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.
3146 @item target (@var{options})
3147 @cindex @code{target} function attribute
3148 Multiple target back ends implement the @code{target} attribute
3149 to specify that a function is to
3150 be compiled with different target options than specified on the
3151 command line. This can be used for instance to have functions
3152 compiled with a different ISA (instruction set architecture) than the
3153 default. You can also use the @samp{#pragma GCC target} pragma to set
3154 more than one function to be compiled with specific target options.
3155 @xref{Function Specific Option Pragmas}, for details about the
3156 @samp{#pragma GCC target} pragma.
3158 For instance, on an x86, you could declare one function with the
3159 @code{target("sse4.1,arch=core2")} attribute and another with
3160 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3161 compiling the first function with @option{-msse4.1} and
3162 @option{-march=core2} options, and the second function with
3163 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3164 to make sure that a function is only invoked on a machine that
3165 supports the particular ISA it is compiled for (for example by using
3166 @code{cpuid} on x86 to determine what feature bits and architecture
3170 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3171 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3174 You can either use multiple
3175 strings separated by commas to specify multiple options,
3176 or separate the options with a comma (@samp{,}) within a single string.
3178 The options supported are specific to each target; refer to @ref{x86
3179 Function Attributes}, @ref{PowerPC Function Attributes},
3180 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3184 @cindex @code{unused} function attribute
3185 This attribute, attached to a function, means that the function is meant
3186 to be possibly unused. GCC does not produce a warning for this
3190 @cindex @code{used} function attribute
3191 This attribute, attached to a function, means that code must be emitted
3192 for the function even if it appears that the function is not referenced.
3193 This is useful, for example, when the function is referenced only in
3196 When applied to a member function of a C++ class template, the
3197 attribute also means that the function is instantiated if the
3198 class itself is instantiated.
3200 @item visibility ("@var{visibility_type}")
3201 @cindex @code{visibility} function attribute
3202 This attribute affects the linkage of the declaration to which it is attached.
3203 There are four supported @var{visibility_type} values: default,
3204 hidden, protected or internal visibility.
3207 void __attribute__ ((visibility ("protected")))
3208 f () @{ /* @r{Do something.} */; @}
3209 int i __attribute__ ((visibility ("hidden")));
3212 The possible values of @var{visibility_type} correspond to the
3213 visibility settings in the ELF gABI.
3216 @c keep this list of visibilities in alphabetical order.
3219 Default visibility is the normal case for the object file format.
3220 This value is available for the visibility attribute to override other
3221 options that may change the assumed visibility of entities.
3223 On ELF, default visibility means that the declaration is visible to other
3224 modules and, in shared libraries, means that the declared entity may be
3227 On Darwin, default visibility means that the declaration is visible to
3230 Default visibility corresponds to ``external linkage'' in the language.
3233 Hidden visibility indicates that the entity declared has a new
3234 form of linkage, which we call ``hidden linkage''. Two
3235 declarations of an object with hidden linkage refer to the same object
3236 if they are in the same shared object.
3239 Internal visibility is like hidden visibility, but with additional
3240 processor specific semantics. Unless otherwise specified by the
3241 psABI, GCC defines internal visibility to mean that a function is
3242 @emph{never} called from another module. Compare this with hidden
3243 functions which, while they cannot be referenced directly by other
3244 modules, can be referenced indirectly via function pointers. By
3245 indicating that a function cannot be called from outside the module,
3246 GCC may for instance omit the load of a PIC register since it is known
3247 that the calling function loaded the correct value.
3250 Protected visibility is like default visibility except that it
3251 indicates that references within the defining module bind to the
3252 definition in that module. That is, the declared entity cannot be
3253 overridden by another module.
3257 All visibilities are supported on many, but not all, ELF targets
3258 (supported when the assembler supports the @samp{.visibility}
3259 pseudo-op). Default visibility is supported everywhere. Hidden
3260 visibility is supported on Darwin targets.
3262 The visibility attribute should be applied only to declarations that
3263 would otherwise have external linkage. The attribute should be applied
3264 consistently, so that the same entity should not be declared with
3265 different settings of the attribute.
3267 In C++, the visibility attribute applies to types as well as functions
3268 and objects, because in C++ types have linkage. A class must not have
3269 greater visibility than its non-static data member types and bases,
3270 and class members default to the visibility of their class. Also, a
3271 declaration without explicit visibility is limited to the visibility
3274 In C++, you can mark member functions and static member variables of a
3275 class with the visibility attribute. This is useful if you know a
3276 particular method or static member variable should only be used from
3277 one shared object; then you can mark it hidden while the rest of the
3278 class has default visibility. Care must be taken to avoid breaking
3279 the One Definition Rule; for example, it is usually not useful to mark
3280 an inline method as hidden without marking the whole class as hidden.
3282 A C++ namespace declaration can also have the visibility attribute.
3285 namespace nspace1 __attribute__ ((visibility ("protected")))
3286 @{ /* @r{Do something.} */; @}
3289 This attribute applies only to the particular namespace body, not to
3290 other definitions of the same namespace; it is equivalent to using
3291 @samp{#pragma GCC visibility} before and after the namespace
3292 definition (@pxref{Visibility Pragmas}).
3294 In C++, if a template argument has limited visibility, this
3295 restriction is implicitly propagated to the template instantiation.
3296 Otherwise, template instantiations and specializations default to the
3297 visibility of their template.
3299 If both the template and enclosing class have explicit visibility, the
3300 visibility from the template is used.
3302 @item warn_unused_result
3303 @cindex @code{warn_unused_result} function attribute
3304 The @code{warn_unused_result} attribute causes a warning to be emitted
3305 if a caller of the function with this attribute does not use its
3306 return value. This is useful for functions where not checking
3307 the result is either a security problem or always a bug, such as
3311 int fn () __attribute__ ((warn_unused_result));
3314 if (fn () < 0) return -1;
3321 results in warning on line 5.
3324 @cindex @code{weak} function attribute
3325 The @code{weak} attribute causes the declaration to be emitted as a weak
3326 symbol rather than a global. This is primarily useful in defining
3327 library functions that can be overridden in user code, though it can
3328 also be used with non-function declarations. Weak symbols are supported
3329 for ELF targets, and also for a.out targets when using the GNU assembler
3333 @itemx weakref ("@var{target}")
3334 @cindex @code{weakref} function attribute
3335 The @code{weakref} attribute marks a declaration as a weak reference.
3336 Without arguments, it should be accompanied by an @code{alias} attribute
3337 naming the target symbol. Optionally, the @var{target} may be given as
3338 an argument to @code{weakref} itself. In either case, @code{weakref}
3339 implicitly marks the declaration as @code{weak}. Without a
3340 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3341 @code{weakref} is equivalent to @code{weak}.
