec_all_numeric (vector float);
15368 int vec_any_eq (vector signed char, vector bool char);
15369 int vec_any_eq (vector signed char, vector signed char);
15370 int vec_any_eq (vector unsigned char, vector bool char);
15371 int vec_any_eq (vector unsigned char, vector unsigned char);
15372 int vec_any_eq (vector bool char, vector bool char);
15373 int vec_any_eq (vector bool char, vector unsigned char);
15374 int vec_any_eq (vector bool char, vector signed char);
15375 int vec_any_eq (vector signed short, vector bool short);
15376 int vec_any_eq (vector signed short, vector signed short);
15377 int vec_any_eq (vector unsigned short, vector bool short);
15378 int vec_any_eq (vector unsigned short, vector unsigned short);
15379 int vec_any_eq (vector bool short, vector bool short);
15380 int vec_any_eq (vector bool short, vector unsigned short);
15381 int vec_any_eq (vector bool short, vector signed short);
15382 int vec_any_eq (vector pixel, vector pixel);
15383 int vec_any_eq (vector signed int, vector bool int);
15384 int vec_any_eq (vector signed int, vector signed int);
15385 int vec_any_eq (vector unsigned int, vector bool int);
15386 int vec_any_eq (vector unsigned int, vector unsigned int);
15387 int vec_any_eq (vector bool int, vector bool int);
15388 int vec_any_eq (vector bool int, vector unsigned int);
15389 int vec_any_eq (vector bool int, vector signed int);
15390 int vec_any_eq (vector float, vector float);
15392 int vec_any_ge (vector signed char, vector bool char);
15393 int vec_any_ge (vector unsigned char, vector bool char);
15394 int vec_any_ge (vector unsigned char, vector unsigned char);
15395 int vec_any_ge (vector signed char, vector signed char);
15396 int vec_any_ge (vector bool char, vector unsigned char);
15397 int vec_any_ge (vector bool char, vector signed char);
15398 int vec_any_ge (vector unsigned short, vector bool short);
15399 int vec_any_ge (vector unsigned short, vector unsigned short);
15400 int vec_any_ge (vector signed short, vector signed short);
15401 int vec_any_ge (vector signed short, vector bool short);
15402 int vec_any_ge (vector bool short, vector unsigned short);
15403 int vec_any_ge (vector bool short, vector signed short);
15404 int vec_any_ge (vector signed int, vector bool int);
15405 int vec_any_ge (vector unsigned int, vector bool int);
15406 int vec_any_ge (vector unsigned int, vector unsigned int);
15407 int vec_any_ge (vector signed int, vector signed int);
15408 int vec_any_ge (vector bool int, vector unsigned int);
15409 int vec_any_ge (vector bool int, vector signed int);
15410 int vec_any_ge (vector float, vector float);
15412 int vec_any_gt (vector bool char, vector unsigned char);
15413 int vec_any_gt (vector unsigned char, vector bool char);
15414 int vec_any_gt (vector unsigned char, vector unsigned char);
15415 int vec_any_gt (vector bool char, vector signed char);
15416 int vec_any_gt (vector signed char, vector bool char);
15417 int vec_any_gt (vector signed char, vector signed char);
15418 int vec_any_gt (vector bool short, vector unsigned short);
15419 int vec_any_gt (vector unsigned short, vector bool short);
15420 int vec_any_gt (vector unsigned short, vector unsigned short);
15421 int vec_any_gt (vector bool short, vector signed short);
15422 int vec_any_gt (vector signed short, vector bool short);
15423 int vec_any_gt (vector signed short, vector signed short);
15424 int vec_any_gt (vector bool int, vector unsigned int);
15425 int vec_any_gt (vector unsigned int, vector bool int);
15426 int vec_any_gt (vector unsigned int, vector unsigned int);
15427 int vec_any_gt (vector bool int, vector signed int);
15428 int vec_any_gt (vector signed int, vector bool int);
15429 int vec_any_gt (vector signed int, vector signed int);
15430 int vec_any_gt (vector float, vector float);
15432 int vec_any_le (vector bool char, vector unsigned char);
15433 int vec_any_le (vector unsigned char, vector bool char);
15434 int vec_any_le (vector unsigned char, vector unsigned char);
15435 int vec_any_le (vector bool char, vector signed char);
15436 int vec_any_le (vector signed char, vector bool char);
15437 int vec_any_le (vector signed char, vector signed char);
15438 int vec_any_le (vector bool short, vector unsigned short);
15439 int vec_any_le (vector unsigned short, vector bool short);
15440 int vec_any_le (vector unsigned short, vector unsigned short);
15441 int vec_any_le (vector bool short, vector signed short);
15442 int vec_any_le (vector signed short, vector bool short);
15443 int vec_any_le (vector signed short, vector signed short);
15444 int vec_any_le (vector bool int, vector unsigned int);
15445 int vec_any_le (vector unsigned int, vector bool int);
15446 int vec_any_le (vector unsigned int, vector unsigned int);
15447 int vec_any_le (vector bool int, vector signed int);
15448 int vec_any_le (vector signed int, vector bool int);
15449 int vec_any_le (vector signed int, vector signed int);
15450 int vec_any_le (vector float, vector float);
15452 int vec_any_lt (vector bool char, vector unsigned char);
15453 int vec_any_lt (vector unsigned char, vector bool char);
15454 int vec_any_lt (vector unsigned char, vector unsigned char);
15455 int vec_any_lt (vector bool char, vector signed char);
15456 int vec_any_lt (vector signed char, vector bool char);
15457 int vec_any_lt (vector signed char, vector signed char);
15458 int vec_any_lt (vector bool short, vector unsigned short);
15459 int vec_any_lt (vector unsigned short, vector bool short);
15460 int vec_any_lt (vector unsigned short, vector unsigned short);
15461 int vec_any_lt (vector bool short, vector signed short);
15462 int vec_any_lt (vector signed short, vector bool short);
15463 int vec_any_lt (vector signed short, vector signed short);
15464 int vec_any_lt (vector bool int, vector unsigned int);
15465 int vec_any_lt (vector unsigned int, vector bool int);
15466 int vec_any_lt (vector unsigned int, vector unsigned int);
15467 int vec_any_lt (vector bool int, vector signed int);
15468 int vec_any_lt (vector signed int, vector bool int);
15469 int vec_any_lt (vector signed int, vector signed int);
15470 int vec_any_lt (vector float, vector float);
15472 int vec_any_nan (vector float);
15474 int vec_any_ne (vector signed char, vector bool char);
15475 int vec_any_ne (vector signed char, vector signed char);
15476 int vec_any_ne (vector unsigned char, vector bool char);
15477 int vec_any_ne (vector unsigned char, vector unsigned char);
15478 int vec_any_ne (vector bool char, vector bool char);
15479 int vec_any_ne (vector bool char, vector unsigned char);
15480 int vec_any_ne (vector bool char, vector signed char);
15481 int vec_any_ne (vector signed short, vector bool short);
15482 int vec_any_ne (vector signed short, vector signed short);
15483 int vec_any_ne (vector unsigned short, vector bool short);
15484 int vec_any_ne (vector unsigned short, vector unsigned short);
15485 int vec_any_ne (vector bool short, vector bool short);
15486 int vec_any_ne (vector bool short, vector unsigned short);
15487 int vec_any_ne (vector bool short, vector signed short);
15488 int vec_any_ne (vector pixel, vector pixel);
15489 int vec_any_ne (vector signed int, vector bool int);
15490 int vec_any_ne (vector signed int, vector signed int);
15491 int vec_any_ne (vector unsigned int, vector bool int);
15492 int vec_any_ne (vector unsigned int, vector unsigned int);
15493 int vec_any_ne (vector bool int, vector bool int);
15494 int vec_any_ne (vector bool int, vector unsigned int);
15495 int vec_any_ne (vector bool int, vector signed int);
15496 int vec_any_ne (vector float, vector float);
15498 int vec_any_nge (vector float, vector float);
15500 int vec_any_ngt (vector float, vector float);
15502 int vec_any_nle (vector float, vector float);
15504 int vec_any_nlt (vector float, vector float);
15506 int vec_any_numeric (vector float);
15508 int vec_any_out (vector float, vector float);
15511 If the vector/scalar (VSX) instruction set is available, the following
15512 additional functions are available:
15515 vector double vec_abs (vector double);
15516 vector double vec_add (vector double, vector double);
15517 vector double vec_and (vector double, vector double);
15518 vector double vec_and (vector double, vector bool long);
15519 vector double vec_and (vector bool long, vector double);
15520 vector long vec_and (vector long, vector long);
15521 vector long vec_and (vector long, vector bool long);
15522 vector long vec_and (vector bool long, vector long);
15523 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15524 vector unsigned long vec_and (vector unsigned long, vector bool long);
15525 vector unsigned long vec_and (vector bool long, vector unsigned long);
15526 vector double vec_andc (vector double, vector double);
15527 vector double vec_andc (vector double, vector bool long);
15528 vector double vec_andc (vector bool long, vector double);
15529 vector long vec_andc (vector long, vector long);
15530 vector long vec_andc (vector long, vector bool long);
15531 vector long vec_andc (vector bool long, vector long);
15532 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15533 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15534 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15535 vector double vec_ceil (vector double);
15536 vector bool long vec_cmpeq (vector double, vector double);
15537 vector bool long vec_cmpge (vector double, vector double);
15538 vector bool long vec_cmpgt (vector double, vector double);
15539 vector bool long vec_cmple (vector double, vector double);
15540 vector bool long vec_cmplt (vector double, vector double);
15541 vector double vec_cpsgn (vector double, vector double);
15542 vector float vec_div (vector float, vector float);
15543 vector double vec_div (vector double, vector double);
15544 vector long vec_div (vector long, vector long);
15545 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15546 vector double vec_floor (vector double);
15547 vector double vec_ld (int, const vector double *);
15548 vector double vec_ld (int, const double *);
15549 vector double vec_ldl (int, const vector double *);
15550 vector double vec_ldl (int, const double *);
15551 vector unsigned char vec_lvsl (int, const volatile double *);
15552 vector unsigned char vec_lvsr (int, const volatile double *);
15553 vector double vec_madd (vector double, vector double, vector double);
15554 vector double vec_max (vector double, vector double);
15555 vector signed long vec_mergeh (vector signed long, vector signed long);
15556 vector signed long vec_mergeh (vector signed long, vector bool long);
15557 vector signed long vec_mergeh (vector bool long, vector signed long);
15558 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15559 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15560 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15561 vector signed long vec_mergel (vector signed long, vector signed long);
15562 vector signed long vec_mergel (vector signed long, vector bool long);
15563 vector signed long vec_mergel (vector bool long, vector signed long);
15564 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15565 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15566 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15567 vector double vec_min (vector double, vector double);
15568 vector float vec_msub (vector float, vector float, vector float);
15569 vector double vec_msub (vector double, vector double, vector double);
15570 vector float vec_mul (vector float, vector float);
15571 vector double vec_mul (vector double, vector double);
15572 vector long vec_mul (vector long, vector long);
15573 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15574 vector float vec_nearbyint (vector float);
15575 vector double vec_nearbyint (vector double);
15576 vector float vec_nmadd (vector float, vector float, vector float);
15577 vector double vec_nmadd (vector double, vector double, vector double);
15578 vector double vec_nmsub (vector double, vector double, vector double);
15579 vector double vec_nor (vector double, vector double);
15580 vector long vec_nor (vector long, vector long);
15581 vector long vec_nor (vector long, vector bool long);
15582 vector long vec_nor (vector bool long, vector long);
15583 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15584 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15585 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15586 vector double vec_or (vector double, vector double);
15587 vector double vec_or (vector double, vector bool long);
15588 vector double vec_or (vector bool long, vector double);
15589 vector long vec_or (vector long, vector long);
15590 vector long vec_or (vector long, vector bool long);
15591 vector long vec_or (vector bool long, vector long);
15592 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15593 vector unsigned long vec_or (vector unsigned long, vector bool long);
15594 vector unsigned long vec_or (vector bool long, vector unsigned long);
15595 vector double vec_perm (vector double, vector double, vector unsigned char);
15596 vector long vec_perm (vector long, vector long, vector unsigned char);
15597 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15598 vector unsigned char);
15599 vector double vec_rint (vector double);
15600 vector double vec_recip (vector double, vector double);
15601 vector double vec_rsqrt (vector double);
15602 vector double vec_rsqrte (vector double);
15603 vector double vec_sel (vector double, vector double, vector bool long);
15604 vector double vec_sel (vector double, vector double, vector unsigned long);
15605 vector long vec_sel (vector long, vector long, vector long);
15606 vector long vec_sel (vector long, vector long, vector unsigned long);
15607 vector long vec_sel (vector long, vector long, vector bool long);
15608 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15610 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15611 vector unsigned long);
15612 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15614 vector double vec_splats (double);
15615 vector signed long vec_splats (signed long);
15616 vector unsigned long vec_splats (unsigned long);
15617 vector float vec_sqrt (vector float);
15618 vector double vec_sqrt (vector double);
15619 void vec_st (vector double, int, vector double *);
15620 void vec_st (vector double, int, double *);
15621 vector double vec_sub (vector double, vector double);
15622 vector double vec_trunc (vector double);
15623 vector double vec_xor (vector double, vector double);
15624 vector double vec_xor (vector double, vector bool long);
15625 vector double vec_xor (vector bool long, vector double);
15626 vector long vec_xor (vector long, vector long);
15627 vector long vec_xor (vector long, vector bool long);
15628 vector long vec_xor (vector bool long, vector long);
15629 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15630 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15631 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15632 int vec_all_eq (vector double, vector double);
15633 int vec_all_ge (vector double, vector double);
15634 int vec_all_gt (vector double, vector double);
15635 int vec_all_le (vector double, vector double);
15636 int vec_all_lt (vector double, vector double);
15637 int vec_all_nan (vector double);
15638 int vec_all_ne (vector double, vector double);
15639 int vec_all_nge (vector double, vector double);
15640 int vec_all_ngt (vector double, vector double);
15641 int vec_all_nle (vector double, vector double);
15642 int vec_all_nlt (vector double, vector double);
15643 int vec_all_numeric (vector double);
15644 int vec_any_eq (vector double, vector double);
15645 int vec_any_ge (vector double, vector double);
15646 int vec_any_gt (vector double, vector double);
15647 int vec_any_le (vector double, vector double);
15648 int vec_any_lt (vector double, vector double);
15649 int vec_any_nan (vector double);
15650 int vec_any_ne (vector double, vector double);
15651 int vec_any_nge (vector double, vector double);
15652 int vec_any_ngt (vector double, vector double);
15653 int vec_any_nle (vector double, vector double);
15654 int vec_any_nlt (vector double, vector double);
15655 int vec_any_numeric (vector double);
15657 vector double vec_vsx_ld (int, const vector double *);
15658 vector double vec_vsx_ld (int, const double *);
15659 vector float vec_vsx_ld (int, const vector float *);
15660 vector float vec_vsx_ld (int, const float *);
15661 vector bool int vec_vsx_ld (int, const vector bool int *);
15662 vector signed int vec_vsx_ld (int, const vector signed int *);
15663 vector signed int vec_vsx_ld (int, const int *);
15664 vector signed int vec_vsx_ld (int, const long *);
15665 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15666 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15667 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15668 vector bool short vec_vsx_ld (int, const vector bool short *);
15669 vector pixel vec_vsx_ld (int, const vector pixel *);
15670 vector signed short vec_vsx_ld (int, const vector signed short *);
15671 vector signed short vec_vsx_ld (int, const short *);
15672 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15673 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15674 vector bool char vec_vsx_ld (int, const vector bool char *);
15675 vector signed char vec_vsx_ld (int, const vector signed char *);
15676 vector signed char vec_vsx_ld (int, const signed char *);
15677 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15678 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15680 void vec_vsx_st (vector double, int, vector double *);
15681 void vec_vsx_st (vector double, int, double *);
15682 void vec_vsx_st (vector float, int, vector float *);
15683 void vec_vsx_st (vector float, int, float *);
15684 void vec_vsx_st (vector signed int, int, vector signed int *);
15685 void vec_vsx_st (vector signed int, int, int *);
15686 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15687 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15688 void vec_vsx_st (vector bool int, int, vector bool int *);
15689 void vec_vsx_st (vector bool int, int, unsigned int *);
15690 void vec_vsx_st (vector bool int, int, int *);
15691 void vec_vsx_st (vector signed short, int, vector signed short *);
15692 void vec_vsx_st (vector signed short, int, short *);
15693 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15694 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15695 void vec_vsx_st (vector bool short, int, vector bool short *);
15696 void vec_vsx_st (vector bool short, int, unsigned short *);
15697 void vec_vsx_st (vector pixel, int, vector pixel *);
15698 void vec_vsx_st (vector pixel, int, unsigned short *);
15699 void vec_vsx_st (vector pixel, int, short *);
15700 void vec_vsx_st (vector bool short, int, short *);
15701 void vec_vsx_st (vector signed char, int, vector signed char *);
15702 void vec_vsx_st (vector signed char, int, signed char *);
15703 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15704 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15705 void vec_vsx_st (vector bool char, int, vector bool char *);
15706 void vec_vsx_st (vector bool char, int, unsigned char *);
15707 void vec_vsx_st (vector bool char, int, signed char *);
15709 vector double vec_xxpermdi (vector double, vector double, int);
15710 vector float vec_xxpermdi (vector float, vector float, int);
15711 vector long long vec_xxpermdi (vector long long, vector long long, int);
15712 vector unsigned long long vec_xxpermdi (vector unsigned long long,
15713 vector unsigned long long, int);
15714 vector int vec_xxpermdi (vector int, vector int, int);
15715 vector unsigned int vec_xxpermdi (vector unsigned int,
15716 vector unsigned int, int);
15717 vector short vec_xxpermdi (vector short, vector short, int);
15718 vector unsigned short vec_xxpermdi (vector unsigned short,
15719 vector unsigned short, int);
15720 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
15721 vector unsigned char vec_xxpermdi (vector unsigned char,
15722 vector unsigned char, int);
15724 vector double vec_xxsldi (vector double, vector double, int);
15725 vector float vec_xxsldi (vector float, vector float, int);
15726 vector long long vec_xxsldi (vector long long, vector long long, int);
15727 vector unsigned long long vec_xxsldi (vector unsigned long long,
15728 vector unsigned long long, int);
15729 vector int vec_xxsldi (vector int, vector int, int);
15730 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
15731 vector short vec_xxsldi (vector short, vector short, int);
15732 vector unsigned short vec_xxsldi (vector unsigned short,
15733 vector unsigned short, int);
15734 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
15735 vector unsigned char vec_xxsldi (vector unsigned char,
15736 vector unsigned char, int);
15739 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
15740 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
15741 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
15742 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
15743 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
15745 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15746 instruction set is available, the following additional functions are
15747 available for both 32-bit and 64-bit targets. For 64-bit targets, you
15748 can use @var{vector long} instead of @var{vector long long},
15749 @var{vector bool long} instead of @var{vector bool long long}, and
15750 @var{vector unsigned long} instead of @var{vector unsigned long long}.
15753 vector long long vec_abs (vector long long);
15755 vector long long vec_add (vector long long, vector long long);
15756 vector unsigned long long vec_add (vector unsigned long long,
15757 vector unsigned long long);
15759 int vec_all_eq (vector long long, vector long long);
15760 int vec_all_eq (vector unsigned long long, vector unsigned long long);
15761 int vec_all_ge (vector long long, vector long long);
15762 int vec_all_ge (vector unsigned long long, vector unsigned long long);
15763 int vec_all_gt (vector long long, vector long long);
15764 int vec_all_gt (vector unsigned long long, vector unsigned long long);
15765 int vec_all_le (vector long long, vector long long);
15766 int vec_all_le (vector unsigned long long, vector unsigned long long);
15767 int vec_all_lt (vector long long, vector long long);
15768 int vec_all_lt (vector unsigned long long, vector unsigned long long);
15769 int vec_all_ne (vector long long, vector long long);
15770 int vec_all_ne (vector unsigned long long, vector unsigned long long);
15772 int vec_any_eq (vector long long, vector long long);
15773 int vec_any_eq (vector unsigned long long, vector unsigned long long);
15774 int vec_any_ge (vector long long, vector long long);
15775 int vec_any_ge (vector unsigned long long, vector unsigned long long);
15776 int vec_any_gt (vector long long, vector long long);
15777 int vec_any_gt (vector unsigned long long, vector unsigned long long);
15778 int vec_any_le (vector long long, vector long long);
15779 int vec_any_le (vector unsigned long long, vector unsigned long long);
15780 int vec_any_lt (vector long long, vector long long);
15781 int vec_any_lt (vector unsigned long long, vector unsigned long long);
15782 int vec_any_ne (vector long long, vector long long);
15783 int vec_any_ne (vector unsigned long long, vector unsigned long long);
15785 vector long long vec_eqv (vector long long, vector long long);
15786 vector long long vec_eqv (vector bool long long, vector long long);
15787 vector long long vec_eqv (vector long long, vector bool long long);
15788 vector unsigned long long vec_eqv (vector unsigned long long,
15789 vector unsigned long long);
15790 vector unsigned long long vec_eqv (vector bool long long,
15791 vector unsigned long long);
15792 vector unsigned long long vec_eqv (vector unsigned long long,
15793 vector bool long long);
15794 vector int vec_eqv (vector int, vector int);
15795 vector int vec_eqv (vector bool int, vector int);
15796 vector int vec_eqv (vector int, vector bool int);
15797 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
15798 vector unsigned int vec_eqv (vector bool unsigned int,
15799 vector unsigned int);
15800 vector unsigned int vec_eqv (vector unsigned int,
15801 vector bool unsigned int);
15802 vector short vec_eqv (vector short, vector short);
15803 vector short vec_eqv (vector bool short, vector short);
15804 vector short vec_eqv (vector short, vector bool short);
15805 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
15806 vector unsigned short vec_eqv (vector bool unsigned short,
15807 vector unsigned short);
15808 vector unsigned short vec_eqv (vector unsigned short,
15809 vector bool unsigned short);
15810 vector signed char vec_eqv (vector signed char, vector signed char);
15811 vector signed char vec_eqv (vector bool signed char, vector signed char);
15812 vector signed char vec_eqv (vector signed char, vector bool signed char);
15813 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
15814 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
15815 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
15817 vector long long vec_max (vector long long, vector long long);
15818 vector unsigned long long vec_max (vector unsigned long long,
15819 vector unsigned long long);
15821 vector signed int vec_mergee (vector signed int, vector signed int);
15822 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
15823 vector bool int vec_mergee (vector bool int, vector bool int);
15825 vector signed int vec_mergeo (vector signed int, vector signed int);
15826 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
15827 vector bool int vec_mergeo (vector bool int, vector bool int);
15829 vector long long vec_min (vector long long, vector long long);
15830 vector unsigned long long vec_min (vector unsigned long long,
15831 vector unsigned long long);
15833 vector long long vec_nand (vector long long, vector long long);
15834 vector long long vec_nand (vector bool long long, vector long long);
15835 vector long long vec_nand (vector long long, vector bool long long);
15836 vector unsigned long long vec_nand (vector unsigned long long,
15837 vector unsigned long long);
15838 vector unsigned long long vec_nand (vector bool long long,
15839 vector unsigned long long);
15840 vector unsigned long long vec_nand (vector unsigned long long,
15841 vector bool long long);
15842 vector int vec_nand (vector int, vector int);
15843 vector int vec_nand (vector bool int, vector int);
15844 vector int vec_nand (vector int, vector bool int);
15845 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
15846 vector unsigned int vec_nand (vector bool unsigned int,
15847 vector unsigned int);
15848 vector unsigned int vec_nand (vector unsigned int,
15849 vector bool unsigned int);
15850 vector short vec_nand (vector short, vector short);
15851 vector short vec_nand (vector bool short, vector short);
15852 vector short vec_nand (vector short, vector bool short);
15853 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
15854 vector unsigned short vec_nand (vector bool unsigned short,
15855 vector unsigned short);
15856 vector unsigned short vec_nand (vector unsigned short,
15857 vector bool unsigned short);
15858 vector signed char vec_nand (vector signed char, vector signed char);
15859 vector signed char vec_nand (vector bool signed char, vector signed char);
15860 vector signed char vec_nand (vector signed char, vector bool signed char);
15861 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
15862 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
15863 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
15865 vector long long vec_orc (vector long long, vector long long);
15866 vector long long vec_orc (vector bool long long, vector long long);
15867 vector long long vec_orc (vector long long, vector bool long long);
15868 vector unsigned long long vec_orc (vector unsigned long long,
15869 vector unsigned long long);
15870 vector unsigned long long vec_orc (vector bool long long,
15871 vector unsigned long long);
15872 vector unsigned long long vec_orc (vector unsigned long long,
15873 vector bool long long);
15874 vector int vec_orc (vector int, vector int);
15875 vector int vec_orc (vector bool int, vector int);
15876 vector int vec_orc (vector int, vector bool int);
15877 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
15878 vector unsigned int vec_orc (vector bool unsigned int,
15879 vector unsigned int);
15880 vector unsigned int vec_orc (vector unsigned int,
15881 vector bool unsigned int);
15882 vector short vec_orc (vector short, vector short);
15883 vector short vec_orc (vector bool short, vector short);
15884 vector short vec_orc (vector short, vector bool short);
15885 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
15886 vector unsigned short vec_orc (vector bool unsigned short,
15887 vector unsigned short);
15888 vector unsigned short vec_orc (vector unsigned short,
15889 vector bool unsigned short);
15890 vector signed char vec_orc (vector signed char, vector signed char);
15891 vector signed char vec_orc (vector bool signed char, vector signed char);
15892 vector signed char vec_orc (vector signed char, vector bool signed char);
15893 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
15894 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
15895 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
15897 vector int vec_pack (vector long long, vector long long);
15898 vector unsigned int vec_pack (vector unsigned long long,
15899 vector unsigned long long);
15900 vector bool int vec_pack (vector bool long long, vector bool long long);
15902 vector int vec_packs (vector long long, vector long long);
15903 vector unsigned int vec_packs (vector unsigned long long,
15904 vector unsigned long long);
15906 vector unsigned int vec_packsu (vector long long, vector long long);
15907 vector unsigned int vec_packsu (vector unsigned long long,
15908 vector unsigned long long);
15910 vector long long vec_rl (vector long long,
15911 vector unsigned long long);
15912 vector long long vec_rl (vector unsigned long long,
15913 vector unsigned long long);
15915 vector long long vec_sl (vector long long, vector unsigned long long);
15916 vector long long vec_sl (vector unsigned long long,
15917 vector unsigned long long);
15919 vector long long vec_sr (vector long long, vector unsigned long long);
15920 vector unsigned long long char vec_sr (vector unsigned long long,
15921 vector unsigned long long);
15923 vector long long vec_sra (vector long long, vector unsigned long long);
15924 vector unsigned long long vec_sra (vector unsigned long long,
15925 vector unsigned long long);
15927 vector long long vec_sub (vector long long, vector long long);
15928 vector unsigned long long vec_sub (vector unsigned long long,
15929 vector unsigned long long);
15931 vector long long vec_unpackh (vector int);
15932 vector unsigned long long vec_unpackh (vector unsigned int);
15934 vector long long vec_unpackl (vector int);
15935 vector unsigned long long vec_unpackl (vector unsigned int);
15937 vector long long vec_vaddudm (vector long long, vector long long);
15938 vector long long vec_vaddudm (vector bool long long, vector long long);
15939 vector long long vec_vaddudm (vector long long, vector bool long long);
15940 vector unsigned long long vec_vaddudm (vector unsigned long long,
15941 vector unsigned long long);
15942 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
15943 vector unsigned long long);
15944 vector unsigned long long vec_vaddudm (vector unsigned long long,
15945 vector bool unsigned long long);
15947 vector long long vec_vbpermq (vector signed char, vector signed char);
15948 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
15950 vector long long vec_cntlz (vector long long);
15951 vector unsigned long long vec_cntlz (vector unsigned long long);
15952 vector int vec_cntlz (vector int);
15953 vector unsigned int vec_cntlz (vector int);
15954 vector short vec_cntlz (vector short);
15955 vector unsigned short vec_cntlz (vector unsigned short);
15956 vector signed char vec_cntlz (vector signed char);
15957 vector unsigned char vec_cntlz (vector unsigned char);
15959 vector long long vec_vclz (vector long long);
15960 vector unsigned long long vec_vclz (vector unsigned long long);
15961 vector int vec_vclz (vector int);
15962 vector unsigned int vec_vclz (vector int);
15963 vector short vec_vclz (vector short);
15964 vector unsigned short vec_vclz (vector unsigned short);
15965 vector signed char vec_vclz (vector signed char);
15966 vector unsigned char vec_vclz (vector unsigned char);
15968 vector signed char vec_vclzb (vector signed char);
15969 vector unsigned char vec_vclzb (vector unsigned char);
15971 vector long long vec_vclzd (vector long long);
15972 vector unsigned long long vec_vclzd (vector unsigned long long);
15974 vector short vec_vclzh (vector short);
15975 vector unsigned short vec_vclzh (vector unsigned short);
15977 vector int vec_vclzw (vector int);
15978 vector unsigned int vec_vclzw (vector int);
15980 vector signed char vec_vgbbd (vector signed char);
15981 vector unsigned char vec_vgbbd (vector unsigned char);
15983 vector long long vec_vmaxsd (vector long long, vector long long);
15985 vector unsigned long long vec_vmaxud (vector unsigned long long,
15986 unsigned vector long long);
15988 vector long long vec_vminsd (vector long long, vector long long);
15990 vector unsigned long long vec_vminud (vector long long,
15993 vector int vec_vpksdss (vector long long, vector long long);
15994 vector unsigned int vec_vpksdss (vector long long, vector long long);
15996 vector unsigned int vec_vpkudus (vector unsigned long long,
15997 vector unsigned long long);
15999 vector int vec_vpkudum (vector long long, vector long long);
16000 vector unsigned int vec_vpkudum (vector unsigned long long,
16001 vector unsigned long long);
16002 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
16004 vector long long vec_vpopcnt (vector long long);
16005 vector unsigned long long vec_vpopcnt (vector unsigned long long);
16006 vector int vec_vpopcnt (vector int);
16007 vector unsigned int vec_vpopcnt (vector int);
16008 vector short vec_vpopcnt (vector short);
16009 vector unsigned short vec_vpopcnt (vector unsigned short);
16010 vector signed char vec_vpopcnt (vector signed char);
16011 vector unsigned char vec_vpopcnt (vector unsigned char);
16013 vector signed char vec_vpopcntb (vector signed char);
16014 vector unsigned char vec_vpopcntb (vector unsigned char);
16016 vector long long vec_vpopcntd (vector long long);
16017 vector unsigned long long vec_vpopcntd (vector unsigned long long);
16019 vector short vec_vpopcnth (vector short);
16020 vector unsigned short vec_vpopcnth (vector unsigned short);
16022 vector int vec_vpopcntw (vector int);
16023 vector unsigned int vec_vpopcntw (vector int);
16025 vector long long vec_vrld (vector long long, vector unsigned long long);
16026 vector unsigned long long vec_vrld (vector unsigned long long,
16027 vector unsigned long long);
16029 vector long long vec_vsld (vector long long, vector unsigned long long);
16030 vector long long vec_vsld (vector unsigned long long,
16031 vector unsigned long long);
16033 vector long long vec_vsrad (vector long long, vector unsigned long long);
16034 vector unsigned long long vec_vsrad (vector unsigned long long,
16035 vector unsigned long long);
16037 vector long long vec_vsrd (vector long long, vector unsigned long long);
16038 vector unsigned long long char vec_vsrd (vector unsigned long long,
16039 vector unsigned long long);
16041 vector long long vec_vsubudm (vector long long, vector long long);
16042 vector long long vec_vsubudm (vector bool long long, vector long long);
16043 vector long long vec_vsubudm (vector long long, vector bool long long);
16044 vector unsigned long long vec_vsubudm (vector unsigned long long,
16045 vector unsigned long long);
16046 vector unsigned long long vec_vsubudm (vector bool long long,
16047 vector unsigned long long);
16048 vector unsigned long long vec_vsubudm (vector unsigned long long,
16049 vector bool long long);
16051 vector long long vec_vupkhsw (vector int);
16052 vector unsigned long long vec_vupkhsw (vector unsigned int);
16054 vector long long vec_vupklsw (vector int);
16055 vector unsigned long long vec_vupklsw (vector int);
16058 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16059 instruction set is available, the following additional functions are
16060 available for 64-bit targets. New vector types
16061 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
16062 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
16065 The normal vector extract, and set operations work on
16066 @var{vector __int128_t} and @var{vector __uint128_t} types,
16067 but the index value must be 0.
16070 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
16071 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
16073 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
16074 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
16076 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
16077 vector __int128_t);
16078 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
16079 vector __uint128_t);
16081 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
16082 vector __int128_t);
16083 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
16084 vector __uint128_t);
16086 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
16087 vector __int128_t);
16088 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
16089 vector __uint128_t);
16091 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
16092 vector __int128_t);
16093 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
16094 vector __uint128_t);
16096 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
16097 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
16099 __int128_t vec_vsubuqm (__int128_t, __int128_t);
16100 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
16102 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
16103 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
16104 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
16105 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
16106 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
16107 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
16108 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
16109 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
16110 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
16111 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
16114 If the cryptographic instructions are enabled (@option{-mcrypto} or
16115 @option{-mcpu=power8}), the following builtins are enabled.
16118 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
16120 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
16121 vector unsigned long long);
16123 vector unsigned long long __builtin_crypto_vcipherlast
16124 (vector unsigned long long,
16125 vector unsigned long long);
16127 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
16128 vector unsigned long long);
16130 vector unsigned long long __builtin_crypto_vncipherlast
16131 (vector unsigned long long,
16132 vector unsigned long long);
16134 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
16135 vector unsigned char,
16136 vector unsigned char);
16138 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
16139 vector unsigned short,
16140 vector unsigned short);
16142 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
16143 vector unsigned int,
16144 vector unsigned int);
16146 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
16147 vector unsigned long long,
16148 vector unsigned long long);
16150 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
16151 vector unsigned char);
16153 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
16154 vector unsigned short);
16156 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
16157 vector unsigned int);
16159 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
16160 vector unsigned long long);
16162 vector unsigned long long __builtin_crypto_vshasigmad
16163 (vector unsigned long long, int, int);
16165 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
16169 The second argument to the @var{__builtin_crypto_vshasigmad} and
16170 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
16171 integer that is 0 or 1. The third argument to these builtin functions
16172 must be a constant integer in the range of 0 to 15.
16174 @node PowerPC Hardware Transactional Memory Built-in Functions
16175 @subsection PowerPC Hardware Transactional Memory Built-in Functions
16176 GCC provides two interfaces for accessing the Hardware Transactional
16177 Memory (HTM) instructions available on some of the PowerPC family
16178 of processors (eg, POWER8). The two interfaces come in a low level
16179 interface, consisting of built-in functions specific to PowerPC and a
16180 higher level interface consisting of inline functions that are common
16181 between PowerPC and S/390.
16183 @subsubsection PowerPC HTM Low Level Built-in Functions
16185 The following low level built-in functions are available with
16186 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16187 They all generate the machine instruction that is part of the name.
16189 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16190 the full 4-bit condition register value set by their associated hardware
16191 instruction. The header file @code{htmintrin.h} defines some macros that can
16192 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16193 returns a simple true or false value depending on whether a transaction was
16194 successfully started or not. The arguments of the builtins match exactly the
16195 type and order of the associated hardware instruction's operands, except for
16196 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16197 Refer to the ISA manual for a description of each instruction's operands.
16200 unsigned int __builtin_tbegin (unsigned int)
16201 unsigned int __builtin_tend (unsigned int)
16203 unsigned int __builtin_tabort (unsigned int)
16204 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16205 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16206 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16207 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16209 unsigned int __builtin_tcheck (void)
16210 unsigned int __builtin_treclaim (unsigned int)
16211 unsigned int __builtin_trechkpt (void)
16212 unsigned int __builtin_tsr (unsigned int)
16215 In addition to the above HTM built-ins, we have added built-ins for
16216 some common extended mnemonics of the HTM instructions:
16219 unsigned int __builtin_tendall (void)
16220 unsigned int __builtin_tresume (void)
16221 unsigned int __builtin_tsuspend (void)
16224 Note that the semantics of the above HTM builtins are required to mimic
16225 the locking semantics used for critical sections. Builtins that are used
16226 to create a new transaction or restart a suspended transaction must have
16227 lock acquisition like semantics while those builtins that end or suspend a
16228 transaction must have lock release like semantics. Specifically, this must
16229 mimic lock semantics as specified by C++11, for example: Lock acquisition is
16230 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
16231 that returns 0, and lock release is as-if an execution of
16232 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
16233 implicit implementation-defined lock used for all transactions. The HTM
16234 instructions associated with with the builtins inherently provide the
16235 correct acquisition and release hardware barriers required. However,
16236 the compiler must also be prohibited from moving loads and stores across
16237 the builtins in a way that would violate their semantics. This has been
16238 accomplished by adding memory barriers to the associated HTM instructions
16239 (which is a conservative approach to provide acquire and release semantics).
16240 Earlier versions of the compiler did not treat the HTM instructions as
16241 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
16242 be used to determine whether the current compiler treats HTM instructions
16243 as memory barriers or not. This allows the user to explicitly add memory
16244 barriers to their code when using an older version of the compiler.
16246 The following set of built-in functions are available to gain access
16247 to the HTM specific special purpose registers.
16250 unsigned long __builtin_get_texasr (void)
16251 unsigned long __builtin_get_texasru (void)
16252 unsigned long __builtin_get_tfhar (void)
16253 unsigned long __builtin_get_tfiar (void)
16255 void __builtin_set_texasr (unsigned long);
16256 void __builtin_set_texasru (unsigned long);
16257 void __builtin_set_tfhar (unsigned long);
16258 void __builtin_set_tfiar (unsigned long);
16261 Example usage of these low level built-in functions may look like:
16264 #include <htmintrin.h>
16266 int num_retries = 10;
16270 if (__builtin_tbegin (0))
16272 /* Transaction State Initiated. */
16273 if (is_locked (lock))
16274 __builtin_tabort (0);
16275 ... transaction code...
16276 __builtin_tend (0);
16281 /* Transaction State Failed. Use locks if the transaction
16282 failure is "persistent" or we've tried too many times. */
16283 if (num_retries-- <= 0
16284 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16286 acquire_lock (lock);
16287 ... non transactional fallback path...
16288 release_lock (lock);
16295 One final built-in function has been added that returns the value of
16296 the 2-bit Transaction State field of the Machine Status Register (MSR)
16297 as stored in @code{CR0}.
16300 unsigned long __builtin_ttest (void)
16303 This built-in can be used to determine the current transaction state
16304 using the following code example:
16307 #include <htmintrin.h>
16309 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16311 if (tx_state == _HTM_TRANSACTIONAL)
16313 /* Code to use in transactional state. */
16315 else if (tx_state == _HTM_NONTRANSACTIONAL)
16317 /* Code to use in non-transactional state. */
16319 else if (tx_state == _HTM_SUSPENDED)
16321 /* Code to use in transaction suspended state. */
16325 @subsubsection PowerPC HTM High Level Inline Functions
16327 The following high level HTM interface is made available by including
16328 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16329 where CPU is `power8' or later. This interface is common between PowerPC
16330 and S/390, allowing users to write one HTM source implementation that
16331 can be compiled and executed on either system.
16334 long __TM_simple_begin (void)
16335 long __TM_begin (void* const TM_buff)
16336 long __TM_end (void)
16337 void __TM_abort (void)
16338 void __TM_named_abort (unsigned char const code)
16339 void __TM_resume (void)
16340 void __TM_suspend (void)
16342 long __TM_is_user_abort (void* const TM_buff)
16343 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16344 long __TM_is_illegal (void* const TM_buff)
16345 long __TM_is_footprint_exceeded (void* const TM_buff)
16346 long __TM_nesting_depth (void* const TM_buff)
16347 long __TM_is_nested_too_deep(void* const TM_buff)
16348 long __TM_is_conflict(void* const TM_buff)
16349 long __TM_is_failure_persistent(void* const TM_buff)
16350 long __TM_failure_address(void* const TM_buff)
16351 long long __TM_failure_code(void* const TM_buff)
16354 Using these common set of HTM inline functions, we can create
16355 a more portable version of the HTM example in the previous
16356 section that will work on either PowerPC or S/390:
16359 #include <htmxlintrin.h>
16361 int num_retries = 10;
16362 TM_buff_type TM_buff;
16366 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16368 /* Transaction State Initiated. */
16369 if (is_locked (lock))
16371 ... transaction code...
16377 /* Transaction State Failed. Use locks if the transaction
16378 failure is "persistent" or we've tried too many times. */
16379 if (num_retries-- <= 0
16380 || __TM_is_failure_persistent (TM_buff))
16382 acquire_lock (lock);
16383 ... non transactional fallback path...