3344 static int x() __attribute__ ((weakref ("y")));
3345 /* is equivalent to... */
3346 static int x() __attribute__ ((weak, weakref, alias ("y")));
3348 static int x() __attribute__ ((weakref));
3349 static int x() __attribute__ ((alias ("y")));
3352 A weak reference is an alias that does not by itself require a
3353 definition to be given for the target symbol. If the target symbol is
3354 only referenced through weak references, then it becomes a @code{weak}
3355 undefined symbol. If it is directly referenced, however, then such
3356 strong references prevail, and a definition is required for the
3357 symbol, not necessarily in the same translation unit.
3359 The effect is equivalent to moving all references to the alias to a
3360 separate translation unit, renaming the alias to the aliased symbol,
3361 declaring it as weak, compiling the two separate translation units and
3362 performing a reloadable link on them.
3364 At present, a declaration to which @code{weakref} is attached can
3365 only be @code{static}.
3370 @cindex lower memory region on the MSP430
3371 @cindex upper memory region on the MSP430
3372 @cindex either memory region on the MSP430
3373 On the MSP430 target these attributes can be used to specify whether
3374 the function or variable should be placed into low memory, high
3375 memory, or the placement should be left to the linker to decide. The
3376 attributes are only significant if compiling for the MSP430X
3379 The attributes work in conjunction with a linker script that has been
3380 augmented to specify where to place sections with a @code{.lower} and
3381 a @code{.upper} prefix. So for example as well as placing the
3382 @code{.data} section the script would also specify the placement of a
3383 @code{.lower.data} and a @code{.upper.data} section. The intention
3384 being that @code{lower} sections are placed into a small but easier to
3385 access memory region and the upper sections are placed into a larger, but
3386 slower to access region.
3388 The @code{either} attribute is special. It tells the linker to place
3389 the object into the corresponding @code{lower} section if there is
3390 room for it. If there is insufficient room then the object is placed
3391 into the corresponding @code{upper} section instead. Note - the
3392 placement algorithm is not very sophisticated. It will not attempt to
3393 find an optimal packing of the @code{lower} sections. It just makes
3394 one pass over the objects and does the best that it can. Using the
3395 @option{-ffunction-sections} and @option{-fdata-sections} command line
3396 options can help the packing however, since they produce smaller,
3397 easier to pack regions.
3400 On the MSP430 a function can be given the @code{reentant} attribute.
3401 This makes the function disable interrupts upon entry and enable
3402 interrupts upon exit. Reentrant functions cannot be @code{naked}.
3405 On the MSP430 a function can be given the @code{critical} attribute.
3406 This makes the function disable interrupts upon entry and restore the
3407 previous interrupt enabled/disabled state upon exit. A function
3408 cannot have both the @code{reentrant} and @code{critical} attributes.
3409 Critical functions cannot be @code{naked}.
3412 On the MSP430 a function can be given the @code{wakeup} attribute.
3413 Such a function must also have the @code{interrupt} attribute. When a
3414 function with the @code{wakeup} attribute exists the processor will be
3415 woken up from any low-power state in which it may be residing.
3419 @c This is the end of the target-independent attribute table
3421 @node AArch64 Function Attributes
3422 @subsection AArch64 Function Attributes
3424 The following target-specific function attributes are available for the
3425 AArch64 target. For the most part, these options mirror the behavior of
3426 similar command-line options (@pxref{AArch64 Options}), but on a
3430 @item general-regs-only
3431 @cindex @code{general-regs-only} function attribute, AArch64
3432 Indicates that no floating-point or Advanced SIMD registers should be
3433 used when generating code for this function. If the function explicitly
3434 uses floating-point code, then the compiler gives an error. This is
3435 the same behavior as that of the command-line option
3436 @option{-mgeneral-regs-only}.
3438 @item fix-cortex-a53-835769
3439 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3440 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3441 applied to this function. To explicitly disable the workaround for this
3442 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3443 This corresponds to the behavior of the command line options
3444 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3447 @cindex @code{cmodel=} function attribute, AArch64
3448 Indicates that code should be generated for a particular code model for
3449 this function. The behavior and permissible arguments are the same as
3450 for the command line option @option{-mcmodel=}.
3453 @cindex @code{strict-align} function attribute, AArch64
3454 Indicates that the compiler should not assume that unaligned memory references
3455 are handled by the system. The behavior is the same as for the command-line
3456 option @option{-mstrict-align}.
3458 @item omit-leaf-frame-pointer
3459 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3460 Indicates that the frame pointer should be omitted for a leaf function call.
3461 To keep the frame pointer, the inverse attribute
3462 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3463 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3464 and @option{-mno-omit-leaf-frame-pointer}.
3467 @cindex @code{tls-dialect=} function attribute, AArch64
3468 Specifies the TLS dialect to use for this function. The behavior and
3469 permissible arguments are the same as for the command-line option
3470 @option{-mtls-dialect=}.
3473 @cindex @code{arch=} function attribute, AArch64
3474 Specifies the architecture version and architectural extensions to use
3475 for this function. The behavior and permissible arguments are the same as
3476 for the @option{-march=} command-line option.
3479 @cindex @code{tune=} function attribute, AArch64
3480 Specifies the core for which to tune the performance of this function.
3481 The behavior and permissible arguments are the same as for the @option{-mtune=}
3482 command-line option.
3485 @cindex @code{cpu=} function attribute, AArch64
3486 Specifies the core for which to tune the performance of this function and also
3487 whose architectural features to use. The behavior and valid arguments are the
3488 same as for the @option{-mcpu=} command-line option.
3492 The above target attributes can be specified as follows:
3495 __attribute__((target("@var{attr-string}")))
3503 where @code{@var{attr-string}} is one of the attribute strings specified above.
3505 Additionally, the architectural extension string may be specified on its
3506 own. This can be used to turn on and off particular architectural extensions
3507 without having to specify a particular architecture version or core. Example:
3510 __attribute__((target("+crc+nocrypto")))
3518 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3519 extension and disables the @code{crypto} extension for the function @code{foo}
3520 without modifying an existing @option{-march=} or @option{-mcpu} option.
3522 Multiple target function attributes can be specified by separating them with
3523 a comma. For example:
3525 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3533 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3534 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3536 @subsubsection Inlining rules
3537 Specifying target attributes on individual functions or performing link-time
3538 optimization across translation units compiled with different target options
3539 can affect function inlining rules:
3541 In particular, a caller function can inline a callee function only if the
3542 architectural features available to the callee are a subset of the features
3543 available to the caller.
3544 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3545 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3546 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3547 because the all the architectural features that function @code{bar} requires
3548 are available to function @code{foo}. Conversely, function @code{bar} cannot
3549 inline function @code{foo}.
3551 Additionally inlining a function compiled with @option{-mstrict-align} into a
3552 function compiled without @code{-mstrict-align} is not allowed.
3553 However, inlining a function compiled without @option{-mstrict-align} into a
3554 function compiled with @option{-mstrict-align} is allowed.
3556 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3557 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3558 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3559 architectural feature rules specified above.
3561 @node ARC Function Attributes
3562 @subsection ARC Function Attributes
3564 These function attributes are supported by the ARC back end:
3568 @cindex @code{interrupt} function attribute, ARC
3569 Use this attribute to indicate
3570 that the specified function is an interrupt handler. The compiler generates
3571 function entry and exit sequences suitable for use in an interrupt handler
3572 when this attribute is present.
3574 On the ARC, you must specify the kind of interrupt to be handled
3575 in a parameter to the interrupt attribute like this:
3578 void f () __attribute__ ((interrupt ("ilink1")));
3581 Permissible values for this parameter are: @w{@code{ilink1}} and
3587 @cindex @code{long_call} function attribute, ARC
3588 @cindex @code{medium_call} function attribute, ARC
3589 @cindex @code{short_call} function attribute, ARC
3590 @cindex indirect calls, ARC
3591 These attributes specify how a particular function is called.
3592 These attributes override the
3593 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3594 command-line switches and @code{#pragma long_calls} settings.
3596 For ARC, a function marked with the @code{long_call} attribute is
3597 always called using register-indirect jump-and-link instructions,
3598 thereby enabling the called function to be placed anywhere within the
3599 32-bit address space. A function marked with the @code{medium_call}
3600 attribute will always be close enough to be called with an unconditional
3601 branch-and-link instruction, which has a 25-bit offset from
3602 the call site. A function marked with the @code{short_call}
3603 attribute will always be close enough to be called with a conditional
3604 branch-and-link instruction, which has a 21-bit offset from
3608 @node ARM Function Attributes
3609 @subsection ARM Function Attributes
3611 These function attributes are supported for ARM targets:
3615 @cindex @code{interrupt} function attribute, ARM
3616 Use this attribute to indicate
3617 that the specified function is an interrupt handler. The compiler generates
3618 function entry and exit sequences suitable for use in an interrupt handler
3619 when this attribute is present.
3621 You can specify the kind of interrupt to be handled by
3622 adding an optional parameter to the interrupt attribute like this:
3625 void f () __attribute__ ((interrupt ("IRQ")));
3629 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3630 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3632 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3633 may be called with a word-aligned stack pointer.
3636 @cindex @code{isr} function attribute, ARM
3637 Use this attribute on ARM to write Interrupt Service Routines. This is an
3638 alias to the @code{interrupt} attribute above.
3642 @cindex @code{long_call} function attribute, ARM
3643 @cindex @code{short_call} function attribute, ARM
3644 @cindex indirect calls, ARM
3645 These attributes specify how a particular function is called.
3646 These attributes override the
3647 @option{-mlong-calls} (@pxref{ARM Options})
3648 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3649 @code{long_call} attribute indicates that the function might be far
3650 away from the call site and require a different (more expensive)
3651 calling sequence. The @code{short_call} attribute always places
3652 the offset to the function from the call site into the @samp{BL}
3653 instruction directly.
3656 @cindex @code{naked} function attribute, ARM
3657 This attribute allows the compiler to construct the
3658 requisite function declaration, while allowing the body of the
3659 function to be assembly code. The specified function will not have
3660 prologue/epilogue sequences generated by the compiler. Only basic
3661 @code{asm} statements can safely be included in naked functions
3662 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3663 basic @code{asm} and C code may appear to work, they cannot be
3664 depended upon to work reliably and are not supported.
3667 @cindex @code{pcs} function attribute, ARM
3669 The @code{pcs} attribute can be used to control the calling convention
3670 used for a function on ARM. The attribute takes an argument that specifies
3671 the calling convention to use.
3673 When compiling using the AAPCS ABI (or a variant of it) then valid
3674 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3675 order to use a variant other than @code{"aapcs"} then the compiler must
3676 be permitted to use the appropriate co-processor registers (i.e., the
3677 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3681 /* Argument passed in r0, and result returned in r0+r1. */
3682 double f2d (float) __attribute__((pcs("aapcs")));
3685 Variadic functions always use the @code{"aapcs"} calling convention and
3686 the compiler rejects attempts to specify an alternative.
3688 @item target (@var{options})
3689 @cindex @code{target} function attribute
3690 As discussed in @ref{Common Function Attributes}, this attribute
3691 allows specification of target-specific compilation options.
3693 On ARM, the following options are allowed:
3697 @cindex @code{target("thumb")} function attribute, ARM
3698 Force code generation in the Thumb (T16/T32) ISA, depending on the
3702 @cindex @code{target("arm")} function attribute, ARM
3703 Force code generation in the ARM (A32) ISA.
3705 Functions from different modes can be inlined in the caller's mode.
3708 @cindex @code{target("fpu=")} function attribute, ARM
3709 Specifies the fpu for which to tune the performance of this function.
3710 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3711 command-line option.
3717 @node AVR Function Attributes
3718 @subsection AVR Function Attributes
3720 These function attributes are supported by the AVR back end:
3724 @cindex @code{interrupt} function attribute, AVR
3725 Use this attribute to indicate
3726 that the specified function is an interrupt handler. The compiler generates
3727 function entry and exit sequences suitable for use in an interrupt handler
3728 when this attribute is present.
3730 On the AVR, the hardware globally disables interrupts when an
3731 interrupt is executed. The first instruction of an interrupt handler
3732 declared with this attribute is a @code{SEI} instruction to
3733 re-enable interrupts. See also the @code{signal} function attribute
3734 that does not insert a @code{SEI} instruction. If both @code{signal} and
3735 @code{interrupt} are specified for the same function, @code{signal}
3736 is silently ignored.