16384 release_lock (lock);
16391 @node RX Built-in Functions
16392 @subsection RX Built-in Functions
16393 GCC supports some of the RX instructions which cannot be expressed in
16394 the C programming language via the use of built-in functions. The
16395 following functions are supported:
16397 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16398 Generates the @code{brk} machine instruction.
16401 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16402 Generates the @code{clrpsw} machine instruction to clear the specified
16403 bit in the processor status word.
16406 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16407 Generates the @code{int} machine instruction to generate an interrupt
16408 with the specified value.
16411 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16412 Generates the @code{machi} machine instruction to add the result of
16413 multiplying the top 16 bits of the two arguments into the
16417 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16418 Generates the @code{maclo} machine instruction to add the result of
16419 multiplying the bottom 16 bits of the two arguments into the
16423 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16424 Generates the @code{mulhi} machine instruction to place the result of
16425 multiplying the top 16 bits of the two arguments into the
16429 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16430 Generates the @code{mullo} machine instruction to place the result of
16431 multiplying the bottom 16 bits of the two arguments into the
16435 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16436 Generates the @code{mvfachi} machine instruction to read the top
16437 32 bits of the accumulator.
16440 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16441 Generates the @code{mvfacmi} machine instruction to read the middle
16442 32 bits of the accumulator.
16445 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16446 Generates the @code{mvfc} machine instruction which reads the control
16447 register specified in its argument and returns its value.
16450 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16451 Generates the @code{mvtachi} machine instruction to set the top
16452 32 bits of the accumulator.
16455 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16456 Generates the @code{mvtaclo} machine instruction to set the bottom
16457 32 bits of the accumulator.
16460 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16461 Generates the @code{mvtc} machine instruction which sets control
16462 register number @code{reg} to @code{val}.
16465 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16466 Generates the @code{mvtipl} machine instruction set the interrupt
16470 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16471 Generates the @code{racw} machine instruction to round the accumulator
16472 according to the specified mode.
16475 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16476 Generates the @code{revw} machine instruction which swaps the bytes in
16477 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16478 and also bits 16--23 occupy bits 24--31 and vice versa.
16481 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16482 Generates the @code{rmpa} machine instruction which initiates a
16483 repeated multiply and accumulate sequence.
16486 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16487 Generates the @code{round} machine instruction which returns the
16488 floating-point argument rounded according to the current rounding mode
16489 set in the floating-point status word register.
16492 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16493 Generates the @code{sat} machine instruction which returns the
16494 saturated value of the argument.
16497 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16498 Generates the @code{setpsw} machine instruction to set the specified
16499 bit in the processor status word.
16502 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16503 Generates the @code{wait} machine instruction.
16506 @node S/390 System z Built-in Functions
16507 @subsection S/390 System z Built-in Functions
16508 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16509 Generates the @code{tbegin} machine instruction starting a
16510 non-constraint hardware transaction. If the parameter is non-NULL the
16511 memory area is used to store the transaction diagnostic buffer and
16512 will be passed as first operand to @code{tbegin}. This buffer can be
16513 defined using the @code{struct __htm_tdb} C struct defined in
16514 @code{htmintrin.h} and must reside on a double-word boundary. The
16515 second tbegin operand is set to @code{0xff0c}. This enables
16516 save/restore of all GPRs and disables aborts for FPR and AR
16517 manipulations inside the transaction body. The condition code set by
16518 the tbegin instruction is returned as integer value. The tbegin
16519 instruction by definition overwrites the content of all FPRs. The
16520 compiler will generate code which saves and restores the FPRs. For
16521 soft-float code it is recommended to used the @code{*_nofloat}
16522 variant. In order to prevent a TDB from being written it is required
16523 to pass an constant zero value as parameter. Passing the zero value
16524 through a variable is not sufficient. Although modifications of
16525 access registers inside the transaction will not trigger an
16526 transaction abort it is not supported to actually modify them. Access
16527 registers do not get saved when entering a transaction. They will have
16528 undefined state when reaching the abort code.
16531 Macros for the possible return codes of tbegin are defined in the
16532 @code{htmintrin.h} header file:
16535 @item _HTM_TBEGIN_STARTED
16536 @code{tbegin} has been executed as part of normal processing. The
16537 transaction body is supposed to be executed.
16538 @item _HTM_TBEGIN_INDETERMINATE
16539 The transaction was aborted due to an indeterminate condition which
16540 might be persistent.
16541 @item _HTM_TBEGIN_TRANSIENT
16542 The transaction aborted due to a transient failure. The transaction
16543 should be re-executed in that case.
16544 @item _HTM_TBEGIN_PERSISTENT
16545 The transaction aborted due to a persistent failure. Re-execution
16546 under same circumstances will not be productive.
16549 @defmac _HTM_FIRST_USER_ABORT_CODE
16550 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16551 specifies the first abort code which can be used for
16552 @code{__builtin_tabort}. Values below this threshold are reserved for
16556 @deftp {Data type} {struct __htm_tdb}
16557 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16558 the structure of the transaction diagnostic block as specified in the
16559 Principles of Operation manual chapter 5-91.
16562 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16563 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16564 Using this variant in code making use of FPRs will leave the FPRs in
16565 undefined state when entering the transaction abort handler code.
16568 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16569 In addition to @code{__builtin_tbegin} a loop for transient failures
16570 is generated. If tbegin returns a condition code of 2 the transaction
16571 will be retried as often as specified in the second argument. The
16572 perform processor assist instruction is used to tell the CPU about the
16573 number of fails so far.
16576 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16577 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16578 restores. Using this variant in code making use of FPRs will leave
16579 the FPRs in undefined state when entering the transaction abort
16583 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16584 Generates the @code{tbeginc} machine instruction starting a constraint
16585 hardware transaction. The second operand is set to @code{0xff08}.
16588 @deftypefn {Built-in Function} int __builtin_tend (void)
16589 Generates the @code{tend} machine instruction finishing a transaction
16590 and making the changes visible to other threads. The condition code
16591 generated by tend is returned as integer value.
16594 @deftypefn {Built-in Function} void __builtin_tabort (int)
16595 Generates the @code{tabort} machine instruction with the specified
16596 abort code. Abort codes from 0 through 255 are reserved and will
16597 result in an error message.
16600 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16601 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16602 integer parameter is loaded into rX and a value of zero is loaded into
16603 rY. The integer parameter specifies the number of times the
16604 transaction repeatedly aborted.
16607 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16608 Generates the @code{etnd} machine instruction. The current nesting
16609 depth is returned as integer value. For a nesting depth of 0 the code
16610 is not executed as part of an transaction.
16613 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16615 Generates the @code{ntstg} machine instruction. The second argument
16616 is written to the first arguments location. The store operation will
16617 not be rolled-back in case of an transaction abort.
16620 @node SH Built-in Functions
16621 @subsection SH Built-in Functions
16622 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16623 families of processors:
16625 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16626 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16627 used by system code that manages threads and execution contexts. The compiler
16628 normally does not generate code that modifies the contents of @samp{GBR} and
16629 thus the value is preserved across function calls. Changing the @samp{GBR}
16630 value in user code must be done with caution, since the compiler might use
16631 @samp{GBR} in order to access thread local variables.
16635 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16636 Returns the value that is currently set in the @samp{GBR} register.
16637 Memory loads and stores that use the thread pointer as a base address are
16638 turned into @samp{GBR} based displacement loads and stores, if possible.
16646 int get_tcb_value (void)
16648 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16649 return ((my_tcb*)__builtin_thread_pointer ())->c;
16655 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16656 Returns the value that is currently set in the @samp{FPSCR} register.
16659 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16660 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16661 preserving the current values of the FR, SZ and PR bits.
16664 @node SPARC VIS Built-in Functions
16665 @subsection SPARC VIS Built-in Functions
16667 GCC supports SIMD operations on the SPARC using both the generic vector
16668 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16669 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16670 switch, the VIS extension is exposed as the following built-in functions:
16673 typedef int v1si __attribute__ ((vector_size (4)));
16674 typedef int v2si __attribute__ ((vector_size (8)));
16675 typedef short v4hi __attribute__ ((vector_size (8)));
16676 typedef short v2hi __attribute__ ((vector_size (4)));
16677 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16678 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16680 void __builtin_vis_write_gsr (int64_t);
16681 int64_t __builtin_vis_read_gsr (void);
16683 void * __builtin_vis_alignaddr (void *, long);
16684 void * __builtin_vis_alignaddrl (void *, long);
16685 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16686 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16687 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16688 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16690 v4hi __builtin_vis_fexpand (v4qi);
16692 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16693 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16694 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16695 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16696 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16697 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16698 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16700 v4qi __builtin_vis_fpack16 (v4hi);
16701 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16702 v2hi __builtin_vis_fpackfix (v2si);
16703 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16705 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
16707 long __builtin_vis_edge8 (void *, void *);
16708 long __builtin_vis_edge8l (void *, void *);
16709 long __builtin_vis_edge16 (void *, void *);
16710 long __builtin_vis_edge16l (void *, void *);
16711 long __builtin_vis_edge32 (void *, void *);
16712 long __builtin_vis_edge32l (void *, void *);
16714 long __builtin_vis_fcmple16 (v4hi, v4hi);
16715 long __builtin_vis_fcmple32 (v2si, v2si);
16716 long __builtin_vis_fcmpne16 (v4hi, v4hi);
16717 long __builtin_vis_fcmpne32 (v2si, v2si);
16718 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
16719 long __builtin_vis_fcmpgt32 (v2si, v2si);
16720 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
16721 long __builtin_vis_fcmpeq32 (v2si, v2si);
16723 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
16724 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
16725 v2si __builtin_vis_fpadd32 (v2si, v2si);
16726 v1si __builtin_vis_fpadd32s (v1si, v1si);
16727 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
16728 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
16729 v2si __builtin_vis_fpsub32 (v2si, v2si);
16730 v1si __builtin_vis_fpsub32s (v1si, v1si);
16732 long __builtin_vis_array8 (long, long);
16733 long __builtin_vis_array16 (long, long);
16734 long __builtin_vis_array32 (long, long);
16737 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
16738 functions also become available:
16741 long __builtin_vis_bmask (long, long);
16742 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
16743 v2si __builtin_vis_bshufflev2si (v2si, v2si);
16744 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
16745 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
16747 long __builtin_vis_edge8n (void *, void *);
16748 long __builtin_vis_edge8ln (void *, void *);
16749 long __builtin_vis_edge16n (void *, void *);
16750 long __builtin_vis_edge16ln (void *, void *);
16751 long __builtin_vis_edge32n (void *, void *);
16752 long __builtin_vis_edge32ln (void *, void *);
16755 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
16756 functions also become available:
16759 void __builtin_vis_cmask8 (long);
16760 void __builtin_vis_cmask16 (long);
16761 void __builtin_vis_cmask32 (long);
16763 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
16765 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
16766 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
16767 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
16768 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
16769 v2si __builtin_vis_fsll16 (v2si, v2si);
16770 v2si __builtin_vis_fslas16 (v2si, v2si);
16771 v2si __builtin_vis_fsrl16 (v2si, v2si);
16772 v2si __builtin_vis_fsra16 (v2si, v2si);
16774 long __builtin_vis_pdistn (v8qi, v8qi);
16776 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
16778 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
16779 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
16781 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
16782 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
16783 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
16784 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
16785 v2si __builtin_vis_fpadds32 (v2si, v2si);
16786 v1si __builtin_vis_fpadds32s (v1si, v1si);
16787 v2si __builtin_vis_fpsubs32 (v2si, v2si);
16788 v1si __builtin_vis_fpsubs32s (v1si, v1si);
16790 long __builtin_vis_fucmple8 (v8qi, v8qi);
16791 long __builtin_vis_fucmpne8 (v8qi, v8qi);
16792 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
16793 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
16795 float __builtin_vis_fhadds (float, float);
16796 double __builtin_vis_fhaddd (double, double);
16797 float __builtin_vis_fhsubs (float, float);
16798 double __builtin_vis_fhsubd (double, double);
16799 float __builtin_vis_fnhadds (float, float);
16800 double __builtin_vis_fnhaddd (double, double);
16802 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
16803 int64_t __builtin_vis_xmulx (int64_t, int64_t);
16804 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
16807 @node SPU Built-in Functions
16808 @subsection SPU Built-in Functions
16810 GCC provides extensions for the SPU processor as described in the
16811 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
16812 found at @uref{http://cell.scei.co.jp/} or
16813 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
16814 implementation differs in several ways.
16819 The optional extension of specifying vector constants in parentheses is
16823 A vector initializer requires no cast if the vector constant is of the
16824 same type as the variable it is initializing.
16827 If @code{signed} or @code{unsigned} is omitted, the signedness of the
16828 vector type is the default signedness of the base type. The default
16829 varies depending on the operating system, so a portable program should
16830 always specify the signedness.
16833 By default, the keyword @code{__vector} is added. The macro
16834 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
16838 GCC allows using a @code{typedef} name as the type specifier for a
16842 For C, overloaded functions are implemented with macros so the following
16846 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
16850 Since @code{spu_add} is a macro, the vector constant in the example
16851 is treated as four separate arguments. Wrap the entire argument in
16852 parentheses for this to work.
16855 The extended version of @code{__builtin_expect} is not supported.
16859 @emph{Note:} Only the interface described in the aforementioned
16860 specification is supported. Internally, GCC uses built-in functions to
16861 implement the required functionality, but these are not supported and
16862 are subject to change without notice.
16864 @node TI C6X Built-in Functions
16865 @subsection TI C6X Built-in Functions
16867 GCC provides intrinsics to access certain instructions of the TI C6X
16868 processors. These intrinsics, listed below, are available after
16869 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
16870 to C6X instructions.
16874 int _sadd (int, int)
16875 int _ssub (int, int)
16876 int _sadd2 (int, int)
16877 int _ssub2 (int, int)
16878 long long _mpy2 (int, int)
16879 long long _smpy2 (int, int)
16880 int _add4 (int, int)
16881 int _sub4 (int, int)
16882 int _saddu4 (int, int)
16884 int _smpy (int, int)
16885 int _smpyh (int, int)
16886 int _smpyhl (int, int)
16887 int _smpylh (int, int)
16889 int _sshl (int, int)
16890 int _subc (int, int)
16892 int _avg2 (int, int)
16893 int _avgu4 (int, int)
16895 int _clrr (int, int)
16896 int _extr (int, int)
16897 int _extru (int, int)
16903 @node TILE-Gx Built-in Functions
16904 @subsection TILE-Gx Built-in Functions
16906 GCC provides intrinsics to access every instruction of the TILE-Gx
16907 processor. The intrinsics are of the form:
16911 unsigned long long __insn_@var{op} (...)
16915 Where @var{op} is the name of the instruction. Refer to the ISA manual
16916 for the complete list of instructions.
16918 GCC also provides intrinsics to directly access the network registers.
16919 The intrinsics are:
16923 unsigned long long __tile_idn0_receive (void)
16924 unsigned long long __tile_idn1_receive (void)
16925 unsigned long long __tile_udn0_receive (void)
16926 unsigned long long __tile_udn1_receive (void)
16927 unsigned long long __tile_udn2_receive (void)
16928 unsigned long long __tile_udn3_receive (void)
16929 void __tile_idn_send (unsigned long long)
16930 void __tile_udn_send (unsigned long long)
16934 The intrinsic @code{void __tile_network_barrier (void)} is used to
16935 guarantee that no network operations before it are reordered with
16938 @node TILEPro Built-in Functions
16939 @subsection TILEPro Built-in Functions
16941 GCC provides intrinsics to access every instruction of the TILEPro
16942 processor. The intrinsics are of the form:
16946 unsigned __insn_@var{op} (...)
16951 where @var{op} is the name of the instruction. Refer to the ISA manual
16952 for the complete list of instructions.
16954 GCC also provides intrinsics to directly access the network registers.
16955 The intrinsics are:
16959 unsigned __tile_idn0_receive (void)
16960 unsigned __tile_idn1_receive (void)
16961 unsigned __tile_sn_receive (void)
16962 unsigned __tile_udn0_receive (void)
16963 unsigned __tile_udn1_receive (void)
16964 unsigned __tile_udn2_receive (void)
16965 unsigned __tile_udn3_receive (void)
16966 void __tile_idn_send (unsigned)
16967 void __tile_sn_send (unsigned)
16968 void __tile_udn_send (unsigned)
16972 The intrinsic @code{void __tile_network_barrier (void)} is used to
16973 guarantee that no network operations before it are reordered with
16976 @node x86 Built-in Functions
16977 @subsection x86 Built-in Functions
16979 These built-in functions are available for the x86-32 and x86-64 family
16980 of computers, depending on the command-line switches used.
16982 If you specify command-line switches such as @option{-msse},
16983 the compiler could use the extended instruction sets even if the built-ins
16984 are not used explicitly in the program. For this reason, applications
16985 that perform run-time CPU detection must compile separate files for each
16986 supported architecture, using the appropriate flags. In particular,
16987 the file containing the CPU detection code should be compiled without
16990 The following machine modes are available for use with MMX built-in functions
16991 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
16992 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
16993 vector of eight 8-bit integers. Some of the built-in functions operate on
16994 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
16996 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
16997 of two 32-bit floating-point values.
16999 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
17000 floating-point values. Some instructions use a vector of four 32-bit
17001 integers, these use @code{V4SI}. Finally, some instructions operate on an
17002 entire vector register, interpreting it as a 128-bit integer, these use mode
17005 In 64-bit mode, the x86-64 family of processors uses additional built-in
17006 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
17007 floating point and @code{TC} 128-bit complex floating-point values.
17009 The following floating-point built-in functions are available in 64-bit
17010 mode. All of them implement the function that is part of the name.
17013 __float128 __builtin_fabsq (__float128)
17014 __float128 __builtin_copysignq (__float128, __float128)
17017 The following built-in function is always available.
17020 @item void __builtin_ia32_pause (void)
17021 Generates the @code{pause} machine instruction with a compiler memory
17025 The following floating-point built-in functions are made available in the
17029 @item __float128 __builtin_infq (void)
17030 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
17031 @findex __builtin_infq
17033 @item __float128 __builtin_huge_valq (void)
17034 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
17035 @findex __builtin_huge_valq
17038 The following built-in functions are always available and can be used to
17039 check the target platform type.
17041 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
17042 This function runs the CPU detection code to check the type of CPU and the
17043 features supported. This built-in function needs to be invoked along with the built-in functions
17044 to check CPU type and features, @code{__builtin_cpu_is} and
17045 @code{__builtin_cpu_supports}, only when used in a function that is
17046 executed before any constructors are called. The CPU detection code is
17047 automatically executed in a very high priority constructor.
17049 For example, this function has to be used in @code{ifunc} resolvers that
17050 check for CPU type using the built-in functions @code{__builtin_cpu_is}
17051 and @code{__builtin_cpu_supports}, or in constructors on targets that
17052 don't support constructor priority.
17055 static void (*resolve_memcpy (void)) (void)
17057 // ifunc resolvers fire before constructors, explicitly call the init
17059 __builtin_cpu_init ();
17060 if (__builtin_cpu_supports ("ssse3"))
17061 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
17063 return default_memcpy;
17066 void *memcpy (void *, const void *, size_t)
17067 __attribute__ ((ifunc ("resolve_memcpy")));
17072 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
17073 This function returns a positive integer if the run-time CPU
17074 is of type @var{cpuname}
17075 and returns @code{0} otherwise. The following CPU names can be detected:
17091 Intel Core i7 Nehalem CPU.
17094 Intel Core i7 Westmere CPU.
17097 Intel Core i7 Sandy Bridge CPU.
17103 AMD Family 10h CPU.
17106 AMD Family 10h Barcelona CPU.
17109 AMD Family 10h Shanghai CPU.
17112 AMD Family 10h Istanbul CPU.
17115 AMD Family 14h CPU.
17118 AMD Family 15h CPU.
17121 AMD Family 15h Bulldozer version 1.
17124 AMD Family 15h Bulldozer version 2.
17127 AMD Family 15h Bulldozer version 3.
17130 AMD Family 15h Bulldozer version 4.
17133 AMD Family 16h CPU.
17136 AMD Family 17h CPU.
17139 Here is an example:
17141 if (__builtin_cpu_is ("corei7"))
17143 do_corei7 (); // Core i7 specific implementation.
17147 do_generic (); // Generic implementation.
17152 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
17153 This function returns a positive integer if the run-time CPU
17154 supports @var{feature}
17155 and returns @code{0} otherwise. The following features can be detected:
17163 POPCNT instruction.
17171 SSSE3 instructions.
17173 SSE4.1 instructions.
17175 SSE4.2 instructions.
17181 AVX512F instructions.
17184 Here is an example:
17186 if (__builtin_cpu_supports ("popcnt"))
17188 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
17192 count = generic_countbits (n); //generic implementation.
17198 The following built-in functions are made available by @option{-mmmx}.
17199 All of them generate the machine instruction that is part of the name.
17202 v8qi __builtin_ia32_paddb (v8qi, v8qi)
17203 v4hi __builtin_ia32_paddw (v4hi, v4hi)
17204 v2si __builtin_ia32_paddd (v2si, v2si)
17205 v8qi __builtin_ia32_psubb (v8qi, v8qi)
17206 v4hi __builtin_ia32_psubw (v4hi, v4hi)
17207 v2si __builtin_ia32_psubd (v2si, v2si)
17208 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17209 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17210 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17211 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17212 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17213 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17214 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17215 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17216 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17217 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17218 di __builtin_ia32_pand (di, di)
17219 di __builtin_ia32_pandn (di,di)
17220 di __builtin_ia32_por (di, di)
17221 di __builtin_ia32_pxor (di, di)
17222 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17223 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17224 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17225 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17226 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17227 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17228 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17229 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17230 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17231 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17232 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17233 v2si __builtin_ia32_punpckldq (v2si, v2si)
17234 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17235 v4hi __builtin_ia32_packssdw (v2si, v2si)
17236 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17238 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17239 v2si __builtin_ia32_pslld (v2si, v2si)
17240 v1di __builtin_ia32_psllq (v1di, v1di)
17241 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17242 v2si __builtin_ia32_psrld (v2si, v2si)
17243 v1di __builtin_ia32_psrlq (v1di, v1di)
17244 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17245 v2si __builtin_ia32_psrad (v2si, v2si)
17246 v4hi __builtin_ia32_psllwi (v4hi, int)
17247 v2si __builtin_ia32_pslldi (v2si, int)
17248 v1di __builtin_ia32_psllqi (v1di, int)
17249 v4hi __builtin_ia32_psrlwi (v4hi, int)
17250 v2si __builtin_ia32_psrldi (v2si, int)
17251 v1di __builtin_ia32_psrlqi (v1di, int)
17252 v4hi __builtin_ia32_psrawi (v4hi, int)
17253 v2si __builtin_ia32_psradi (v2si, int)
17257 The following built-in functions are made available either with
17258 @option{-msse}, or with a combination of @option{-m3dnow} and
17259 @option{-march=athlon}. All of them generate the machine
17260 instruction that is part of the name.
17263 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17264 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17265 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17266 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17267 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17268 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17269 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17270 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17271 int __builtin_ia32_pmovmskb (v8qi)
17272 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17273 void __builtin_ia32_movntq (di *, di)
17274 void __builtin_ia32_sfence (void)
17277 The following built-in functions are available when @option{-msse} is used.
17278 All of them generate the machine instruction that is part of the name.
17281 int __builtin_ia32_comieq (v4sf, v4sf)
17282 int __builtin_ia32_comineq (v4sf, v4sf)
17283 int __builtin_ia32_comilt (v4sf, v4sf)
17284 int __builtin_ia32_comile (v4sf, v4sf)
17285 int __builtin_ia32_comigt (v4sf, v4sf)
17286 int __builtin_ia32_comige (v4sf, v4sf)
17287 int __builtin_ia32_ucomieq (v4sf, v4sf)
17288 int __builtin_ia32_ucomineq (v4sf, v4sf)
17289 int __builtin_ia32_ucomilt (v4sf, v4sf)
17290 int __builtin_ia32_ucomile (v4sf, v4sf)
17291 int __builtin_ia32_ucomigt (v4sf, v4sf)
17292 int __builtin_ia32_ucomige (v4sf, v4sf)
17293 v4sf __builtin_ia32_addps (v4sf, v4sf)
17294 v4sf __builtin_ia32_subps (v4sf, v4sf)
17295 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17296 v4sf __builtin_ia32_divps (v4sf, v4sf)
17297 v4sf __builtin_ia32_addss (v4sf, v4sf)
17298 v4sf __builtin_ia32_subss (v4sf, v4sf)
17299 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17300 v4sf __builtin_ia32_divss (v4sf, v4sf)
17301 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17302 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17303 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17304 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17305 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17306 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17307 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17308 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17309 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17310 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17311 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17312 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17313 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17314 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17315 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17316 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17317 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17318 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17319 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17320 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17321 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17322 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17323 v4sf __builtin_ia32_minps (v4sf, v4sf)
17324 v4sf __builtin_ia32_minss (v4sf, v4sf)
17325 v4sf __builtin_ia32_andps (v4sf, v4sf)
17326 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17327 v4sf __builtin_ia32_orps (v4sf, v4sf)
17328 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17329 v4sf __builtin_ia32_movss (v4sf, v4sf)
17330 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17331 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17332 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17333 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17334 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17335 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17336 v2si __builtin_ia32_cvtps2pi (v4sf)
17337 int __builtin_ia32_cvtss2si (v4sf)
17338 v2si __builtin_ia32_cvttps2pi (v4sf)
17339 int __builtin_ia32_cvttss2si (v4sf)
17340 v4sf __builtin_ia32_rcpps (v4sf)
17341 v4sf __builtin_ia32_rsqrtps (v4sf)
17342 v4sf __builtin_ia32_sqrtps (v4sf)
17343 v4sf __builtin_ia32_rcpss (v4sf)
17344 v4sf __builtin_ia32_rsqrtss (v4sf)
17345 v4sf __builtin_ia32_sqrtss (v4sf)
17346 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17347 void __builtin_ia32_movntps (float *, v4sf)
17348 int __builtin_ia32_movmskps (v4sf)
17351 The following built-in functions are available when @option{-msse} is used.
17354 @item v4sf __builtin_ia32_loadups (float *)
17355 Generates the @code{movups} machine instruction as a load from memory.
17356 @item void __builtin_ia32_storeups (float *, v4sf)
17357 Generates the @code{movups} machine instruction as a store to memory.
17358 @item v4sf __builtin_ia32_loadss (float *)
17359 Generates the @code{movss} machine instruction as a load from memory.
17360 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17361 Generates the @code{movhps} machine instruction as a load from memory.
17362 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17363 Generates the @code{movlps} machine instruction as a load from memory
17364 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17365 Generates the @code{movhps} machine instruction as a store to memory.
17366 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17367 Generates the @code{movlps} machine instruction as a store to memory.
17370 The following built-in functions are available when @option{-msse2} is used.
17371 All of them generate the machine instruction that is part of the name.
17374 int __builtin_ia32_comisdeq (v2df, v2df)
17375 int __builtin_ia32_comisdlt (v2df, v2df)
17376 int __builtin_ia32_comisdle (v2df, v2df)
17377 int __builtin_ia32_comisdgt (v2df, v2df)
17378 int __builtin_ia32_comisdge (v2df, v2df)
17379 int __builtin_ia32_comisdneq (v2df, v2df)
17380 int __builtin_ia32_ucomisdeq (v2df, v2df)
17381 int __builtin_ia32_ucomisdlt (v2df, v2df)
17382 int __builtin_ia32_ucomisdle (v2df, v2df)
17383 int __builtin_ia32_ucomisdgt (v2df, v2df)
17384 int __builtin_ia32_ucomisdge (v2df, v2df)
17385 int __builtin_ia32_ucomisdneq (v2df, v2df)
17386 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17387 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17388 v2df __builtin_ia32_cmplepd (v2df, v2df)
17389 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17390 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17391 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17392 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17393 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17394 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17395 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17396 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17397 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17398 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17399 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17400 v2df __builtin_ia32_cmplesd (v2df, v2df)
17401 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17402 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17403 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17404 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17405 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17406 v2di __builtin_ia32_paddq (v2di, v2di)
17407 v2di __builtin_ia32_psubq (v2di, v2di)
17408 v2df __builtin_ia32_addpd (v2df, v2df)
17409 v2df __builtin_ia32_subpd (v2df, v2df)
17410 v2df __builtin_ia32_mulpd (v2df, v2df)
17411 v2df __builtin_ia32_divpd (v2df, v2df)
17412 v2df __builtin_ia32_addsd (v2df, v2df)
17413 v2df __builtin_ia32_subsd (v2df, v2df)
17414 v2df __builtin_ia32_mulsd (v2df, v2df)
17415 v2df __builtin_ia32_divsd (v2df, v2df)
17416 v2df __builtin_ia32_minpd (v2df, v2df)
17417 v2df __builtin_ia32_maxpd (v2df, v2df)
17418 v2df __builtin_ia32_minsd (v2df, v2df)
17419 v2df __builtin_ia32_maxsd (v2df, v2df)
17420 v2df __builtin_ia32_andpd (v2df, v2df)
17421 v2df __builtin_ia32_andnpd (v2df, v2df)
17422 v2df __builtin_ia32_orpd (v2df, v2df)
17423 v2df __builtin_ia32_xorpd (v2df, v2df)
17424 v2df __builtin_ia32_movsd (v2df, v2df)
17425 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17426 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17427 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17428 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17429 v4si __builtin_ia32_paddd128 (v4si, v4si)
17430 v2di __builtin_ia32_paddq128 (v2di, v2di)
17431 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17432 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17433 v4si __builtin_ia32_psubd128 (v4si, v4si)
17434 v2di __builtin_ia32_psubq128 (v2di, v2di)
17435 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17436 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17437 v2di __builtin_ia32_pand128 (v2di, v2di)
17438 v2di __builtin_ia32_pandn128 (v2di, v2di)
17439 v2di __builtin_ia32_por128 (v2di, v2di)
17440 v2di __builtin_ia32_pxor128 (v2di, v2di)
17441 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17442 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17443 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17444 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17445 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17446 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17447 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17448 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17449 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17450 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17451 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17452 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17453 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17454 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17455 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17456 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17457 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17458 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17459 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17460 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17461 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17462 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17463 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17464 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17465 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17466 v2df __builtin_ia32_loadupd (double *)
17467 void __builtin_ia32_storeupd (double *, v2df)
17468 v2df __builtin_ia32_loadhpd (v2df, double const *)
17469 v2df __builtin_ia32_loadlpd (v2df, double const *)
17470 int __builtin_ia32_movmskpd (v2df)
17471 int __builtin_ia32_pmovmskb128 (v16qi)
17472 void __builtin_ia32_movnti (int *, int)
17473 void __builtin_ia32_movnti64 (long long int *, long long int)
17474 void __builtin_ia32_movntpd (double *, v2df)
17475 void __builtin_ia32_movntdq (v2df *, v2df)
17476 v4si __builtin_ia32_pshufd (v4si, int)
17477 v8hi __builtin_ia32_pshuflw (v8hi, int)
17478 v8hi __builtin_ia32_pshufhw (v8hi, int)
17479 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17480 v2df __builtin_ia32_sqrtpd (v2df)
17481 v2df __builtin_ia32_sqrtsd (v2df)
17482 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17483 v2df __builtin_ia32_cvtdq2pd (v4si)
17484 v4sf __builtin_ia32_cvtdq2ps (v4si)
17485 v4si __builtin_ia32_cvtpd2dq (v2df)
17486 v2si __builtin_ia32_cvtpd2pi (v2df)
17487 v4sf __builtin_ia32_cvtpd2ps (v2df)
17488 v4si __builtin_ia32_cvttpd2dq (v2df)
17489 v2si __builtin_ia32_cvttpd2pi (v2df)
17490 v2df __builtin_ia32_cvtpi2pd (v2si)
17491 int __builtin_ia32_cvtsd2si (v2df)
17492 int __builtin_ia32_cvttsd2si (v2df)
17493 long long __builtin_ia32_cvtsd2si64 (v2df)
17494 long long __builtin_ia32_cvttsd2si64 (v2df)
17495 v4si __builtin_ia32_cvtps2dq (v4sf)
17496 v2df __builtin_ia32_cvtps2pd (v4sf)
17497 v4si __builtin_ia32_cvttps2dq (v4sf)
17498 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17499 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17500 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17501 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17502 void __builtin_ia32_clflush (const void *)
17503 void __builtin_ia32_lfence (void)
17504 void __builtin_ia32_mfence (void)
17505 v16qi __builtin_ia32_loaddqu (const char *)
17506 void __builtin_ia32_storedqu (char *, v16qi)
17507 v1di __builtin_ia32_pmuludq (v2si, v2si)
17508 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17509 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17510 v4si __builtin_ia32_pslld128 (v4si, v4si)
17511 v2di __builtin_ia32_psllq128 (v2di, v2di)
17512 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17513 v4si __builtin_ia32_psrld128 (v4si, v4si)
17514 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17515 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17516 v4si __builtin_ia32_psrad128 (v4si, v4si)
17517 v2di __builtin_ia32_pslldqi128 (v2di, int)
17518 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17519 v4si __builtin_ia32_pslldi128 (v4si, int)
17520 v2di __builtin_ia32_psllqi128 (v2di, int)
17521 v2di __builtin_ia32_psrldqi128 (v2di, int)
17522 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17523 v4si __builtin_ia32_psrldi128 (v4si, int)
17524 v2di __builtin_ia32_psrlqi128 (v2di, int)
17525 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17526 v4si __builtin_ia32_psradi128 (v4si, int)
17527 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17528 v2di __builtin_ia32_movq128 (v2di)
17531 The following built-in functions are available when @option{-msse3} is used.
17532 All of them generate the machine instruction that is part of the name.
17535 v2df __builtin_ia32_addsubpd (v2df, v2df)
17536 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17537 v2df __builtin_ia32_haddpd (v2df, v2df)
17538 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17539 v2df __builtin_ia32_hsubpd (v2df, v2df)
17540 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17541 v16qi __builtin_ia32_lddqu (char const *)
17542 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17543 v4sf __builtin_ia32_movshdup (v4sf)
17544 v4sf __builtin_ia32_movsldup (v4sf)
17545 void __builtin_ia32_mwait (unsigned int, unsigned int)
17548 The following built-in functions are available when @option{-mssse3} is used.
17549 All of them generate the machine instruction that is part of the name.
17552 v2si __builtin_ia32_phaddd (v2si, v2si)
17553 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17554 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17555 v2si __builtin_ia32_phsubd (v2si, v2si)
17556 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17557 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17558 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17559 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17560 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17561 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17562 v2si __builtin_ia32_psignd (v2si, v2si)
17563 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17564 v1di __builtin_ia32_palignr (v1di, v1di, int)
17565 v8qi __builtin_ia32_pabsb (v8qi)
17566 v2si __builtin_ia32_pabsd (v2si)
17567 v4hi __builtin_ia32_pabsw (v4hi)
17570 The following built-in functions are available when @option{-mssse3} is used.
17571 All of them generate the machine instruction that is part of the name.
17574 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17575 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17576 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17577 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17578 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17579 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17580 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17581 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17582 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17583 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17584 v4si __builtin_ia32_psignd128 (v4si, v4si)
17585 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17586 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17587 v16qi __builtin_ia32_pabsb128 (v16qi)
17588 v4si __builtin_ia32_pabsd128 (v4si)
17589 v8hi __builtin_ia32_pabsw128 (v8hi)
17592 The following built-in functions are available when @option{-msse4.1} is
17593 used. All of them generate the machine instruction that is part of the
17597 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17598 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17599 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17600 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17601 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17602 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17603 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17604 v2di __builtin_ia32_movntdqa (v2di *);
17605 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17606 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17607 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17608 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17609 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17610 v8hi __builtin_ia32_phminposuw128 (v8hi)
17611 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17612 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17613 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17614 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17615 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17616 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17617 v4si __builtin_ia32_pminud128 (v4si, v4si)
17618 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17619 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17620 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17621 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17622 v2di __builtin_ia32_pmovsxdq128 (v4si)
17623 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17624 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17625 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17626 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17627 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17628 v2di __builtin_ia32_pmovzxdq128 (v4si)
17629 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17630 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17631 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17632 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17633 int __builtin_ia32_ptestc128 (v2di, v2di)
17634 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17635 int __builtin_ia32_ptestz128 (v2di, v2di)
17636 v2df __builtin_ia32_roundpd (v2df, const int)
17637 v4sf __builtin_ia32_roundps (v4sf, const int)
17638 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17639 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17642 The following built-in functions are available when @option{-msse4.1} is
17646 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17647 Generates the @code{insertps} machine instruction.
17648 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17649 Generates the @code{pextrb} machine instruction.
17650 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17651 Generates the @code{pinsrb} machine instruction.
17652 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17653 Generates the @code{pinsrd} machine instruction.
17654 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17655 Generates the @code{pinsrq} machine instruction in 64bit mode.
17658 The following built-in functions are changed to generate new SSE4.1
17659 instructions when @option{-msse4.1} is used.
17662 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17663 Generates the @code{extractps} machine instruction.
17664 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17665 Generates the @code{pextrd} machine instruction.
17666 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17667 Generates the @code{pextrq} machine instruction in 64bit mode.
17670 The following built-in functions are available when @option{-msse4.2} is
17671 used. All of them generate the machine instruction that is part of the
17675 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17676 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17677 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17678 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17679 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17680 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17681 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17682 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17683 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17684 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17685 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17686 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17687 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17688 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17689 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17692 The following built-in functions are available when @option{-msse4.2} is
17696 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17697 Generates the @code{crc32b} machine instruction.
17698 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17699 Generates the @code{crc32w} machine instruction.
17700 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17701 Generates the @code{crc32l} machine instruction.
17702 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17703 Generates the @code{crc32q} machine instruction.
17706 The following built-in functions are changed to generate new SSE4.2
17707 instructions when @option{-msse4.2} is used.
17710 @item int __builtin_popcount (unsigned int)
17711 Generates the @code{popcntl} machine instruction.
17712 @item int __builtin_popcountl (unsigned long)
17713 Generates the @code{popcntl} or @code{popcntq} machine instruction,
17714 depending on the size of @code{unsigned long}.