3739 @cindex @code{naked} function attribute, AVR
3740 This attribute allows the compiler to construct the
3741 requisite function declaration, while allowing the body of the
3742 function to be assembly code. The specified function will not have
3743 prologue/epilogue sequences generated by the compiler. Only basic
3744 @code{asm} statements can safely be included in naked functions
3745 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3746 basic @code{asm} and C code may appear to work, they cannot be
3747 depended upon to work reliably and are not supported.
3751 @cindex @code{OS_main} function attribute, AVR
3752 @cindex @code{OS_task} function attribute, AVR
3753 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3754 do not save/restore any call-saved register in their prologue/epilogue.
3756 The @code{OS_main} attribute can be used when there @emph{is
3757 guarantee} that interrupts are disabled at the time when the function
3758 is entered. This saves resources when the stack pointer has to be
3759 changed to set up a frame for local variables.
3761 The @code{OS_task} attribute can be used when there is @emph{no
3762 guarantee} that interrupts are disabled at that time when the function
3763 is entered like for, e@.g@. task functions in a multi-threading operating
3764 system. In that case, changing the stack pointer register is
3765 guarded by save/clear/restore of the global interrupt enable flag.
3767 The differences to the @code{naked} function attribute are:
3769 @item @code{naked} functions do not have a return instruction whereas
3770 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3771 @code{RETI} return instruction.
3772 @item @code{naked} functions do not set up a frame for local variables
3773 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3778 @cindex @code{signal} function attribute, AVR
3779 Use this attribute on the AVR to indicate that the specified
3780 function is an interrupt handler. The compiler generates function
3781 entry and exit sequences suitable for use in an interrupt handler when this
3782 attribute is present.
3784 See also the @code{interrupt} function attribute.
3786 The AVR hardware globally disables interrupts when an interrupt is executed.
3787 Interrupt handler functions defined with the @code{signal} attribute
3788 do not re-enable interrupts. It is save to enable interrupts in a
3789 @code{signal} handler. This ``save'' only applies to the code
3790 generated by the compiler and not to the IRQ layout of the
3791 application which is responsibility of the application.
3793 If both @code{signal} and @code{interrupt} are specified for the same
3794 function, @code{signal} is silently ignored.
3797 @node Blackfin Function Attributes
3798 @subsection Blackfin Function Attributes
3800 These function attributes are supported by the Blackfin back end:
3804 @item exception_handler
3805 @cindex @code{exception_handler} function attribute
3806 @cindex exception handler functions, Blackfin
3807 Use this attribute on the Blackfin to indicate that the specified function
3808 is an exception handler. The compiler generates function entry and
3809 exit sequences suitable for use in an exception handler when this
3810 attribute is present.
3812 @item interrupt_handler
3813 @cindex @code{interrupt_handler} function attribute, Blackfin
3814 Use this attribute to
3815 indicate that the specified function is an interrupt handler. The compiler
3816 generates function entry and exit sequences suitable for use in an
3817 interrupt handler when this attribute is present.
3820 @cindex @code{kspisusp} function attribute, Blackfin
3821 @cindex User stack pointer in interrupts on the Blackfin
3822 When used together with @code{interrupt_handler}, @code{exception_handler}
3823 or @code{nmi_handler}, code is generated to load the stack pointer
3824 from the USP register in the function prologue.
3827 @cindex @code{l1_text} function attribute, Blackfin
3828 This attribute specifies a function to be placed into L1 Instruction
3829 SRAM@. The function is put into a specific section named @code{.l1.text}.
3830 With @option{-mfdpic}, function calls with a such function as the callee
3831 or caller uses inlined PLT.
3834 @cindex @code{l2} function attribute, Blackfin
3835 This attribute specifies a function to be placed into L2
3836 SRAM. The function is put into a specific section named
3837 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3842 @cindex indirect calls, Blackfin
3843 @cindex @code{longcall} function attribute, Blackfin
3844 @cindex @code{shortcall} function attribute, Blackfin
3845 The @code{longcall} attribute
3846 indicates that the function might be far away from the call site and
3847 require a different (more expensive) calling sequence. The
3848 @code{shortcall} attribute indicates that the function is always close
3849 enough for the shorter calling sequence to be used. These attributes
3850 override the @option{-mlongcall} switch.
3853 @cindex @code{nesting} function attribute, Blackfin
3854 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3855 Use this attribute together with @code{interrupt_handler},
3856 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3857 entry code should enable nested interrupts or exceptions.
3860 @cindex @code{nmi_handler} function attribute, Blackfin
3861 @cindex NMI handler functions on the Blackfin processor
3862 Use this attribute on the Blackfin to indicate that the specified function
3863 is an NMI handler. The compiler generates function entry and
3864 exit sequences suitable for use in an NMI handler when this
3865 attribute is present.
3868 @cindex @code{saveall} function attribute, Blackfin
3869 @cindex save all registers on the Blackfin
3870 Use this attribute to indicate that
3871 all registers except the stack pointer should be saved in the prologue
3872 regardless of whether they are used or not.
3875 @node CR16 Function Attributes
3876 @subsection CR16 Function Attributes
3878 These function attributes are supported by the CR16 back end:
3882 @cindex @code{interrupt} function attribute, CR16
3883 Use this attribute to indicate
3884 that the specified function is an interrupt handler. The compiler generates
3885 function entry and exit sequences suitable for use in an interrupt handler
3886 when this attribute is present.
3889 @node Epiphany Function Attributes
3890 @subsection Epiphany Function Attributes
3892 These function attributes are supported by the Epiphany back end:
3896 @cindex @code{disinterrupt} function attribute, Epiphany
3897 This attribute causes the compiler to emit
3898 instructions to disable interrupts for the duration of the given
3901 @item forwarder_section
3902 @cindex @code{forwarder_section} function attribute, Epiphany
3903 This attribute modifies the behavior of an interrupt handler.
3904 The interrupt handler may be in external memory which cannot be
3905 reached by a branch instruction, so generate a local memory trampoline
3906 to transfer control. The single parameter identifies the section where
3907 the trampoline is placed.
3910 @cindex @code{interrupt} function attribute, Epiphany
3911 Use this attribute to indicate
3912 that the specified function is an interrupt handler. The compiler generates
3913 function entry and exit sequences suitable for use in an interrupt handler
3914 when this attribute is present. It may also generate
3915 a special section with code to initialize the interrupt vector table.