17715 @item int __builtin_popcountll (unsigned long long)
17716 Generates the @code{popcntq} machine instruction.
17719 The following built-in functions are available when @option{-mavx} is
17720 used. All of them generate the machine instruction that is part of the
17724 v4df __builtin_ia32_addpd256 (v4df,v4df)
17725 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
17726 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
17727 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
17728 v4df __builtin_ia32_andnpd256 (v4df,v4df)
17729 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
17730 v4df __builtin_ia32_andpd256 (v4df,v4df)
17731 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
17732 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
17733 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
17734 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
17735 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
17736 v2df __builtin_ia32_cmppd (v2df,v2df,int)
17737 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
17738 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
17739 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
17740 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
17741 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
17742 v4df __builtin_ia32_cvtdq2pd256 (v4si)
17743 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
17744 v4si __builtin_ia32_cvtpd2dq256 (v4df)
17745 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
17746 v8si __builtin_ia32_cvtps2dq256 (v8sf)
17747 v4df __builtin_ia32_cvtps2pd256 (v4sf)
17748 v4si __builtin_ia32_cvttpd2dq256 (v4df)
17749 v8si __builtin_ia32_cvttps2dq256 (v8sf)
17750 v4df __builtin_ia32_divpd256 (v4df,v4df)
17751 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
17752 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
17753 v4df __builtin_ia32_haddpd256 (v4df,v4df)
17754 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
17755 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
17756 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
17757 v32qi __builtin_ia32_lddqu256 (pcchar)
17758 v32qi __builtin_ia32_loaddqu256 (pcchar)
17759 v4df __builtin_ia32_loadupd256 (pcdouble)
17760 v8sf __builtin_ia32_loadups256 (pcfloat)
17761 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
17762 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
17763 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
17764 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
17765 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
17766 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
17767 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
17768 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
17769 v4df __builtin_ia32_maxpd256 (v4df,v4df)
17770 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
17771 v4df __builtin_ia32_minpd256 (v4df,v4df)
17772 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
17773 v4df __builtin_ia32_movddup256 (v4df)
17774 int __builtin_ia32_movmskpd256 (v4df)
17775 int __builtin_ia32_movmskps256 (v8sf)
17776 v8sf __builtin_ia32_movshdup256 (v8sf)
17777 v8sf __builtin_ia32_movsldup256 (v8sf)
17778 v4df __builtin_ia32_mulpd256 (v4df,v4df)
17779 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
17780 v4df __builtin_ia32_orpd256 (v4df,v4df)
17781 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
17782 v2df __builtin_ia32_pd_pd256 (v4df)
17783 v4df __builtin_ia32_pd256_pd (v2df)
17784 v4sf __builtin_ia32_ps_ps256 (v8sf)
17785 v8sf __builtin_ia32_ps256_ps (v4sf)
17786 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
17787 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
17788 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
17789 v8sf __builtin_ia32_rcpps256 (v8sf)
17790 v4df __builtin_ia32_roundpd256 (v4df,int)
17791 v8sf __builtin_ia32_roundps256 (v8sf,int)
17792 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
17793 v8sf __builtin_ia32_rsqrtps256 (v8sf)
17794 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
17795 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
17796 v4si __builtin_ia32_si_si256 (v8si)
17797 v8si __builtin_ia32_si256_si (v4si)
17798 v4df __builtin_ia32_sqrtpd256 (v4df)
17799 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
17800 v8sf __builtin_ia32_sqrtps256 (v8sf)
17801 void __builtin_ia32_storedqu256 (pchar,v32qi)
17802 void __builtin_ia32_storeupd256 (pdouble,v4df)
17803 void __builtin_ia32_storeups256 (pfloat,v8sf)
17804 v4df __builtin_ia32_subpd256 (v4df,v4df)
17805 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
17806 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
17807 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
17808 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
17809 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
17810 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
17811 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
17812 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
17813 v4sf __builtin_ia32_vbroadcastss (pcfloat)
17814 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
17815 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
17816 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
17817 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
17818 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
17819 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
17820 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
17821 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
17822 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
17823 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
17824 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
17825 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
17826 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
17827 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
17828 v2df __builtin_ia32_vpermilpd (v2df,int)
17829 v4df __builtin_ia32_vpermilpd256 (v4df,int)
17830 v4sf __builtin_ia32_vpermilps (v4sf,int)
17831 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
17832 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
17833 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
17834 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
17835 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
17836 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
17837 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
17838 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
17839 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
17840 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
17841 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
17842 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
17843 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
17844 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
17845 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
17846 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
17847 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
17848 void __builtin_ia32_vzeroall (void)
17849 void __builtin_ia32_vzeroupper (void)
17850 v4df __builtin_ia32_xorpd256 (v4df,v4df)
17851 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
17854 The following built-in functions are available when @option{-mavx2} is
17855 used. All of them generate the machine instruction that is part of the
17859 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
17860 v32qi __builtin_ia32_pabsb256 (v32qi)
17861 v16hi __builtin_ia32_pabsw256 (v16hi)
17862 v8si __builtin_ia32_pabsd256 (v8si)
17863 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
17864 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
17865 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
17866 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
17867 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
17868 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
17869 v8si __builtin_ia32_paddd256 (v8si,v8si)
17870 v4di __builtin_ia32_paddq256 (v4di,v4di)
17871 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
17872 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
17873 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
17874 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
17875 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
17876 v4di __builtin_ia32_andsi256 (v4di,v4di)
17877 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
17878 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
17879 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
17880 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
17881 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
17882 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
17883 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
17884 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
17885 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
17886 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
17887 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
17888 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
17889 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
17890 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
17891 v8si __builtin_ia32_phaddd256 (v8si,v8si)
17892 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
17893 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
17894 v8si __builtin_ia32_phsubd256 (v8si,v8si)
17895 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
17896 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
17897 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
17898 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
17899 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
17900 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
17901 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
17902 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
17903 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
17904 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
17905 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
17906 v8si __builtin_ia32_pminsd256 (v8si,v8si)
17907 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
17908 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
17909 v8si __builtin_ia32_pminud256 (v8si,v8si)
17910 int __builtin_ia32_pmovmskb256 (v32qi)
17911 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
17912 v8si __builtin_ia32_pmovsxbd256 (v16qi)
17913 v4di __builtin_ia32_pmovsxbq256 (v16qi)
17914 v8si __builtin_ia32_pmovsxwd256 (v8hi)
17915 v4di __builtin_ia32_pmovsxwq256 (v8hi)
17916 v4di __builtin_ia32_pmovsxdq256 (v4si)
17917 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
17918 v8si __builtin_ia32_pmovzxbd256 (v16qi)
17919 v4di __builtin_ia32_pmovzxbq256 (v16qi)
17920 v8si __builtin_ia32_pmovzxwd256 (v8hi)
17921 v4di __builtin_ia32_pmovzxwq256 (v8hi)
17922 v4di __builtin_ia32_pmovzxdq256 (v4si)
17923 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
17924 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
17925 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
17926 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
17927 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
17928 v8si __builtin_ia32_pmulld256 (v8si,v8si)
17929 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
17930 v4di __builtin_ia32_por256 (v4di,v4di)
17931 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
17932 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
17933 v8si __builtin_ia32_pshufd256 (v8si,int)
17934 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
17935 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
17936 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
17937 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
17938 v8si __builtin_ia32_psignd256 (v8si,v8si)
17939 v4di __builtin_ia32_pslldqi256 (v4di,int)
17940 v16hi __builtin_ia32_psllwi256 (16hi,int)
17941 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
17942 v8si __builtin_ia32_pslldi256 (v8si,int)
17943 v8si __builtin_ia32_pslld256(v8si,v4si)
17944 v4di __builtin_ia32_psllqi256 (v4di,int)
17945 v4di __builtin_ia32_psllq256(v4di,v2di)
17946 v16hi __builtin_ia32_psrawi256 (v16hi,int)
17947 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
17948 v8si __builtin_ia32_psradi256 (v8si,int)
17949 v8si __builtin_ia32_psrad256 (v8si,v4si)
17950 v4di __builtin_ia32_psrldqi256 (v4di, int)
17951 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
17952 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
17953 v8si __builtin_ia32_psrldi256 (v8si,int)
17954 v8si __builtin_ia32_psrld256 (v8si,v4si)
17955 v4di __builtin_ia32_psrlqi256 (v4di,int)
17956 v4di __builtin_ia32_psrlq256(v4di,v2di)
17957 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
17958 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
17959 v8si __builtin_ia32_psubd256 (v8si,v8si)
17960 v4di __builtin_ia32_psubq256 (v4di,v4di)
17961 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
17962 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
17963 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
17964 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
17965 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
17966 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
17967 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
17968 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
17969 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
17970 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
17971 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
17972 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
17973 v4di __builtin_ia32_pxor256 (v4di,v4di)
17974 v4di __builtin_ia32_movntdqa256 (pv4di)
17975 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
17976 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
17977 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
17978 v4di __builtin_ia32_vbroadcastsi256 (v2di)
17979 v4si __builtin_ia32_pblendd128 (v4si,v4si)
17980 v8si __builtin_ia32_pblendd256 (v8si,v8si)
17981 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
17982 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
17983 v8si __builtin_ia32_pbroadcastd256 (v4si)
17984 v4di __builtin_ia32_pbroadcastq256 (v2di)
17985 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
17986 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
17987 v4si __builtin_ia32_pbroadcastd128 (v4si)
17988 v2di __builtin_ia32_pbroadcastq128 (v2di)
17989 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
17990 v4df __builtin_ia32_permdf256 (v4df,int)
17991 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
17992 v4di __builtin_ia32_permdi256 (v4di,int)
17993 v4di __builtin_ia32_permti256 (v4di,v4di,int)
17994 v4di __builtin_ia32_extract128i256 (v4di,int)
17995 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
17996 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
17997 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
17998 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
17999 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
18000 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
18001 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
18002 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
18003 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
18004 v8si __builtin_ia32_psllv8si (v8si,v8si)
18005 v4si __builtin_ia32_psllv4si (v4si,v4si)
18006 v4di __builtin_ia32_psllv4di (v4di,v4di)
18007 v2di __builtin_ia32_psllv2di (v2di,v2di)
18008 v8si __builtin_ia32_psrav8si (v8si,v8si)
18009 v4si __builtin_ia32_psrav4si (v4si,v4si)
18010 v8si __builtin_ia32_psrlv8si (v8si,v8si)
18011 v4si __builtin_ia32_psrlv4si (v4si,v4si)
18012 v4di __builtin_ia32_psrlv4di (v4di,v4di)
18013 v2di __builtin_ia32_psrlv2di (v2di,v2di)
18014 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
18015 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
18016 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
18017 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
18018 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
18019 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
18020 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
18021 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
18022 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
18023 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
18024 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
18025 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
18026 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
18027 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
18028 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
18029 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
18032 The following built-in functions are available when @option{-maes} is
18033 used. All of them generate the machine instruction that is part of the
18037 v2di __builtin_ia32_aesenc128 (v2di, v2di)
18038 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
18039 v2di __builtin_ia32_aesdec128 (v2di, v2di)
18040 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
18041 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
18042 v2di __builtin_ia32_aesimc128 (v2di)
18045 The following built-in function is available when @option{-mpclmul} is
18049 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
18050 Generates the @code{pclmulqdq} machine instruction.
18053 The following built-in function is available when @option{-mfsgsbase} is
18054 used. All of them generate the machine instruction that is part of the
18058 unsigned int __builtin_ia32_rdfsbase32 (void)
18059 unsigned long long __builtin_ia32_rdfsbase64 (void)
18060 unsigned int __builtin_ia32_rdgsbase32 (void)
18061 unsigned long long __builtin_ia32_rdgsbase64 (void)
18062 void _writefsbase_u32 (unsigned int)
18063 void _writefsbase_u64 (unsigned long long)
18064 void _writegsbase_u32 (unsigned int)
18065 void _writegsbase_u64 (unsigned long long)
18068 The following built-in function is available when @option{-mrdrnd} is
18069 used. All of them generate the machine instruction that is part of the
18073 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
18074 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
18075 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
18078 The following built-in functions are available when @option{-msse4a} is used.
18079 All of them generate the machine instruction that is part of the name.
18082 void __builtin_ia32_movntsd (double *, v2df)
18083 void __builtin_ia32_movntss (float *, v4sf)
18084 v2di __builtin_ia32_extrq (v2di, v16qi)
18085 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
18086 v2di __builtin_ia32_insertq (v2di, v2di)
18087 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
18090 The following built-in functions are available when @option{-mxop} is used.
18092 v2df __builtin_ia32_vfrczpd (v2df)
18093 v4sf __builtin_ia32_vfrczps (v4sf)
18094 v2df __builtin_ia32_vfrczsd (v2df)
18095 v4sf __builtin_ia32_vfrczss (v4sf)
18096 v4df __builtin_ia32_vfrczpd256 (v4df)
18097 v8sf __builtin_ia32_vfrczps256 (v8sf)
18098 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
18099 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
18100 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
18101 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
18102 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
18103 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
18104 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
18105 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
18106 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
18107 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
18108 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
18109 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
18110 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
18111 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
18112 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18113 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
18114 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
18115 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
18116 v4si __builtin_ia32_vpcomequd (v4si, v4si)
18117 v2di __builtin_ia32_vpcomequq (v2di, v2di)
18118 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
18119 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18120 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
18121 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
18122 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
18123 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
18124 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
18125 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
18126 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
18127 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
18128 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
18129 v4si __builtin_ia32_vpcomged (v4si, v4si)
18130 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
18131 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
18132 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
18133 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
18134 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
18135 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
18136 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
18137 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
18138 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
18139 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
18140 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
18141 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
18142 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
18143 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
18144 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
18145 v4si __builtin_ia32_vpcomled (v4si, v4si)
18146 v2di __builtin_ia32_vpcomleq (v2di, v2di)
18147 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
18148 v4si __builtin_ia32_vpcomleud (v4si, v4si)
18149 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
18150 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
18151 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
18152 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
18153 v4si __builtin_ia32_vpcomltd (v4si, v4si)
18154 v2di __builtin_ia32_vpcomltq (v2di, v2di)
18155 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
18156 v4si __builtin_ia32_vpcomltud (v4si, v4si)
18157 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
18158 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
18159 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
18160 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
18161 v4si __builtin_ia32_vpcomned (v4si, v4si)
18162 v2di __builtin_ia32_vpcomneq (v2di, v2di)
18163 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
18164 v4si __builtin_ia32_vpcomneud (v4si, v4si)
18165 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
18166 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
18167 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
18168 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
18169 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
18170 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
18171 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
18172 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
18173 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
18174 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
18175 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
18176 v4si __builtin_ia32_vphaddbd (v16qi)
18177 v2di __builtin_ia32_vphaddbq (v16qi)
18178 v8hi __builtin_ia32_vphaddbw (v16qi)
18179 v2di __builtin_ia32_vphadddq (v4si)
18180 v4si __builtin_ia32_vphaddubd (v16qi)
18181 v2di __builtin_ia32_vphaddubq (v16qi)
18182 v8hi __builtin_ia32_vphaddubw (v16qi)
18183 v2di __builtin_ia32_vphaddudq (v4si)
18184 v4si __builtin_ia32_vphadduwd (v8hi)
18185 v2di __builtin_ia32_vphadduwq (v8hi)
18186 v4si __builtin_ia32_vphaddwd (v8hi)
18187 v2di __builtin_ia32_vphaddwq (v8hi)
18188 v8hi __builtin_ia32_vphsubbw (v16qi)
18189 v2di __builtin_ia32_vphsubdq (v4si)
18190 v4si __builtin_ia32_vphsubwd (v8hi)
18191 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
18192 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
18193 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
18194 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
18195 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
18196 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
18197 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
18198 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
18199 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
18200 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
18201 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
18202 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
18203 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
18204 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
18205 v4si __builtin_ia32_vprotd (v4si, v4si)
18206 v2di __builtin_ia32_vprotq (v2di, v2di)
18207 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18208 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18209 v4si __builtin_ia32_vpshad (v4si, v4si)
18210 v2di __builtin_ia32_vpshaq (v2di, v2di)
18211 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18212 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18213 v4si __builtin_ia32_vpshld (v4si, v4si)
18214 v2di __builtin_ia32_vpshlq (v2di, v2di)
18215 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18218 The following built-in functions are available when @option{-mfma4} is used.
18219 All of them generate the machine instruction that is part of the name.
18222 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18223 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18224 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18225 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18226 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18227 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18228 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18229 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18230 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18231 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18232 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18233 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18234 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18235 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18236 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18237 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18238 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18239 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18240 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18241 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18242 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18243 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18244 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18245 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18246 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18247 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18248 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18249 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18250 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18251 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18252 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18253 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18257 The following built-in functions are available when @option{-mlwp} is used.
18260 void __builtin_ia32_llwpcb16 (void *);
18261 void __builtin_ia32_llwpcb32 (void *);
18262 void __builtin_ia32_llwpcb64 (void *);
18263 void * __builtin_ia32_llwpcb16 (void);
18264 void * __builtin_ia32_llwpcb32 (void);
18265 void * __builtin_ia32_llwpcb64 (void);
18266 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18267 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18268 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18269 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18270 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18271 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18274 The following built-in functions are available when @option{-mbmi} is used.
18275 All of them generate the machine instruction that is part of the name.
18277 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18278 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18281 The following built-in functions are available when @option{-mbmi2} is used.
18282 All of them generate the machine instruction that is part of the name.
18284 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18285 unsigned int _pdep_u32 (unsigned int, unsigned int)
18286 unsigned int _pext_u32 (unsigned int, unsigned int)
18287 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18288 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18289 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18292 The following built-in functions are available when @option{-mlzcnt} is used.
18293 All of them generate the machine instruction that is part of the name.
18295 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18296 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18297 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18300 The following built-in functions are available when @option{-mfxsr} is used.
18301 All of them generate the machine instruction that is part of the name.
18303 void __builtin_ia32_fxsave (void *)
18304 void __builtin_ia32_fxrstor (void *)
18305 void __builtin_ia32_fxsave64 (void *)
18306 void __builtin_ia32_fxrstor64 (void *)
18309 The following built-in functions are available when @option{-mxsave} is used.
18310 All of them generate the machine instruction that is part of the name.
18312 void __builtin_ia32_xsave (void *, long long)
18313 void __builtin_ia32_xrstor (void *, long long)
18314 void __builtin_ia32_xsave64 (void *, long long)
18315 void __builtin_ia32_xrstor64 (void *, long long)
18318 The following built-in functions are available when @option{-mxsaveopt} is used.
18319 All of them generate the machine instruction that is part of the name.
18321 void __builtin_ia32_xsaveopt (void *, long long)
18322 void __builtin_ia32_xsaveopt64 (void *, long long)
18325 The following built-in functions are available when @option{-mtbm} is used.
18326 Both of them generate the immediate form of the bextr machine instruction.
18328 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18329 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18333 The following built-in functions are available when @option{-m3dnow} is used.
18334 All of them generate the machine instruction that is part of the name.
18337 void __builtin_ia32_femms (void)
18338 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18339 v2si __builtin_ia32_pf2id (v2sf)
18340 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18341 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18342 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18343 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18344 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18345 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18346 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18347 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18348 v2sf __builtin_ia32_pfrcp (v2sf)
18349 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18350 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18351 v2sf __builtin_ia32_pfrsqrt (v2sf)
18352 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18353 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18354 v2sf __builtin_ia32_pi2fd (v2si)
18355 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18358 The following built-in functions are available when both @option{-m3dnow}
18359 and @option{-march=athlon} are used. All of them generate the machine
18360 instruction that is part of the name.
18363 v2si __builtin_ia32_pf2iw (v2sf)
18364 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18365 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18366 v2sf __builtin_ia32_pi2fw (v2si)
18367 v2sf __builtin_ia32_pswapdsf (v2sf)
18368 v2si __builtin_ia32_pswapdsi (v2si)
18371 The following built-in functions are available when @option{-mrtm} is used
18372 They are used for restricted transactional memory. These are the internal
18373 low level functions. Normally the functions in
18374 @ref{x86 transactional memory intrinsics} should be used instead.
18377 int __builtin_ia32_xbegin ()
18378 void __builtin_ia32_xend ()
18379 void __builtin_ia32_xabort (status)
18380 int __builtin_ia32_xtest ()
18383 The following built-in functions are available when @option{-mmwaitx} is used.
18384 All of them generate the machine instruction that is part of the name.
18386 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18387 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18390 @node x86 transactional memory intrinsics
18391 @subsection x86 Transactional Memory Intrinsics
18393 These hardware transactional memory intrinsics for x86 allow you to use
18394 memory transactions with RTM (Restricted Transactional Memory).
18395 This support is enabled with the @option{-mrtm} option.
18396 For using HLE (Hardware Lock Elision) see
18397 @ref{x86 specific memory model extensions for transactional memory} instead.
18399 A memory transaction commits all changes to memory in an atomic way,
18400 as visible to other threads. If the transaction fails it is rolled back
18401 and all side effects discarded.
18403 Generally there is no guarantee that a memory transaction ever succeeds
18404 and suitable fallback code always needs to be supplied.
18406 @deftypefn {RTM Function} {unsigned} _xbegin ()
18407 Start a RTM (Restricted Transactional Memory) transaction.
18408 Returns @code{_XBEGIN_STARTED} when the transaction
18409 started successfully (note this is not 0, so the constant has to be
18410 explicitly tested).
18412 If the transaction aborts, all side-effects
18413 are undone and an abort code encoded as a bit mask is returned.
18414 The following macros are defined:
18417 @item _XABORT_EXPLICIT
18418 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18419 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18420 @item _XABORT_RETRY
18421 Transaction retry is possible.
18422 @item _XABORT_CONFLICT
18423 Transaction abort due to a memory conflict with another thread.
18424 @item _XABORT_CAPACITY
18425 Transaction abort due to the transaction using too much memory.
18426 @item _XABORT_DEBUG
18427 Transaction abort due to a debug trap.
18428 @item _XABORT_NESTED
18429 Transaction abort in an inner nested transaction.
18432 There is no guarantee
18433 any transaction ever succeeds, so there always needs to be a valid
18437 @deftypefn {RTM Function} {void} _xend ()
18438 Commit the current transaction. When no transaction is active this faults.
18439 All memory side-effects of the transaction become visible
18440 to other threads in an atomic manner.
18443 @deftypefn {RTM Function} {int} _xtest ()
18444 Return a nonzero value if a transaction is currently active, otherwise 0.
18447 @deftypefn {RTM Function} {void} _xabort (status)
18448 Abort the current transaction. When no transaction is active this is a no-op.
18449 The @var{status} is an 8-bit constant; its value is encoded in the return
18450 value from @code{_xbegin}.
18453 Here is an example showing handling for @code{_XABORT_RETRY}
18454 and a fallback path for other failures:
18457 #include <immintrin.h>
18459 int n_tries, max_tries;
18460 unsigned status = _XABORT_EXPLICIT;
18463 for (n_tries = 0; n_tries < max_tries; n_tries++)
18465 status = _xbegin ();
18466 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18469 if (status == _XBEGIN_STARTED)
18471 ... transaction code...
18476 ... non-transactional fallback path...
18481 Note that, in most cases, the transactional and non-transactional code
18482 must synchronize together to ensure consistency.
18484 @node Target Format Checks
18485 @section Format Checks Specific to Particular Target Machines
18487 For some target machines, GCC supports additional options to the
18489 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18492 * Solaris Format Checks::
18493 * Darwin Format Checks::
18496 @node Solaris Format Checks
18497 @subsection Solaris Format Checks
18499 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18500 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18501 conversions, and the two-argument @code{%b} conversion for displaying
18502 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18504 @node Darwin Format Checks
18505 @subsection Darwin Format Checks
18507 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18508 attribute context. Declarations made with such attribution are parsed for correct syntax
18509 and format argument types. However, parsing of the format string itself is currently undefined
18510 and is not carried out by this version of the compiler.
18512 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18513 also be used as format arguments. Note that the relevant headers are only likely to be
18514 available on Darwin (OSX) installations. On such installations, the XCode and system
18515 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18516 associated functions.
18519 @section Pragmas Accepted by GCC
18521 @cindex @code{#pragma}
18523 GCC supports several types of pragmas, primarily in order to compile
18524 code originally written for other compilers. Note that in general
18525 we do not recommend the use of pragmas; @xref{Function Attributes},
18526 for further explanation.
18529 * AArch64 Pragmas::
18533 * RS/6000 and PowerPC Pragmas::
18535 * Solaris Pragmas::
18536 * Symbol-Renaming Pragmas::
18537 * Structure-Layout Pragmas::
18539 * Diagnostic Pragmas::
18540 * Visibility Pragmas::
18541 * Push/Pop Macro Pragmas::
18542 * Function Specific Option Pragmas::
18543 * Loop-Specific Pragmas::
18546 @node AArch64 Pragmas
18547 @subsection AArch64 Pragmas
18549 The pragmas defined by the AArch64 target correspond to the AArch64
18550 target function attributes. They can be specified as below:
18552 #pragma GCC target("string")
18555 where @code{@var{string}} can be any string accepted as an AArch64 target
18556 attribute. @xref{AArch64 Function Attributes}, for more details
18557 on the permissible values of @code{string}.
18560 @subsection ARM Pragmas
18562 The ARM target defines pragmas for controlling the default addition of
18563 @code{long_call} and @code{short_call} attributes to functions.
18564 @xref{Function Attributes}, for information about the effects of these
18569 @cindex pragma, long_calls
18570 Set all subsequent functions to have the @code{long_call} attribute.
18572 @item no_long_calls
18573 @cindex pragma, no_long_calls
18574 Set all subsequent functions to have the @code{short_call} attribute.
18576 @item long_calls_off
18577 @cindex pragma, long_calls_off
18578 Do not affect the @code{long_call} or @code{short_call} attributes of
18579 subsequent functions.
18583 @subsection M32C Pragmas
18586 @item GCC memregs @var{number}
18587 @cindex pragma, memregs
18588 Overrides the command-line option @code{-memregs=} for the current
18589 file. Use with care! This pragma must be before any function in the
18590 file, and mixing different memregs values in different objects may
18591 make them incompatible. This pragma is useful when a
18592 performance-critical function uses a memreg for temporary values,
18593 as it may allow you to reduce the number of memregs used.
18595 @item ADDRESS @var{name} @var{address}
18596 @cindex pragma, address
18597 For any declared symbols matching @var{name}, this does three things
18598 to that symbol: it forces the symbol to be located at the given
18599 address (a number), it forces the symbol to be volatile, and it
18600 changes the symbol's scope to be static. This pragma exists for
18601 compatibility with other compilers, but note that the common
18602 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18606 #pragma ADDRESS port3 0x103
18613 @subsection MeP Pragmas
18617 @item custom io_volatile (on|off)
18618 @cindex pragma, custom io_volatile
18619 Overrides the command-line option @code{-mio-volatile} for the current
18620 file. Note that for compatibility with future GCC releases, this
18621 option should only be used once before any @code{io} variables in each
18624 @item GCC coprocessor available @var{registers}
18625 @cindex pragma, coprocessor available
18626 Specifies which coprocessor registers are available to the register
18627 allocator. @var{registers} may be a single register, register range
18628 separated by ellipses, or comma-separated list of those. Example:
18631 #pragma GCC coprocessor available $c0...$c10, $c28
18634 @item GCC coprocessor call_saved @var{registers}
18635 @cindex pragma, coprocessor call_saved
18636 Specifies which coprocessor registers are to be saved and restored by
18637 any function using them. @var{registers} may be a single register,
18638 register range separated by ellipses, or comma-separated list of
18642 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18645 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18646 @cindex pragma, coprocessor subclass
18647 Creates and defines a register class. These register classes can be
18648 used by inline @code{asm} constructs. @var{registers} may be a single
18649 register, register range separated by ellipses, or comma-separated
18650 list of those. Example:
18653 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18655 asm ("cpfoo %0" : "=B" (x));
18658 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18659 @cindex pragma, disinterrupt
18660 For the named functions, the compiler adds code to disable interrupts
18661 for the duration of those functions. If any functions so named
18662 are not encountered in the source, a warning is emitted that the pragma is
18663 not used. Examples:
18666 #pragma disinterrupt foo
18667 #pragma disinterrupt bar, grill
18668 int foo () @{ @dots{} @}
18671 @item GCC call @var{name} , @var{name} @dots{}
18672 @cindex pragma, call
18673 For the named functions, the compiler always uses a register-indirect
18674 call model when calling the named functions. Examples:
18683 @node RS/6000 and PowerPC Pragmas
18684 @subsection RS/6000 and PowerPC Pragmas
18686 The RS/6000 and PowerPC targets define one pragma for controlling
18687 whether or not the @code{longcall} attribute is added to function
18688 declarations by default. This pragma overrides the @option{-mlongcall}
18689 option, but not the @code{longcall} and @code{shortcall} attributes.
18690 @xref{RS/6000 and PowerPC Options}, for more information about when long
18691 calls are and are not necessary.
18695 @cindex pragma, longcall
18696 Apply the @code{longcall} attribute to all subsequent function
18700 Do not apply the @code{longcall} attribute to subsequent function
18704 @c Describe h8300 pragmas here.
18705 @c Describe sh pragmas here.
18706 @c Describe v850 pragmas here.
18708 @node Darwin Pragmas
18709 @subsection Darwin Pragmas
18711 The following pragmas are available for all architectures running the
18712 Darwin operating system. These are useful for compatibility with other
18716 @item mark @var{tokens}@dots{}
18717 @cindex pragma, mark
18718 This pragma is accepted, but has no effect.
18720 @item options align=@var{alignment}
18721 @cindex pragma, options align
18722 This pragma sets the alignment of fields in structures. The values of
18723 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
18724 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
18725 properly; to restore the previous setting, use @code{reset} for the
18728 @item segment @var{tokens}@dots{}
18729 @cindex pragma, segment
18730 This pragma is accepted, but has no effect.
18732 @item unused (@var{var} [, @var{var}]@dots{})
18733 @cindex pragma, unused
18734 This pragma declares variables to be possibly unused. GCC does not
18735 produce warnings for the listed variables. The effect is similar to
18736 that of the @code{unused} attribute, except that this pragma may appear
18737 anywhere within the variables' scopes.
18740 @node Solaris Pragmas
18741 @subsection Solaris Pragmas
18743 The Solaris target supports @code{#pragma redefine_extname}
18744 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
18745 @code{#pragma} directives for compatibility with the system compiler.
18748 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
18749 @cindex pragma, align
18751 Increase the minimum alignment of each @var{variable} to @var{alignment}.
18752 This is the same as GCC's @code{aligned} attribute @pxref{Variable
18753 Attributes}). Macro expansion occurs on the arguments to this pragma
18754 when compiling C and Objective-C@. It does not currently occur when
18755 compiling C++, but this is a bug which may be fixed in a future
18758 @item fini (@var{function} [, @var{function}]...)
18759 @cindex pragma, fini
18761 This pragma causes each listed @var{function} to be called after
18762 main, or during shared module unloading, by adding a call to the
18763 @code{.fini} section.
18765 @item init (@var{function} [, @var{function}]...)
18766 @cindex pragma, init
18768 This pragma causes each listed @var{function} to be called during
18769 initialization (before @code{main}) or during shared module loading, by
18770 adding a call to the @code{.init} section.
18774 @node Symbol-Renaming Pragmas
18775 @subsection Symbol-Renaming Pragmas
18777 GCC supports a @code{#pragma} directive that changes the name used in
18778 assembly for a given declaration. While this pragma is supported on all
18779 platforms, it is intended primarily to provide compatibility with the
18780 Solaris system headers. This effect can also be achieved using the asm
18781 labels extension (@pxref{Asm Labels}).
18784 @item redefine_extname @var{oldname} @var{newname}
18785 @cindex pragma, redefine_extname
18787 This pragma gives the C function @var{oldname} the assembly symbol
18788 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
18789 is defined if this pragma is available (currently on all platforms).
18792 This pragma and the asm labels extension interact in a complicated
18793 manner. Here are some corner cases you may want to be aware of:
18796 @item This pragma silently applies only to declarations with external
18797 linkage. Asm labels do not have this restriction.
18799 @item In C++, this pragma silently applies only to declarations with
18800 ``C'' linkage. Again, asm labels do not have this restriction.
18802 @item If either of the ways of changing the assembly name of a
18803 declaration are applied to a declaration whose assembly name has
18804 already been determined (either by a previous use of one of these
18805 features, or because the compiler needed the assembly name in order to
18806 generate code), and the new name is different, a warning issues and
18807 the name does not change.
18809 @item The @var{oldname} used by @code{#pragma redefine_extname} is
18810 always the C-language name.
18813 @node Structure-Layout Pragmas
18814 @subsection Structure-Layout Pragmas
18816 For compatibility with Microsoft Windows compilers, GCC supports a
18817 set of @code{#pragma} directives that change the maximum alignment of
18818 members of structures (other than zero-width bit-fields), unions, and
18819 classes subsequently defined. The @var{n} value below always is required
18820 to be a small power of two and specifies the new alignment in bytes.
18823 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
18824 @item @code{#pragma pack()} sets the alignment to the one that was in
18825 effect when compilation started (see also command-line option
18826 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
18827 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
18828 setting on an internal stack and then optionally sets the new alignment.
18829 @item @code{#pragma pack(pop)} restores the alignment setting to the one
18830 saved at the top of the internal stack (and removes that stack entry).
18831 Note that @code{#pragma pack([@var{n}])} does not influence this internal
18832 stack; thus it is possible to have @code{#pragma pack(push)} followed by
18833 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
18834 @code{#pragma pack(pop)}.
18837 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
18838 directive which lays out structures and unions subsequently defined as the
18839 documented @code{__attribute__ ((ms_struct))}.
18842 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
18843 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
18844 @item @code{#pragma ms_struct reset} goes back to the default layout.
18847 Most targets also support the @code{#pragma scalar_storage_order} directive
18848 which lays out structures and unions subsequently defined as the documented
18849 @code{__attribute__ ((scalar_storage_order))}.
18852 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
18853 of the scalar fields to big-endian.
18854 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
18855 of the scalar fields to little-endian.
18856 @item @code{#pragma scalar_storage_order default} goes back to the endianness
18857 that was in effect when compilation started (see also command-line option
18858 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
18862 @subsection Weak Pragmas
18864 For compatibility with SVR4, GCC supports a set of @code{#pragma}
18865 directives for declaring symbols to be weak, and defining weak
18869 @item #pragma weak @var{symbol}
18870 @cindex pragma, weak
18871 This pragma declares @var{symbol} to be weak, as if the declaration
18872 had the attribute of the same name. The pragma may appear before
18873 or after the declaration of @var{symbol}. It is not an error for
18874 @var{symbol} to never be defined at all.
18876 @item #pragma weak @var{symbol1} = @var{symbol2}
18877 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
18878 It is an error if @var{symbol2} is not defined in the current
18882 @node Diagnostic Pragmas
18883 @subsection Diagnostic Pragmas
18885 GCC allows the user to selectively enable or disable certain types of
18886 diagnostics, and change the kind of the diagnostic. For example, a
18887 project's policy might require that all sources compile with
18888 @option{-Werror} but certain files might have exceptions allowing
18889 specific types of warnings. Or, a project might selectively enable
18890 diagnostics and treat them as errors depending on which preprocessor
18891 macros are defined.
18894 @item #pragma GCC diagnostic @var{kind} @var{option}
18895 @cindex pragma, diagnostic
18897 Modifies the disposition of a diagnostic. Note that not all
18898 diagnostics are modifiable; at the moment only warnings (normally
18899 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
18900 Use @option{-fdiagnostics-show-option} to determine which diagnostics
18901 are controllable and which option controls them.
18903 @var{kind} is @samp{error} to treat this diagnostic as an error,
18904 @samp{warning} to treat it like a warning (even if @option{-Werror} is
18905 in effect), or @samp{ignored} if the diagnostic is to be ignored.
18906 @var{option} is a double quoted string that matches the command-line
18910 #pragma GCC diagnostic warning "-Wformat"
18911 #pragma GCC diagnostic error "-Wformat"
18912 #pragma GCC diagnostic ignored "-Wformat"
18915 Note that these pragmas override any command-line options. GCC keeps
18916 track of the location of each pragma, and issues diagnostics according
18917 to the state as of that point in the source file. Thus, pragmas occurring
18918 after a line do not affect diagnostics caused by that line.
18920 @item #pragma GCC diagnostic push
18921 @itemx #pragma GCC diagnostic pop
18923 Causes GCC to remember the state of the diagnostics as of each
18924 @code{push}, and restore to that point at each @code{pop}. If a
18925 @code{pop} has no matching @code{push}, the command-line options are
18929 #pragma GCC diagnostic error "-Wuninitialized"
18930 foo(a); /* error is given for this one */
18931 #pragma GCC diagnostic push
18932 #pragma GCC diagnostic ignored "-Wuninitialized"
18933 foo(b); /* no diagnostic for this one */
18934 #pragma GCC diagnostic pop
18935 foo(c); /* error is given for this one */
18936 #pragma GCC diagnostic pop
18937 foo(d); /* depends on command-line options */
18942 GCC also offers a simple mechanism for printing messages during
18946 @item #pragma message @var{string}
18947 @cindex pragma, diagnostic
18949 Prints @var{string} as a compiler message on compilation. The message
18950 is informational only, and is neither a compilation warning nor an error.
18953 #pragma message "Compiling " __FILE__ "..."
18956 @var{string} may be parenthesized, and is printed with location
18957 information. For example,
18960 #define DO_PRAGMA(x) _Pragma (#x)
18961 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
18963 TODO(Remember to fix this)
18967 prints @samp{/tmp/file.c:4: note: #pragma message:
18968 TODO - Remember to fix this}.
18972 @node Visibility Pragmas
18973 @subsection Visibility Pragmas
18976 @item #pragma GCC visibility push(@var{visibility})
18977 @itemx #pragma GCC visibility pop
18978 @cindex pragma, visibility
18980 This pragma allows the user to set the visibility for multiple
18981 declarations without having to give each a visibility attribute
18982 (@pxref{Function Attributes}).
18984 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
18985 declarations. Class members and template specializations are not
18986 affected; if you want to override the visibility for a particular
18987 member or instantiation, you must use an attribute.
18992 @node Push/Pop Macro Pragmas
18993 @subsection Push/Pop Macro Pragmas
18995 For compatibility with Microsoft Windows compilers, GCC supports
18996 @samp{#pragma push_macro(@var{"macro_name"})}
18997 and @samp{#pragma pop_macro(@var{"macro_name"})}.
19000 @item #pragma push_macro(@var{"macro_name"})
19001 @cindex pragma, push_macro
19002 This pragma saves the value of the macro named as @var{macro_name} to
19003 the top of the stack for this macro.
19005 @item #pragma pop_macro(@var{"macro_name"})
19006 @cindex pragma, pop_macro
19007 This pragma sets the value of the macro named as @var{macro_name} to
19008 the value on top of the stack for this macro. If the stack for
19009 @var{macro_name} is empty, the value of the macro remains unchanged.
19016 #pragma push_macro("X")
19019 #pragma pop_macro("X")
19024 In this example, the definition of X as 1 is saved by @code{#pragma
19025 push_macro} and restored by @code{#pragma pop_macro}.
19027 @node Function Specific Option Pragmas
19028 @subsection Function Specific Option Pragmas
19031 @item #pragma GCC target (@var{"string"}...)
19032 @cindex pragma GCC target
19034 This pragma allows you to set target specific options for functions
19035 defined later in the source file. One or more strings can be
19036 specified. Each function that is defined after this point is as
19037 if @code{attribute((target("STRING")))} was specified for that
19038 function. The parenthesis around the options is optional.
19039 @xref{Function Attributes}, for more information about the
19040 @code{target} attribute and the attribute syntax.
19042 The @code{#pragma GCC target} pragma is presently implemented for
19043 x86, PowerPC, and Nios II targets only.
19047 @item #pragma GCC optimize (@var{"string"}...)
19048 @cindex pragma GCC optimize
19050 This pragma allows you to set global optimization options for functions
19051 defined later in the source file. One or more strings can be
19052 specified. Each function that is defined after this point is as
19053 if @code{attribute((optimize("STRING")))} was specified for that
19054 function. The parenthesis around the options is optional.
19055 @xref{Function Attributes}, for more information about the
19056 @code{optimize} attribute and the attribute syntax.
19060 @item #pragma GCC push_options
19061 @itemx #pragma GCC pop_options
19062 @cindex pragma GCC push_options
19063 @cindex pragma GCC pop_options
19065 These pragmas maintain a stack of the current target and optimization
19066 options. It is intended for include files where you temporarily want
19067 to switch to using a different @samp{#pragma GCC target} or
19068 @samp{#pragma GCC optimize} and then to pop back to the previous
19073 @item #pragma GCC reset_options
19074 @cindex pragma GCC reset_options
19076 This pragma clears the current @code{#pragma GCC target} and
19077 @code{#pragma GCC optimize} to use the default switches as specified
19078 on the command line.
19081 @node Loop-Specific Pragmas
19082 @subsection Loop-Specific Pragmas
19085 @item #pragma GCC ivdep
19086 @cindex pragma GCC ivdep
19089 With this pragma, the programmer asserts that there are no loop-carried
19090 dependencies which would prevent consecutive iterations of
19091 the following loop from executing concurrently with SIMD
19092 (single instruction multiple data) instructions.
19094 For example, the compiler can only unconditionally vectorize the following
19095 loop with the pragma:
19098 void foo (int n, int *a, int *b, int *c)
19102 for (i = 0; i < n; ++i)
19103 a[i] = b[i] + c[i];
19108 In this example, using the @code{restrict} qualifier had the same
19109 effect. In the following example, that would not be possible. Assume
19110 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
19111 that it can unconditionally vectorize the following loop:
19114 void ignore_vec_dep (int *a, int k, int c, int m)
19117 for (int i = 0; i < m; i++)
19118 a[i] = a[i + k] * c;
19123 @node Unnamed Fields
19124 @section Unnamed Structure and Union Fields
19125 @cindex @code{struct}
19126 @cindex @code{union}
19128 As permitted by ISO C11 and for compatibility with other compilers,
19129 GCC allows you to define
19130 a structure or union that contains, as fields, structures and unions
19131 without names. For example:
19145 In this example, you are able to access members of the unnamed
19146 union with code like @samp{foo.b}. Note that only unnamed structs and
19147 unions are allowed, you may not have, for example, an unnamed
19150 You must never create such structures that cause ambiguous field definitions.
19151 For example, in this structure:
19163 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
19164 The compiler gives errors for such constructs.
19166 @opindex fms-extensions
19167 Unless @option{-fms-extensions} is used, the unnamed field must be a
19168 structure or union definition without a tag (for example, @samp{struct
19169 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
19170 also be a definition with a tag such as @samp{struct foo @{ int a;
19171 @};}, a reference to a previously defined structure or union such as
19172 @samp{struct foo;}, or a reference to a @code{typedef} name for a
19173 previously defined structure or union type.
19175 @opindex fplan9-extensions
19176 The option @option{-fplan9-extensions} enables
19177 @option{-fms-extensions} as well as two other extensions. First, a
19178 pointer to a structure is automatically converted to a pointer to an
19179 anonymous field for assignments and function calls. For example:
19182 struct s1 @{ int a; @};
19183 struct s2 @{ struct s1; @};
19184 extern void f1 (struct s1 *);
19185 void f2 (struct s2 *p) @{ f1 (p); @}
19189 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
19190 converted into a pointer to the anonymous field.
19192 Second, when the type of an anonymous field is a @code{typedef} for a
19193 @code{struct} or @code{union}, code may refer to the field using the
19194 name of the @code{typedef}.