3917 On Epiphany targets one or more optional parameters can be added like this:
3920 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3923 Permissible values for these parameters are: @w{@code{reset}},
3924 @w{@code{software_exception}}, @w{@code{page_miss}},
3925 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3926 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3927 Multiple parameters indicate that multiple entries in the interrupt
3928 vector table should be initialized for this function, i.e.@: for each
3929 parameter @w{@var{name}}, a jump to the function is emitted in
3930 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3931 entirely, in which case no interrupt vector table entry is provided.
3933 Note that interrupts are enabled inside the function
3934 unless the @code{disinterrupt} attribute is also specified.
3936 The following examples are all valid uses of these attributes on
3939 void __attribute__ ((interrupt)) universal_handler ();
3940 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3941 void __attribute__ ((interrupt ("dma0, dma1")))
3942 universal_dma_handler ();
3943 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3944 fast_timer_handler ();
3945 void __attribute__ ((interrupt ("dma0, dma1"),
3946 forwarder_section ("tramp")))
3947 external_dma_handler ();
3952 @cindex @code{long_call} function attribute, Epiphany
3953 @cindex @code{short_call} function attribute, Epiphany
3954 @cindex indirect calls, Epiphany
3955 These attributes specify how a particular function is called.
3956 These attributes override the
3957 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3958 command-line switch and @code{#pragma long_calls} settings.
3962 @node H8/300 Function Attributes
3963 @subsection H8/300 Function Attributes
3965 These function attributes are available for H8/300 targets:
3968 @item function_vector
3969 @cindex @code{function_vector} function attribute, H8/300
3970 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3971 that the specified function should be called through the function vector.
3972 Calling a function through the function vector reduces code size; however,
3973 the function vector has a limited size (maximum 128 entries on the H8/300
3974 and 64 entries on the H8/300H and H8S)
3975 and shares space with the interrupt vector.
3977 @item interrupt_handler
3978 @cindex @code{interrupt_handler} function attribute, H8/300
3979 Use this attribute on the H8/300, H8/300H, and H8S to
3980 indicate that the specified function is an interrupt handler. The compiler
3981 generates function entry and exit sequences suitable for use in an
3982 interrupt handler when this attribute is present.
3985 @cindex @code{saveall} function attribute, H8/300
3986 @cindex save all registers on the H8/300, H8/300H, and H8S
3987 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3988 all registers except the stack pointer should be saved in the prologue
3989 regardless of whether they are used or not.
3992 @node IA-64 Function Attributes
3993 @subsection IA-64 Function Attributes
3995 These function attributes are supported on IA-64 targets:
3998 @item syscall_linkage
3999 @cindex @code{syscall_linkage} function attribute, IA-64
4000 This attribute is used to modify the IA-64 calling convention by marking
4001 all input registers as live at all function exits. This makes it possible
4002 to restart a system call after an interrupt without having to save/restore
4003 the input registers. This also prevents kernel data from leaking into
4007 @cindex @code{version_id} function attribute, IA-64
4008 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4009 symbol to contain a version string, thus allowing for function level
4010 versioning. HP-UX system header files may use function level versioning
4011 for some system calls.
4014 extern int foo () __attribute__((version_id ("20040821")));
4018 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4021 @node M32C Function Attributes
4022 @subsection M32C Function Attributes
4024 These function attributes are supported by the M32C back end:
4028 @cindex @code{bank_switch} function attribute, M32C
4029 When added to an interrupt handler with the M32C port, causes the
4030 prologue and epilogue to use bank switching to preserve the registers
4031 rather than saving them on the stack.
4033 @item fast_interrupt
4034 @cindex @code{fast_interrupt} function attribute, M32C
4035 Use this attribute on the M32C port to indicate that the specified
4036 function is a fast interrupt handler. This is just like the
4037 @code{interrupt} attribute, except that @code{freit} is used to return
4038 instead of @code{reit}.
4040 @item function_vector
4041 @cindex @code{function_vector} function attribute, M16C/M32C
4042 On M16C/M32C targets, the @code{function_vector} attribute declares a
4043 special page subroutine call function. Use of this attribute reduces
4044 the code size by 2 bytes for each call generated to the
4045 subroutine. The argument to the attribute is the vector number entry
4046 from the special page vector table which contains the 16 low-order
4047 bits of the subroutine's entry address. Each vector table has special
4048 page number (18 to 255) that is used in @code{jsrs} instructions.
4049 Jump addresses of the routines are generated by adding 0x0F0000 (in
4050 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4051 2-byte addresses set in the vector table. Therefore you need to ensure
4052 that all the special page vector routines should get mapped within the
4053 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4056 In the following example 2 bytes are saved for each call to
4057 function @code{foo}.
4060 void foo (void) __attribute__((function_vector(0x18)));
4071 If functions are defined in one file and are called in another file,
4072 then be sure to write this declaration in both files.
4074 This attribute is ignored for R8C target.
4077 @cindex @code{interrupt} function attribute, M32C
4078 Use this attribute to indicate
4079 that the specified function is an interrupt handler. The compiler generates
4080 function entry and exit sequences suitable for use in an interrupt handler
4081 when this attribute is present.
4084 @node M32R/D Function Attributes
4085 @subsection M32R/D Function Attributes
4087 These function attributes are supported by the M32R/D back end:
4091 @cindex @code{interrupt} function attribute, M32R/D
4092 Use this attribute to indicate
4093 that the specified function is an interrupt handler. The compiler generates
4094 function entry and exit sequences suitable for use in an interrupt handler
4095 when this attribute is present.
4097 @item model (@var{model-name})
4098 @cindex @code{model} function attribute, M32R/D
4099 @cindex function addressability on the M32R/D
4101 On the M32R/D, use this attribute to set the addressability of an
4102 object, and of the code generated for a function. The identifier
4103 @var{model-name} is one of @code{small}, @code{medium}, or
4104 @code{large}, representing each of the code models.
4106 Small model objects live in the lower 16MB of memory (so that their
4107 addresses can be loaded with the @code{ld24} instruction), and are
4108 callable with the @code{bl} instruction.
4110 Medium model objects may live anywhere in the 32-bit address space (the
4111 compiler generates @code{seth/add3} instructions to load their addresses),
4112 and are callable with the @code{bl} instruction.
4114 Large model objects may live anywhere in the 32-bit address space (the
4115 compiler generates @code{seth/add3} instructions to load their addresses),
4116 and may not be reachable with the @code{bl} instruction (the compiler
4117 generates the much slower @code{seth/add3/jl} instruction sequence).
4120 @node m68k Function Attributes
4121 @subsection m68k Function Attributes
4123 These function attributes are supported by the m68k back end:
4127 @itemx interrupt_handler
4128 @cindex @code{interrupt} function attribute, m68k
4129 @cindex @code{interrupt_handler} function attribute, m68k
4130 Use this attribute to
4131 indicate that the specified function is an interrupt handler. The compiler
4132 generates function entry and exit sequences suitable for use in an
4133 interrupt handler when this attribute is present. Either name may be used.
4135 @item interrupt_thread
4136 @cindex @code{interrupt_thread} function attribute, fido
4137 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4138 that the specified function is an interrupt handler that is designed
4139 to run as a thread. The compiler omits generate prologue/epilogue
4140 sequences and replaces the return instruction with a @code{sleep}
4141 instruction. This attribute is available only on fido.
4144 @node MCORE Function Attributes
4145 @subsection MCORE Function Attributes
4147 These function attributes are supported by the MCORE back end:
4151 @cindex @code{naked} function attribute, MCORE
4152 This attribute allows the compiler to construct the
4153 requisite function declaration, while allowing the body of the
4154 function to be assembly code. The specified function will not have
4155 prologue/epilogue sequences generated by the compiler. Only basic
4156 @code{asm} statements can safely be included in naked functions
4157 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4158 basic @code{asm} and C code may appear to work, they cannot be
4159 depended upon to work reliably and are not supported.
4162 @node MeP Function Attributes
4163 @subsection MeP Function Attributes
4165 These function attributes are supported by the MeP back end:
4169 @cindex @code{disinterrupt} function attribute, MeP
4170 On MeP targets, this attribute causes the compiler to emit
4171 instructions to disable interrupts for the duration of the given
4175 @cindex @code{interrupt} function attribute, MeP
4176 Use this attribute to indicate
4177 that the specified function is an interrupt handler. The compiler generates
4178 function entry and exit sequences suitable for use in an interrupt handler
4179 when this attribute is present.
4182 @cindex @code{near} function attribute, MeP
4183 This attribute causes the compiler to assume the called
4184 function is close enough to use the normal calling convention,
4185 overriding the @option{-mtf} command-line option.
4188 @cindex @code{far} function attribute, MeP
4189 On MeP targets this causes the compiler to use a calling convention
4190 that assumes the called function is too far away for the built-in
4194 @cindex @code{vliw} function attribute, MeP
4195 The @code{vliw} attribute tells the compiler to emit
4196 instructions in VLIW mode instead of core mode. Note that this
4197 attribute is not allowed unless a VLIW coprocessor has been configured
4198 and enabled through command-line options.
4201 @node MicroBlaze Function Attributes
4202 @subsection MicroBlaze Function Attributes
4204 These function attributes are supported on MicroBlaze targets:
4207 @item save_volatiles
4208 @cindex @code{save_volatiles} function attribute, MicroBlaze
4209 Use this attribute to indicate that the function is
4210 an interrupt handler. All volatile registers (in addition to non-volatile
4211 registers) are saved in the function prologue. If the function is a leaf
4212 function, only volatiles used by the function are saved. A normal function
4213 return is generated instead of a return from interrupt.
4216 @cindex @code{break_handler} function attribute, MicroBlaze
4217 @cindex break handler functions
4218 Use this attribute to indicate that
4219 the specified function is a break handler. The compiler generates function
4220 entry and exit sequences suitable for use in an break handler when this
4221 attribute is present. The return from @code{break_handler} is done through
4222 the @code{rtbd} instead of @code{rtsd}.
4225 void f () __attribute__ ((break_handler));
4229 @node Microsoft Windows Function Attributes
4230 @subsection Microsoft Windows Function Attributes
4232 The following attributes are available on Microsoft Windows and Symbian OS
4237 @cindex @code{dllexport} function attribute
4238 @cindex @code{__declspec(dllexport)}
4239 On Microsoft Windows targets and Symbian OS targets the
4240 @code{dllexport} attribute causes the compiler to provide a global
4241 pointer to a pointer in a DLL, so that it can be referenced with the
4242 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4243 name is formed by combining @code{_imp__} and the function or variable
4246 You can use @code{__declspec(dllexport)} as a synonym for
4247 @code{__attribute__ ((dllexport))} for compatibility with other
4250 On systems that support the @code{visibility} attribute, this
4251 attribute also implies ``default'' visibility. It is an error to
4252 explicitly specify any other visibility.
4254 GCC's default behavior is to emit all inline functions with the
4255 @code{dllexport} attribute. Since this can cause object file-size bloat,
4256 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4257 ignore the attribute for inlined functions unless the
4258 @option{-fkeep-inline-functions} flag is used instead.
4260 The attribute is ignored for undefined symbols.
4262 When applied to C++ classes, the attribute marks defined non-inlined
4263 member functions and static data members as exports. Static consts
4264 initialized in-class are not marked unless they are also defined
4267 For Microsoft Windows targets there are alternative methods for
4268 including the symbol in the DLL's export table such as using a
4269 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4270 the @option{--export-all} linker flag.
4273 @cindex @code{dllimport} function attribute
4274 @cindex @code{__declspec(dllimport)}
4275 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4276 attribute causes the compiler to reference a function or variable via
4277 a global pointer to a pointer that is set up by the DLL exporting the
4278 symbol. The attribute implies @code{extern}. On Microsoft Windows
4279 targets, the pointer name is formed by combining @code{_imp__} and the
4280 function or variable name.
4282 You can use @code{__declspec(dllimport)} as a synonym for
4283 @code{__attribute__ ((dllimport))} for compatibility with other
4286 On systems that support the @code{visibility} attribute, this
4287 attribute also implies ``default'' visibility. It is an error to
4288 explicitly specify any other visibility.
4290 Currently, the attribute is ignored for inlined functions. If the
4291 attribute is applied to a symbol @emph{definition}, an error is reported.
4292 If a symbol previously declared @code{dllimport} is later defined, the
4293 attribute is ignored in subsequent references, and a warning is emitted.