19197 typedef struct @{ int a; @} s1;
19198 struct s2 @{ s1; @};
19199 s1 f1 (struct s2 *p) @{ return p->s1; @}
19202 These usages are only permitted when they are not ambiguous.
19205 @section Thread-Local Storage
19206 @cindex Thread-Local Storage
19207 @cindex @acronym{TLS}
19208 @cindex @code{__thread}
19210 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
19211 are allocated such that there is one instance of the variable per extant
19212 thread. The runtime model GCC uses to implement this originates
19213 in the IA-64 processor-specific ABI, but has since been migrated
19214 to other processors as well. It requires significant support from
19215 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
19216 system libraries (@file{libc.so} and @file{libpthread.so}), so it
19217 is not available everywhere.
19219 At the user level, the extension is visible with a new storage
19220 class keyword: @code{__thread}. For example:
19224 extern __thread struct state s;
19225 static __thread char *p;
19228 The @code{__thread} specifier may be used alone, with the @code{extern}
19229 or @code{static} specifiers, but with no other storage class specifier.
19230 When used with @code{extern} or @code{static}, @code{__thread} must appear
19231 immediately after the other storage class specifier.
19233 The @code{__thread} specifier may be applied to any global, file-scoped
19234 static, function-scoped static, or static data member of a class. It may
19235 not be applied to block-scoped automatic or non-static data member.
19237 When the address-of operator is applied to a thread-local variable, it is
19238 evaluated at run time and returns the address of the current thread's
19239 instance of that variable. An address so obtained may be used by any
19240 thread. When a thread terminates, any pointers to thread-local variables
19241 in that thread become invalid.
19243 No static initialization may refer to the address of a thread-local variable.
19245 In C++, if an initializer is present for a thread-local variable, it must
19246 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19249 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19250 ELF Handling For Thread-Local Storage} for a detailed explanation of
19251 the four thread-local storage addressing models, and how the runtime
19252 is expected to function.
19255 * C99 Thread-Local Edits::
19256 * C++98 Thread-Local Edits::
19259 @node C99 Thread-Local Edits
19260 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19262 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19263 that document the exact semantics of the language extension.
19267 @cite{5.1.2 Execution environments}
19269 Add new text after paragraph 1
19272 Within either execution environment, a @dfn{thread} is a flow of
19273 control within a program. It is implementation defined whether
19274 or not there may be more than one thread associated with a program.
19275 It is implementation defined how threads beyond the first are
19276 created, the name and type of the function called at thread
19277 startup, and how threads may be terminated. However, objects
19278 with thread storage duration shall be initialized before thread
19283 @cite{6.2.4 Storage durations of objects}
19285 Add new text before paragraph 3
19288 An object whose identifier is declared with the storage-class
19289 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19290 Its lifetime is the entire execution of the thread, and its
19291 stored value is initialized only once, prior to thread startup.
19295 @cite{6.4.1 Keywords}
19297 Add @code{__thread}.
19300 @cite{6.7.1 Storage-class specifiers}
19302 Add @code{__thread} to the list of storage class specifiers in
19305 Change paragraph 2 to
19308 With the exception of @code{__thread}, at most one storage-class
19309 specifier may be given [@dots{}]. The @code{__thread} specifier may
19310 be used alone, or immediately following @code{extern} or
19314 Add new text after paragraph 6
19317 The declaration of an identifier for a variable that has
19318 block scope that specifies @code{__thread} shall also
19319 specify either @code{extern} or @code{static}.
19321 The @code{__thread} specifier shall be used only with
19326 @node C++98 Thread-Local Edits
19327 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19329 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19330 that document the exact semantics of the language extension.
19334 @b{[intro.execution]}
19336 New text after paragraph 4
19339 A @dfn{thread} is a flow of control within the abstract machine.
19340 It is implementation defined whether or not there may be more than
19344 New text after paragraph 7
19347 It is unspecified whether additional action must be taken to
19348 ensure when and whether side effects are visible to other threads.
19354 Add @code{__thread}.
19357 @b{[basic.start.main]}
19359 Add after paragraph 5
19362 The thread that begins execution at the @code{main} function is called
19363 the @dfn{main thread}. It is implementation defined how functions
19364 beginning threads other than the main thread are designated or typed.
19365 A function so designated, as well as the @code{main} function, is called
19366 a @dfn{thread startup function}. It is implementation defined what
19367 happens if a thread startup function returns. It is implementation
19368 defined what happens to other threads when any thread calls @code{exit}.
19372 @b{[basic.start.init]}
19374 Add after paragraph 4
19377 The storage for an object of thread storage duration shall be
19378 statically initialized before the first statement of the thread startup
19379 function. An object of thread storage duration shall not require
19380 dynamic initialization.
19384 @b{[basic.start.term]}
19386 Add after paragraph 3
19389 The type of an object with thread storage duration shall not have a
19390 non-trivial destructor, nor shall it be an array type whose elements
19391 (directly or indirectly) have non-trivial destructors.
19397 Add ``thread storage duration'' to the list in paragraph 1.
19402 Thread, static, and automatic storage durations are associated with
19403 objects introduced by declarations [@dots{}].
19406 Add @code{__thread} to the list of specifiers in paragraph 3.
19409 @b{[basic.stc.thread]}
19411 New section before @b{[basic.stc.static]}
19414 The keyword @code{__thread} applied to a non-local object gives the
19415 object thread storage duration.
19417 A local variable or class data member declared both @code{static}
19418 and @code{__thread} gives the variable or member thread storage
19423 @b{[basic.stc.static]}
19428 All objects that have neither thread storage duration, dynamic
19429 storage duration nor are local [@dots{}].
19435 Add @code{__thread} to the list in paragraph 1.
19440 With the exception of @code{__thread}, at most one
19441 @var{storage-class-specifier} shall appear in a given
19442 @var{decl-specifier-seq}. The @code{__thread} specifier may
19443 be used alone, or immediately following the @code{extern} or
19444 @code{static} specifiers. [@dots{}]
19447 Add after paragraph 5
19450 The @code{__thread} specifier can be applied only to the names of objects
19451 and to anonymous unions.
19457 Add after paragraph 6
19460 Non-@code{static} members shall not be @code{__thread}.
19464 @node Binary constants
19465 @section Binary Constants using the @samp{0b} Prefix
19466 @cindex Binary constants using the @samp{0b} prefix
19468 Integer constants can be written as binary constants, consisting of a
19469 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19470 @samp{0B}. This is particularly useful in environments that operate a
19471 lot on the bit level (like microcontrollers).
19473 The following statements are identical:
19482 The type of these constants follows the same rules as for octal or
19483 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19486 @node C++ Extensions
19487 @chapter Extensions to the C++ Language
19488 @cindex extensions, C++ language
19489 @cindex C++ language extensions
19491 The GNU compiler provides these extensions to the C++ language (and you
19492 can also use most of the C language extensions in your C++ programs). If you
19493 want to write code that checks whether these features are available, you can
19494 test for the GNU compiler the same way as for C programs: check for a
19495 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19496 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19497 Predefined Macros,cpp,The GNU C Preprocessor}).
19500 * C++ Volatiles:: What constitutes an access to a volatile object.
19501 * Restricted Pointers:: C99 restricted pointers and references.
19502 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19503 * C++ Interface:: You can use a single C++ header file for both
19504 declarations and definitions.
19505 * Template Instantiation:: Methods for ensuring that exactly one copy of
19506 each needed template instantiation is emitted.
19507 * Bound member functions:: You can extract a function pointer to the
19508 method denoted by a @samp{->*} or @samp{.*} expression.
19509 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19510 * Function Multiversioning:: Declaring multiple function versions.
19511 * Namespace Association:: Strong using-directives for namespace association.
19512 * Type Traits:: Compiler support for type traits.
19513 * C++ Concepts:: Improved support for generic programming.
19514 * Java Exceptions:: Tweaking exception handling to work with Java.
19515 * Deprecated Features:: Things will disappear from G++.
19516 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19519 @node C++ Volatiles
19520 @section When is a Volatile C++ Object Accessed?
19521 @cindex accessing volatiles
19522 @cindex volatile read
19523 @cindex volatile write
19524 @cindex volatile access
19526 The C++ standard differs from the C standard in its treatment of
19527 volatile objects. It fails to specify what constitutes a volatile
19528 access, except to say that C++ should behave in a similar manner to C
19529 with respect to volatiles, where possible. However, the different
19530 lvalueness of expressions between C and C++ complicate the behavior.
19531 G++ behaves the same as GCC for volatile access, @xref{C
19532 Extensions,,Volatiles}, for a description of GCC's behavior.
19534 The C and C++ language specifications differ when an object is
19535 accessed in a void context:
19538 volatile int *src = @var{somevalue};
19542 The C++ standard specifies that such expressions do not undergo lvalue
19543 to rvalue conversion, and that the type of the dereferenced object may
19544 be incomplete. The C++ standard does not specify explicitly that it
19545 is lvalue to rvalue conversion that is responsible for causing an
19546 access. There is reason to believe that it is, because otherwise
19547 certain simple expressions become undefined. However, because it
19548 would surprise most programmers, G++ treats dereferencing a pointer to
19549 volatile object of complete type as GCC would do for an equivalent
19550 type in C@. When the object has incomplete type, G++ issues a
19551 warning; if you wish to force an error, you must force a conversion to
19552 rvalue with, for instance, a static cast.
19554 When using a reference to volatile, G++ does not treat equivalent
19555 expressions as accesses to volatiles, but instead issues a warning that
19556 no volatile is accessed. The rationale for this is that otherwise it
19557 becomes difficult to determine where volatile access occur, and not
19558 possible to ignore the return value from functions returning volatile
19559 references. Again, if you wish to force a read, cast the reference to
19562 G++ implements the same behavior as GCC does when assigning to a
19563 volatile object---there is no reread of the assigned-to object, the
19564 assigned rvalue is reused. Note that in C++ assignment expressions
19565 are lvalues, and if used as an lvalue, the volatile object is
19566 referred to. For instance, @var{vref} refers to @var{vobj}, as
19567 expected, in the following example:
19571 volatile int &vref = vobj = @var{something};
19574 @node Restricted Pointers
19575 @section Restricting Pointer Aliasing
19576 @cindex restricted pointers
19577 @cindex restricted references
19578 @cindex restricted this pointer
19580 As with the C front end, G++ understands the C99 feature of restricted pointers,
19581 specified with the @code{__restrict__}, or @code{__restrict} type
19582 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19583 language flag, @code{restrict} is not a keyword in C++.
19585 In addition to allowing restricted pointers, you can specify restricted
19586 references, which indicate that the reference is not aliased in the local
19590 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19597 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19598 @var{rref} refers to a (different) unaliased integer.
19600 You may also specify whether a member function's @var{this} pointer is
19601 unaliased by using @code{__restrict__} as a member function qualifier.
19604 void T::fn () __restrict__
19611 Within the body of @code{T::fn}, @var{this} has the effective
19612 definition @code{T *__restrict__ const this}. Notice that the
19613 interpretation of a @code{__restrict__} member function qualifier is
19614 different to that of @code{const} or @code{volatile} qualifier, in that it
19615 is applied to the pointer rather than the object. This is consistent with
19616 other compilers that implement restricted pointers.
19618 As with all outermost parameter qualifiers, @code{__restrict__} is
19619 ignored in function definition matching. This means you only need to
19620 specify @code{__restrict__} in a function definition, rather than
19621 in a function prototype as well.
19623 @node Vague Linkage
19624 @section Vague Linkage
19625 @cindex vague linkage
19627 There are several constructs in C++ that require space in the object
19628 file but are not clearly tied to a single translation unit. We say that
19629 these constructs have ``vague linkage''. Typically such constructs are
19630 emitted wherever they are needed, though sometimes we can be more
19634 @item Inline Functions
19635 Inline functions are typically defined in a header file which can be
19636 included in many different compilations. Hopefully they can usually be
19637 inlined, but sometimes an out-of-line copy is necessary, if the address
19638 of the function is taken or if inlining fails. In general, we emit an
19639 out-of-line copy in all translation units where one is needed. As an
19640 exception, we only emit inline virtual functions with the vtable, since
19641 it always requires a copy.
19643 Local static variables and string constants used in an inline function
19644 are also considered to have vague linkage, since they must be shared
19645 between all inlined and out-of-line instances of the function.
19649 C++ virtual functions are implemented in most compilers using a lookup
19650 table, known as a vtable. The vtable contains pointers to the virtual
19651 functions provided by a class, and each object of the class contains a
19652 pointer to its vtable (or vtables, in some multiple-inheritance
19653 situations). If the class declares any non-inline, non-pure virtual
19654 functions, the first one is chosen as the ``key method'' for the class,
19655 and the vtable is only emitted in the translation unit where the key
19658 @emph{Note:} If the chosen key method is later defined as inline, the
19659 vtable is still emitted in every translation unit that defines it.
19660 Make sure that any inline virtuals are declared inline in the class
19661 body, even if they are not defined there.
19663 @item @code{type_info} objects
19664 @cindex @code{type_info}
19666 C++ requires information about types to be written out in order to
19667 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19668 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19669 object is written out along with the vtable so that @samp{dynamic_cast}
19670 can determine the dynamic type of a class object at run time. For all
19671 other types, we write out the @samp{type_info} object when it is used: when
19672 applying @samp{typeid} to an expression, throwing an object, or
19673 referring to a type in a catch clause or exception specification.
19675 @item Template Instantiations
19676 Most everything in this section also applies to template instantiations,
19677 but there are other options as well.
19678 @xref{Template Instantiation,,Where's the Template?}.
19682 When used with GNU ld version 2.8 or later on an ELF system such as
19683 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
19684 these constructs will be discarded at link time. This is known as
19687 On targets that don't support COMDAT, but do support weak symbols, GCC
19688 uses them. This way one copy overrides all the others, but
19689 the unused copies still take up space in the executable.
19691 For targets that do not support either COMDAT or weak symbols,
19692 most entities with vague linkage are emitted as local symbols to
19693 avoid duplicate definition errors from the linker. This does not happen
19694 for local statics in inlines, however, as having multiple copies
19695 almost certainly breaks things.
19697 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
19698 another way to control placement of these constructs.
19700 @node C++ Interface
19701 @section C++ Interface and Implementation Pragmas
19703 @cindex interface and implementation headers, C++
19704 @cindex C++ interface and implementation headers
19705 @cindex pragmas, interface and implementation
19707 @code{#pragma interface} and @code{#pragma implementation} provide the
19708 user with a way of explicitly directing the compiler to emit entities
19709 with vague linkage (and debugging information) in a particular
19712 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
19713 by COMDAT support and the ``key method'' heuristic
19714 mentioned in @ref{Vague Linkage}. Using them can actually cause your
19715 program to grow due to unnecessary out-of-line copies of inline
19719 @item #pragma interface
19720 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
19721 @kindex #pragma interface
19722 Use this directive in @emph{header files} that define object classes, to save
19723 space in most of the object files that use those classes. Normally,
19724 local copies of certain information (backup copies of inline member
19725 functions, debugging information, and the internal tables that implement
19726 virtual functions) must be kept in each object file that includes class
19727 definitions. You can use this pragma to avoid such duplication. When a
19728 header file containing @samp{#pragma interface} is included in a
19729 compilation, this auxiliary information is not generated (unless
19730 the main input source file itself uses @samp{#pragma implementation}).
19731 Instead, the object files contain references to be resolved at link
19734 The second form of this directive is useful for the case where you have
19735 multiple headers with the same name in different directories. If you
19736 use this form, you must specify the same string to @samp{#pragma
19739 @item #pragma implementation
19740 @itemx #pragma implementation "@var{objects}.h"
19741 @kindex #pragma implementation
19742 Use this pragma in a @emph{main input file}, when you want full output from
19743 included header files to be generated (and made globally visible). The
19744 included header file, in turn, should use @samp{#pragma interface}.
19745 Backup copies of inline member functions, debugging information, and the
19746 internal tables used to implement virtual functions are all generated in
19747 implementation files.
19749 @cindex implied @code{#pragma implementation}
19750 @cindex @code{#pragma implementation}, implied
19751 @cindex naming convention, implementation headers
19752 If you use @samp{#pragma implementation} with no argument, it applies to
19753 an include file with the same basename@footnote{A file's @dfn{basename}
19754 is the name stripped of all leading path information and of trailing
19755 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
19756 file. For example, in @file{allclass.cc}, giving just
19757 @samp{#pragma implementation}
19758 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
19760 Use the string argument if you want a single implementation file to
19761 include code from multiple header files. (You must also use
19762 @samp{#include} to include the header file; @samp{#pragma
19763 implementation} only specifies how to use the file---it doesn't actually
19766 There is no way to split up the contents of a single header file into
19767 multiple implementation files.
19770 @cindex inlining and C++ pragmas
19771 @cindex C++ pragmas, effect on inlining
19772 @cindex pragmas in C++, effect on inlining
19773 @samp{#pragma implementation} and @samp{#pragma interface} also have an
19774 effect on function inlining.
19776 If you define a class in a header file marked with @samp{#pragma
19777 interface}, the effect on an inline function defined in that class is
19778 similar to an explicit @code{extern} declaration---the compiler emits
19779 no code at all to define an independent version of the function. Its
19780 definition is used only for inlining with its callers.
19782 @opindex fno-implement-inlines
19783 Conversely, when you include the same header file in a main source file
19784 that declares it as @samp{#pragma implementation}, the compiler emits
19785 code for the function itself; this defines a version of the function
19786 that can be found via pointers (or by callers compiled without
19787 inlining). If all calls to the function can be inlined, you can avoid
19788 emitting the function by compiling with @option{-fno-implement-inlines}.
19789 If any calls are not inlined, you will get linker errors.
19791 @node Template Instantiation
19792 @section Where's the Template?
19793 @cindex template instantiation
19795 C++ templates were the first language feature to require more
19796 intelligence from the environment than was traditionally found on a UNIX
19797 system. Somehow the compiler and linker have to make sure that each
19798 template instance occurs exactly once in the executable if it is needed,
19799 and not at all otherwise. There are two basic approaches to this
19800 problem, which are referred to as the Borland model and the Cfront model.
19803 @item Borland model
19804 Borland C++ solved the template instantiation problem by adding the code
19805 equivalent of common blocks to their linker; the compiler emits template
19806 instances in each translation unit that uses them, and the linker
19807 collapses them together. The advantage of this model is that the linker
19808 only has to consider the object files themselves; there is no external
19809 complexity to worry about. The disadvantage is that compilation time
19810 is increased because the template code is being compiled repeatedly.
19811 Code written for this model tends to include definitions of all
19812 templates in the header file, since they must be seen to be
19816 The AT&T C++ translator, Cfront, solved the template instantiation
19817 problem by creating the notion of a template repository, an
19818 automatically maintained place where template instances are stored. A
19819 more modern version of the repository works as follows: As individual
19820 object files are built, the compiler places any template definitions and
19821 instantiations encountered in the repository. At link time, the link
19822 wrapper adds in the objects in the repository and compiles any needed
19823 instances that were not previously emitted. The advantages of this
19824 model are more optimal compilation speed and the ability to use the
19825 system linker; to implement the Borland model a compiler vendor also
19826 needs to replace the linker. The disadvantages are vastly increased
19827 complexity, and thus potential for error; for some code this can be
19828 just as transparent, but in practice it can been very difficult to build
19829 multiple programs in one directory and one program in multiple
19830 directories. Code written for this model tends to separate definitions
19831 of non-inline member templates into a separate file, which should be
19832 compiled separately.
19835 G++ implements the Borland model on targets where the linker supports it,
19836 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
19837 Otherwise G++ implements neither automatic model.
19839 You have the following options for dealing with template instantiations:
19843 Do nothing. Code written for the Borland model works fine, but
19844 each translation unit contains instances of each of the templates it
19845 uses. The duplicate instances will be discarded by the linker, but in
19846 a large program, this can lead to an unacceptable amount of code
19847 duplication in object files or shared libraries.
19849 Duplicate instances of a template can be avoided by defining an explicit
19850 instantiation in one object file, and preventing the compiler from doing
19851 implicit instantiations in any other object files by using an explicit
19852 instantiation declaration, using the @code{extern template} syntax:
19855 extern template int max (int, int);
19858 This syntax is defined in the C++ 2011 standard, but has been supported by
19859 G++ and other compilers since well before 2011.
19861 Explicit instantiations can be used for the largest or most frequently
19862 duplicated instances, without having to know exactly which other instances
19863 are used in the rest of the program. You can scatter the explicit
19864 instantiations throughout your program, perhaps putting them in the
19865 translation units where the instances are used or the translation units
19866 that define the templates themselves; you can put all of the explicit
19867 instantiations you need into one big file; or you can create small files
19874 template class Foo<int>;
19875 template ostream& operator <<
19876 (ostream&, const Foo<int>&);
19880 for each of the instances you need, and create a template instantiation
19881 library from those.