4294 The attribute is also overridden by a subsequent declaration as
4297 When applied to C++ classes, the attribute marks non-inlined
4298 member functions and static data members as imports. However, the
4299 attribute is ignored for virtual methods to allow creation of vtables
4302 On the SH Symbian OS target the @code{dllimport} attribute also has
4303 another affect---it can cause the vtable and run-time type information
4304 for a class to be exported. This happens when the class has a
4305 dllimported constructor or a non-inline, non-pure virtual function
4306 and, for either of those two conditions, the class also has an inline
4307 constructor or destructor and has a key function that is defined in
4308 the current translation unit.
4310 For Microsoft Windows targets the use of the @code{dllimport}
4311 attribute on functions is not necessary, but provides a small
4312 performance benefit by eliminating a thunk in the DLL@. The use of the
4313 @code{dllimport} attribute on imported variables can be avoided by passing the
4314 @option{--enable-auto-import} switch to the GNU linker. As with
4315 functions, using the attribute for a variable eliminates a thunk in
4318 One drawback to using this attribute is that a pointer to a
4319 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4320 address. However, a pointer to a @emph{function} with the
4321 @code{dllimport} attribute can be used as a constant initializer; in
4322 this case, the address of a stub function in the import lib is
4323 referenced. On Microsoft Windows targets, the attribute can be disabled
4324 for functions by setting the @option{-mnop-fun-dllimport} flag.
4327 @node MIPS Function Attributes
4328 @subsection MIPS Function Attributes
4330 These function attributes are supported by the MIPS back end:
4334 @cindex @code{interrupt} function attribute, MIPS
4335 Use this attribute to indicate that the specified function is an interrupt
4336 handler. The compiler generates function entry and exit sequences suitable
4337 for use in an interrupt handler when this attribute is present.
4338 An optional argument is supported for the interrupt attribute which allows
4339 the interrupt mode to be described. By default GCC assumes the external
4340 interrupt controller (EIC) mode is in use, this can be explicitly set using
4341 @code{eic}. When interrupts are non-masked then the requested Interrupt
4342 Priority Level (IPL) is copied to the current IPL which has the effect of only
4343 enabling higher priority interrupts. To use vectored interrupt mode use
4344 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4345 the behaviour of the non-masked interrupt support and GCC will arrange to mask
4346 all interrupts from sw0 up to and including the specified interrupt vector.
4348 You can use the following attributes to modify the behavior
4349 of an interrupt handler:
4351 @item use_shadow_register_set
4352 @cindex @code{use_shadow_register_set} function attribute, MIPS
4353 Assume that the handler uses a shadow register set, instead of
4354 the main general-purpose registers. An optional argument @code{intstack} is
4355 supported to indicate that the shadow register set contains a valid stack
4358 @item keep_interrupts_masked
4359 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4360 Keep interrupts masked for the whole function. Without this attribute,
4361 GCC tries to reenable interrupts for as much of the function as it can.
4363 @item use_debug_exception_return
4364 @cindex @code{use_debug_exception_return} function attribute, MIPS
4365 Return using the @code{deret} instruction. Interrupt handlers that don't
4366 have this attribute return using @code{eret} instead.
4369 You can use any combination of these attributes, as shown below:
4371 void __attribute__ ((interrupt)) v0 ();
4372 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4373 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4374 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4375 void __attribute__ ((interrupt, use_shadow_register_set,
4376 keep_interrupts_masked)) v4 ();
4377 void __attribute__ ((interrupt, use_shadow_register_set,
4378 use_debug_exception_return)) v5 ();
4379 void __attribute__ ((interrupt, keep_interrupts_masked,
4380 use_debug_exception_return)) v6 ();
4381 void __attribute__ ((interrupt, use_shadow_register_set,
4382 keep_interrupts_masked,
4383 use_debug_exception_return)) v7 ();
4384 void __attribute__ ((interrupt("eic"))) v8 ();
4385 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4391 @cindex indirect calls, MIPS
4392 @cindex @code{long_call} function attribute, MIPS
4393 @cindex @code{near} function attribute, MIPS
4394 @cindex @code{far} function attribute, MIPS
4395 These attributes specify how a particular function is called on MIPS@.
4396 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4397 command-line switch. The @code{long_call} and @code{far} attributes are
4398 synonyms, and cause the compiler to always call
4399 the function by first loading its address into a register, and then using
4400 the contents of that register. The @code{near} attribute has the opposite
4401 effect; it specifies that non-PIC calls should be made using the more
4402 efficient @code{jal} instruction.
4406 @cindex @code{mips16} function attribute, MIPS
4407 @cindex @code{nomips16} function attribute, MIPS
4409 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4410 function attributes to locally select or turn off MIPS16 code generation.
4411 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4412 while MIPS16 code generation is disabled for functions with the
4413 @code{nomips16} attribute. These attributes override the
4414 @option{-mips16} and @option{-mno-mips16} options on the command line
4415 (@pxref{MIPS Options}).
4417 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4418 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4419 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4420 may interact badly with some GCC extensions such as @code{__builtin_apply}
4421 (@pxref{Constructing Calls}).
4423 @item micromips, MIPS
4424 @itemx nomicromips, MIPS
4425 @cindex @code{micromips} function attribute
4426 @cindex @code{nomicromips} function attribute
4428 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4429 function attributes to locally select or turn off microMIPS code generation.
4430 A function with the @code{micromips} attribute is emitted as microMIPS code,
4431 while microMIPS code generation is disabled for functions with the
4432 @code{nomicromips} attribute. These attributes override the
4433 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4434 (@pxref{MIPS Options}).
4436 When compiling files containing mixed microMIPS and non-microMIPS code, the
4437 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4439 not that within individual functions. Mixed microMIPS and non-microMIPS code
4440 may interact badly with some GCC extensions such as @code{__builtin_apply}
4441 (@pxref{Constructing Calls}).
4444 @cindex @code{nocompression} function attribute, MIPS
4445 On MIPS targets, you can use the @code{nocompression} function attribute
4446 to locally turn off MIPS16 and microMIPS code generation. This attribute
4447 overrides the @option{-mips16} and @option{-mmicromips} options on the
4448 command line (@pxref{MIPS Options}).
4451 @node MSP430 Function Attributes
4452 @subsection MSP430 Function Attributes
4454 These function attributes are supported by the MSP430 back end:
4458 @cindex @code{critical} function attribute, MSP430
4459 Critical functions disable interrupts upon entry and restore the
4460 previous interrupt state upon exit. Critical functions cannot also
4461 have the @code{naked} or @code{reentrant} attributes. They can have
4462 the @code{interrupt} attribute.
4465 @cindex @code{interrupt} function attribute, MSP430
4466 Use this attribute to indicate
4467 that the specified function is an interrupt handler. The compiler generates
4468 function entry and exit sequences suitable for use in an interrupt handler
4469 when this attribute is present.
4471 You can provide an argument to the interrupt
4472 attribute which specifies a name or number. If the argument is a
4473 number it indicates the slot in the interrupt vector table (0 - 31) to
4474 which this handler should be assigned. If the argument is a name it
4475 is treated as a symbolic name for the vector slot. These names should
4476 match up with appropriate entries in the linker script. By default
4477 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4478 @code{reset} for vector 31 are recognized.
4481 @cindex @code{naked} function attribute, MSP430
4482 This attribute allows the compiler to construct the
4483 requisite function declaration, while allowing the body of the
4484 function to be assembly code. The specified function will not have
4485 prologue/epilogue sequences generated by the compiler. Only basic
4486 @code{asm} statements can safely be included in naked functions
4487 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4488 basic @code{asm} and C code may appear to work, they cannot be
4489 depended upon to work reliably and are not supported.
4492 @cindex @code{reentrant} function attribute, MSP430
4493 Reentrant functions disable interrupts upon entry and enable them
4494 upon exit. Reentrant functions cannot also have the @code{naked}
4495 or @code{critical} attributes. They can have the @code{interrupt}
4499 @cindex @code{wakeup} function attribute, MSP430
4500 This attribute only applies to interrupt functions. It is silently
4501 ignored if applied to a non-interrupt function. A wakeup interrupt
4502 function will rouse the processor from any low-power state that it
4503 might be in when the function exits.
4506 @node NDS32 Function Attributes
4507 @subsection NDS32 Function Attributes
4509 These function attributes are supported by the NDS32 back end:
4513 @cindex @code{exception} function attribute
4514 @cindex exception handler functions, NDS32
4515 Use this attribute on the NDS32 target to indicate that the specified function
4516 is an exception handler. The compiler will generate corresponding sections
4517 for use in an exception handler.
4520 @cindex @code{interrupt} function attribute, NDS32
4521 On NDS32 target, this attribute indicates that the specified function
4522 is an interrupt handler. The compiler generates corresponding sections
4523 for use in an interrupt handler. You can use the following attributes
4524 to modify the behavior:
4527 @cindex @code{nested} function attribute, NDS32
4528 This interrupt service routine is interruptible.
4530 @cindex @code{not_nested} function attribute, NDS32
4531 This interrupt service routine is not interruptible.
4533 @cindex @code{nested_ready} function attribute, NDS32
4534 This interrupt service routine is interruptible after @code{PSW.GIE}
4535 (global interrupt enable) is set. This allows interrupt service routine to
4536 finish some short critical code before enabling interrupts.
4538 @cindex @code{save_all} function attribute, NDS32
4539 The system will help save all registers into stack before entering
4542 @cindex @code{partial_save} function attribute, NDS32
4543 The system will help save caller registers into stack before entering
4548 @cindex @code{naked} function attribute, NDS32
4549 This attribute allows the compiler to construct the
4550 requisite function declaration, while allowing the body of the
4551 function to be assembly code. The specified function will not have
4552 prologue/epilogue sequences generated by the compiler. Only basic
4553 @code{asm} statements can safely be included in naked functions
4554 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4555 basic @code{asm} and C code may appear to work, they cannot be
4556 depended upon to work reliably and are not supported.
4559 @cindex @code{reset} function attribute, NDS32
4560 @cindex reset handler functions
4561 Use this attribute on the NDS32 target to indicate that the specified function
4562 is a reset handler. The compiler will generate corresponding sections
4563 for use in a reset handler. You can use the following attributes
4564 to provide extra exception handling:
4567 @cindex @code{nmi} function attribute, NDS32
4568 Provide a user-defined function to handle NMI exception.
4570 @cindex @code{warm} function attribute, NDS32
4571 Provide a user-defined function to handle warm reset exception.
4575 @node Nios II Function Attributes
4576 @subsection Nios II Function Attributes
4578 These function attributes are supported by the Nios II back end:
4581 @item target (@var{options})
4582 @cindex @code{target} function attribute
4583 As discussed in @ref{Common Function Attributes}, this attribute
4584 allows specification of target-specific compilation options.
4586 When compiling for Nios II, the following options are allowed:
4589 @item custom-@var{insn}=@var{N}
4590 @itemx no-custom-@var{insn}
4591 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4592 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4593 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4594 custom instruction with encoding @var{N} when generating code that uses
4595 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4596 the custom instruction @var{insn}.
4597 These target attributes correspond to the
4598 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4599 command-line options, and support the same set of @var{insn} keywords.
4600 @xref{Nios II Options}, for more information.
4602 @item custom-fpu-cfg=@var{name}
4603 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4604 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4605 command-line option, to select a predefined set of custom instructions
4607 @xref{Nios II Options}, for more information.
4611 @node PowerPC Function Attributes
4612 @subsection PowerPC Function Attributes
4614 These function attributes are supported by the PowerPC back end:
4619 @cindex indirect calls, PowerPC
4620 @cindex @code{longcall} function attribute, PowerPC
4621 @cindex @code{shortcall} function attribute, PowerPC
4622 The @code{longcall} attribute
4623 indicates that the function might be far away from the call site and
4624 require a different (more expensive) calling sequence. The
4625 @code{shortcall} attribute indicates that the function is always close
4626 enough for the shorter calling sequence to be used. These attributes
4627 override both the @option{-mlongcall} switch and
4628 the @code{#pragma longcall} setting.
4630 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4631 calls are necessary.
4633 @item target (@var{options})
4634 @cindex @code{target} function attribute
4635 As discussed in @ref{Common Function Attributes}, this attribute
4636 allows specification of target-specific compilation options.
4638 On the PowerPC, the following options are allowed:
4643 @cindex @code{target("altivec")} function attribute, PowerPC
4644 Generate code that uses (does not use) AltiVec instructions. In
4645 32-bit code, you cannot enable AltiVec instructions unless
4646 @option{-mabi=altivec} is used on the command line.
4650 @cindex @code{target("cmpb")} function attribute, PowerPC
4651 Generate code that uses (does not use) the compare bytes instruction
4652 implemented on the POWER6 processor and other processors that support
4653 the PowerPC V2.05 architecture.