19883 This is the simplest option, but also offers flexibility and
19884 fine-grained control when necessary. It is also the most portable
19885 alternative and programs using this approach will work with most modern
19890 Compile your template-using code with @option{-frepo}. The compiler
19891 generates files with the extension @samp{.rpo} listing all of the
19892 template instantiations used in the corresponding object files that
19893 could be instantiated there; the link wrapper, @samp{collect2},
19894 then updates the @samp{.rpo} files to tell the compiler where to place
19895 those instantiations and rebuild any affected object files. The
19896 link-time overhead is negligible after the first pass, as the compiler
19897 continues to place the instantiations in the same files.
19899 This can be a suitable option for application code written for the Borland
19900 model, as it usually just works. Code written for the Cfront model
19901 needs to be modified so that the template definitions are available at
19902 one or more points of instantiation; usually this is as simple as adding
19903 @code{#include <tmethods.cc>} to the end of each template header.
19905 For library code, if you want the library to provide all of the template
19906 instantiations it needs, just try to link all of its object files
19907 together; the link will fail, but cause the instantiations to be
19908 generated as a side effect. Be warned, however, that this may cause
19909 conflicts if multiple libraries try to provide the same instantiations.
19910 For greater control, use explicit instantiation as described in the next
19914 @opindex fno-implicit-templates
19915 Compile your code with @option{-fno-implicit-templates} to disable the
19916 implicit generation of template instances, and explicitly instantiate
19917 all the ones you use. This approach requires more knowledge of exactly
19918 which instances you need than do the others, but it's less
19919 mysterious and allows greater control if you want to ensure that only
19920 the intended instances are used.
19922 If you are using Cfront-model code, you can probably get away with not
19923 using @option{-fno-implicit-templates} when compiling files that don't
19924 @samp{#include} the member template definitions.
19926 If you use one big file to do the instantiations, you may want to
19927 compile it without @option{-fno-implicit-templates} so you get all of the
19928 instances required by your explicit instantiations (but not by any
19929 other files) without having to specify them as well.
19931 In addition to forward declaration of explicit instantiations
19932 (with @code{extern}), G++ has extended the template instantiation
19933 syntax to support instantiation of the compiler support data for a
19934 template class (i.e.@: the vtable) without instantiating any of its
19935 members (with @code{inline}), and instantiation of only the static data
19936 members of a template class, without the support data or member
19937 functions (with @code{static}):
19940 inline template class Foo<int>;
19941 static template class Foo<int>;
19945 @node Bound member functions
19946 @section Extracting the Function Pointer from a Bound Pointer to Member Function
19948 @cindex pointer to member function
19949 @cindex bound pointer to member function
19951 In C++, pointer to member functions (PMFs) are implemented using a wide
19952 pointer of sorts to handle all the possible call mechanisms; the PMF
19953 needs to store information about how to adjust the @samp{this} pointer,
19954 and if the function pointed to is virtual, where to find the vtable, and
19955 where in the vtable to look for the member function. If you are using
19956 PMFs in an inner loop, you should really reconsider that decision. If
19957 that is not an option, you can extract the pointer to the function that
19958 would be called for a given object/PMF pair and call it directly inside
19959 the inner loop, to save a bit of time.
19961 Note that you still pay the penalty for the call through a
19962 function pointer; on most modern architectures, such a call defeats the
19963 branch prediction features of the CPU@. This is also true of normal
19964 virtual function calls.
19966 The syntax for this extension is
19970 extern int (A::*fp)();
19971 typedef int (*fptr)(A *);
19973 fptr p = (fptr)(a.*fp);
19976 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
19977 no object is needed to obtain the address of the function. They can be
19978 converted to function pointers directly:
19981 fptr p1 = (fptr)(&A::foo);
19984 @opindex Wno-pmf-conversions
19985 You must specify @option{-Wno-pmf-conversions} to use this extension.
19987 @node C++ Attributes
19988 @section C++-Specific Variable, Function, and Type Attributes
19990 Some attributes only make sense for C++ programs.
19993 @item abi_tag ("@var{tag}", ...)
19994 @cindex @code{abi_tag} function attribute
19995 @cindex @code{abi_tag} variable attribute
19996 @cindex @code{abi_tag} type attribute
19997 The @code{abi_tag} attribute can be applied to a function, variable, or class
19998 declaration. It modifies the mangled name of the entity to
19999 incorporate the tag name, in order to distinguish the function or
20000 class from an earlier version with a different ABI; perhaps the class
20001 has changed size, or the function has a different return type that is
20002 not encoded in the mangled name.
20004 The attribute can also be applied to an inline namespace, but does not
20005 affect the mangled name of the namespace; in this case it is only used
20006 for @option{-Wabi-tag} warnings and automatic tagging of functions and
20007 variables. Tagging inline namespaces is generally preferable to
20008 tagging individual declarations, but the latter is sometimes
20009 necessary, such as when only certain members of a class need to be
20012 The argument can be a list of strings of arbitrary length. The
20013 strings are sorted on output, so the order of the list is
20016 A redeclaration of an entity must not add new ABI tags,
20017 since doing so would change the mangled name.
20019 The ABI tags apply to a name, so all instantiations and
20020 specializations of a template have the same tags. The attribute will
20021 be ignored if applied to an explicit specialization or instantiation.
20023 The @option{-Wabi-tag} flag enables a warning about a class which does
20024 not have all the ABI tags used by its subobjects and virtual functions; for users with code
20025 that needs to coexist with an earlier ABI, using this option can help
20026 to find all affected types that need to be tagged.
20028 When a type involving an ABI tag is used as the type of a variable or
20029 return type of a function where that tag is not already present in the
20030 signature of the function, the tag is automatically applied to the
20031 variable or function. @option{-Wabi-tag} also warns about this
20032 situation; this warning can be avoided by explicitly tagging the
20033 variable or function or moving it into a tagged inline namespace.
20035 @item init_priority (@var{priority})
20036 @cindex @code{init_priority} variable attribute
20038 In Standard C++, objects defined at namespace scope are guaranteed to be
20039 initialized in an order in strict accordance with that of their definitions
20040 @emph{in a given translation unit}. No guarantee is made for initializations
20041 across translation units. However, GNU C++ allows users to control the
20042 order of initialization of objects defined at namespace scope with the
20043 @code{init_priority} attribute by specifying a relative @var{priority},
20044 a constant integral expression currently bounded between 101 and 65535
20045 inclusive. Lower numbers indicate a higher priority.
20047 In the following example, @code{A} would normally be created before
20048 @code{B}, but the @code{init_priority} attribute reverses that order:
20051 Some_Class A __attribute__ ((init_priority (2000)));
20052 Some_Class B __attribute__ ((init_priority (543)));
20056 Note that the particular values of @var{priority} do not matter; only their
20059 @item java_interface
20060 @cindex @code{java_interface} type attribute
20062 This type attribute informs C++ that the class is a Java interface. It may
20063 only be applied to classes declared within an @code{extern "Java"} block.
20064 Calls to methods declared in this interface are dispatched using GCJ's
20065 interface table mechanism, instead of regular virtual table dispatch.
20068 @cindex @code{warn_unused} type attribute
20070 For C++ types with non-trivial constructors and/or destructors it is
20071 impossible for the compiler to determine whether a variable of this
20072 type is truly unused if it is not referenced. This type attribute
20073 informs the compiler that variables of this type should be warned
20074 about if they appear to be unused, just like variables of fundamental
20077 This attribute is appropriate for types which just represent a value,
20078 such as @code{std::string}; it is not appropriate for types which
20079 control a resource, such as @code{std::mutex}.
20081 This attribute is also accepted in C, but it is unnecessary because C
20082 does not have constructors or destructors.
20086 See also @ref{Namespace Association}.
20088 @node Function Multiversioning
20089 @section Function Multiversioning
20090 @cindex function versions
20092 With the GNU C++ front end, for x86 targets, you may specify multiple
20093 versions of a function, where each function is specialized for a
20094 specific target feature. At runtime, the appropriate version of the
20095 function is automatically executed depending on the characteristics of
20096 the execution platform. Here is an example.
20099 __attribute__ ((target ("default")))
20102 // The default version of foo.
20106 __attribute__ ((target ("sse4.2")))
20109 // foo version for SSE4.2
20113 __attribute__ ((target ("arch=atom")))
20116 // foo version for the Intel ATOM processor
20120 __attribute__ ((target ("arch=amdfam10")))
20123 // foo version for the AMD Family 0x10 processors.
20130 assert ((*p) () == foo ());
20135 In the above example, four versions of function foo are created. The
20136 first version of foo with the target attribute "default" is the default
20137 version. This version gets executed when no other target specific
20138 version qualifies for execution on a particular platform. A new version
20139 of foo is created by using the same function signature but with a
20140 different target string. Function foo is called or a pointer to it is
20141 taken just like a regular function. GCC takes care of doing the
20142 dispatching to call the right version at runtime. Refer to the
20143 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
20144 Function Multiversioning} for more details.
20146 @node Namespace Association
20147 @section Namespace Association
20149 @strong{Caution:} The semantics of this extension are equivalent
20150 to C++ 2011 inline namespaces. Users should use inline namespaces
20151 instead as this extension will be removed in future versions of G++.
20153 A using-directive with @code{__attribute ((strong))} is stronger
20154 than a normal using-directive in two ways:
20158 Templates from the used namespace can be specialized and explicitly
20159 instantiated as though they were members of the using namespace.
20162 The using namespace is considered an associated namespace of all
20163 templates in the used namespace for purposes of argument-dependent
20167 The used namespace must be nested within the using namespace so that
20168 normal unqualified lookup works properly.
20170 This is useful for composing a namespace transparently from
20171 implementation namespaces. For example:
20176 template <class T> struct A @{ @};
20178 using namespace debug __attribute ((__strong__));
20179 template <> struct A<int> @{ @}; // @r{OK to specialize}
20181 template <class T> void f (A<T>);
20186 f (std::A<float>()); // @r{lookup finds} std::f
20192 @section Type Traits
20194 The C++ front end implements syntactic extensions that allow
20195 compile-time determination of
20196 various characteristics of a type (or of a
20200 @item __has_nothrow_assign (type)
20201 If @code{type} is const qualified or is a reference type then the trait is
20202 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
20203 is true, else if @code{type} is a cv class or union type with copy assignment
20204 operators that are known not to throw an exception then the trait is true,
20205 else it is false. Requires: @code{type} shall be a complete type,
20206 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20208 @item __has_nothrow_copy (type)
20209 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
20210 @code{type} is a cv class or union type with copy constructors that
20211 are known not to throw an exception then the trait is true, else it is false.
20212 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
20213 @code{void}, or an array of unknown bound.
20215 @item __has_nothrow_constructor (type)
20216 If @code{__has_trivial_constructor (type)} is true then the trait is
20217 true, else if @code{type} is a cv class or union type (or array
20218 thereof) with a default constructor that is known not to throw an
20219 exception then the trait is true, else it is false. Requires:
20220 @code{type} shall be a complete type, (possibly cv-qualified)
20221 @code{void}, or an array of unknown bound.
20223 @item __has_trivial_assign (type)
20224 If @code{type} is const qualified or is a reference type then the trait is
20225 false. Otherwise if @code{__is_pod (type)} is true then the trait is
20226 true, else if @code{type} is a cv class or union type with a trivial
20227 copy assignment ([class.copy]) then the trait is true, else it is
20228 false. Requires: @code{type} shall be a complete type, (possibly
20229 cv-qualified) @code{void}, or an array of unknown bound.
20231 @item __has_trivial_copy (type)
20232 If @code{__is_pod (type)} is true or @code{type} is a reference type
20233 then the trait is true, else if @code{type} is a cv class or union type
20234 with a trivial copy constructor ([class.copy]) then the trait
20235 is true, else it is false. Requires: @code{type} shall be a complete
20236 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20238 @item __has_trivial_constructor (type)
20239 If @code{__is_pod (type)} is true then the trait is true, else if
20240 @code{type} is a cv class or union type (or array thereof) with a
20241 trivial default constructor ([class.ctor]) then the trait is true,
20242 else it is false. Requires: @code{type} shall be a complete
20243 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20245 @item __has_trivial_destructor (type)
20246 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20247 the trait is true, else if @code{type} is a cv class or union type (or
20248 array thereof) with a trivial destructor ([class.dtor]) then the trait
20249 is true, else it is false. Requires: @code{type} shall be a complete
20250 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20252 @item __has_virtual_destructor (type)
20253 If @code{type} is a class type with a virtual destructor
20254 ([class.dtor]) then the trait is true, else it is false. Requires:
20255 @code{type} shall be a complete type, (possibly cv-qualified)
20256 @code{void}, or an array of unknown bound.
20258 @item __is_abstract (type)
20259 If @code{type} is an abstract class ([class.abstract]) then the trait
20260 is true, else it is false. Requires: @code{type} shall be a complete
20261 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20263 @item __is_base_of (base_type, derived_type)
20264 If @code{base_type} is a base class of @code{derived_type}
20265 ([class.derived]) then the trait is true, otherwise it is false.
20266 Top-level cv qualifications of @code{base_type} and
20267 @code{derived_type} are ignored. For the purposes of this trait, a
20268 class type is considered is own base. Requires: if @code{__is_class
20269 (base_type)} and @code{__is_class (derived_type)} are true and
20270 @code{base_type} and @code{derived_type} are not the same type
20271 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20272 type. Diagnostic is produced if this requirement is not met.
20274 @item __is_class (type)
20275 If @code{type} is a cv class type, and not a union type
20276 ([basic.compound]) the trait is true, else it is false.
20278 @item __is_empty (type)
20279 If @code{__is_class (type)} is false then the trait is false.
20280 Otherwise @code{type} is considered empty if and only if: @code{type}
20281 has no non-static data members, or all non-static data members, if
20282 any, are bit-fields of length 0, and @code{type} has no virtual
20283 members, and @code{type} has no virtual base classes, and @code{type}
20284 has no base classes @code{base_type} for which
20285 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20286 be a complete type, (possibly cv-qualified) @code{void}, or an array
20289 @item __is_enum (type)
20290 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20291 true, else it is false.
20293 @item __is_literal_type (type)
20294 If @code{type} is a literal type ([basic.types]) the trait is
20295 true, else it is false. Requires: @code{type} shall be a complete type,
20296 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20298 @item __is_pod (type)
20299 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20300 else it is false. Requires: @code{type} shall be a complete type,
20301 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20303 @item __is_polymorphic (type)
20304 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20305 is true, else it is false. Requires: @code{type} shall be a complete
20306 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20308 @item __is_standard_layout (type)
20309 If @code{type} is a standard-layout type ([basic.types]) the trait is
20310 true, else it is false. Requires: @code{type} shall be a complete
20311 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20313 @item __is_trivial (type)
20314 If @code{type} is a trivial type ([basic.types]) the trait is
20315 true, else it is false. Requires: @code{type} shall be a complete
20316 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20318 @item __is_union (type)
20319 If @code{type} is a cv union type ([basic.compound]) the trait is
20320 true, else it is false.
20322 @item __underlying_type (type)
20323 The underlying type of @code{type}. Requires: @code{type} shall be
20324 an enumeration type ([dcl.enum]).
20330 @section C++ Concepts
20332 C++ concepts provide much-improved support for generic programming. In
20333 particular, they allow the specification of constraints on template arguments.
20334 The constraints are used to extend the usual overloading and partial
20335 specialization capabilities of the language, allowing generic data structures
20336 and algorithms to be ``refined'' based on their properties rather than their
20339 The following keywords are reserved for concepts.
20343 States an expression as an assumption, and if possible, verifies that the
20344 assumption is valid. For example, @code{assume(n > 0)}.
20347 Introduces an axiom definition. Axioms introduce requirements on values.
20350 Introduces a universally quantified object in an axiom. For example,
20351 @code{forall (int n) n + 0 == n}).
20354 Introduces a concept definition. Concepts are sets of syntactic and semantic
20355 requirements on types and their values.
20358 Introduces constraints on template arguments or requirements for a member
20359 function of a class template.
20363 The front end also exposes a number of internal mechanism that can be used
20364 to simplify the writing of type traits. Note that some of these traits are
20365 likely to be removed in the future.
20368 @item __is_same (type1, type2)
20369 A binary type trait: true whenever the type arguments are the same.
20374 @node Java Exceptions
20375 @section Java Exceptions
20377 The Java language uses a slightly different exception handling model
20378 from C++. Normally, GNU C++ automatically detects when you are
20379 writing C++ code that uses Java exceptions, and handle them
20380 appropriately. However, if C++ code only needs to execute destructors
20381 when Java exceptions are thrown through it, GCC guesses incorrectly.
20382 Sample problematic code is:
20385 struct S @{ ~S(); @};
20386 extern void bar(); // @r{is written in Java, and may throw exceptions}
20395 The usual effect of an incorrect guess is a link failure, complaining of
20396 a missing routine called @samp{__gxx_personality_v0}.
20398 You can inform the compiler that Java exceptions are to be used in a
20399 translation unit, irrespective of what it might think, by writing
20400 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20401 @samp{#pragma} must appear before any functions that throw or catch
20402 exceptions, or run destructors when exceptions are thrown through them.
20404 You cannot mix Java and C++ exceptions in the same translation unit. It
20405 is believed to be safe to throw a C++ exception from one file through
20406 another file compiled for the Java exception model, or vice versa, but
20407 there may be bugs in this area.
20409 @node Deprecated Features
20410 @section Deprecated Features
20412 In the past, the GNU C++ compiler was extended to experiment with new
20413 features, at a time when the C++ language was still evolving. Now that
20414 the C++ standard is complete, some of those features are superseded by
20415 superior alternatives. Using the old features might cause a warning in
20416 some cases that the feature will be dropped in the future. In other
20417 cases, the feature might be gone already.
20419 While the list below is not exhaustive, it documents some of the options
20420 that are now deprecated:
20423 @item -fexternal-templates
20424 @itemx -falt-external-templates
20425 These are two of the many ways for G++ to implement template
20426 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20427 defines how template definitions have to be organized across
20428 implementation units. G++ has an implicit instantiation mechanism that
20429 should work just fine for standard-conforming code.
20431 @item -fstrict-prototype
20432 @itemx -fno-strict-prototype
20433 Previously it was possible to use an empty prototype parameter list to
20434 indicate an unspecified number of parameters (like C), rather than no
20435 parameters, as C++ demands. This feature has been removed, except where
20436 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20439 G++ allows a virtual function returning @samp{void *} to be overridden
20440 by one returning a different pointer type. This extension to the
20441 covariant return type rules is now deprecated and will be removed from a
20444 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20445 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20446 and are now removed from G++. Code using these operators should be
20447 modified to use @code{std::min} and @code{std::max} instead.
20449 The named return value extension has been deprecated, and is now
20452 The use of initializer lists with new expressions has been deprecated,
20453 and is now removed from G++.
20455 Floating and complex non-type template parameters have been deprecated,
20456 and are now removed from G++.
20458 The implicit typename extension has been deprecated and is now
20461 The use of default arguments in function pointers, function typedefs
20462 and other places where they are not permitted by the standard is
20463 deprecated and will be removed from a future version of G++.
20465 G++ allows floating-point literals to appear in integral constant expressions,
20466 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20467 This extension is deprecated and will be removed from a future version.
20469 G++ allows static data members of const floating-point type to be declared
20470 with an initializer in a class definition. The standard only allows
20471 initializers for static members of const integral types and const
20472 enumeration types so this extension has been deprecated and will be removed
20473 from a future version.
20475 @node Backwards Compatibility
20476 @section Backwards Compatibility
20477 @cindex Backwards Compatibility
20478 @cindex ARM [Annotated C++ Reference Manual]
20480 Now that there is a definitive ISO standard C++, G++ has a specification
20481 to adhere to. The C++ language evolved over time, and features that
20482 used to be acceptable in previous drafts of the standard, such as the ARM
20483 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20484 compilation of C++ written to such drafts, G++ contains some backwards
20485 compatibilities. @emph{All such backwards compatibility features are
20486 liable to disappear in future versions of G++.} They should be considered
20487 deprecated. @xref{Deprecated Features}.
20491 If a variable is declared at for scope, it used to remain in scope until
20492 the end of the scope that contained the for statement (rather than just
20493 within the for scope). G++ retains this, but issues a warning, if such a
20494 variable is accessed outside the for scope.
20496 @item Implicit C language
20497 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20498 scope to set the language. On such systems, all header files are
20499 implicitly scoped inside a C language scope. Also, an empty prototype
20500 @code{()} is treated as an unspecified number of arguments, rather
20501 than no arguments, as C++ demands.
20504 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20505 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr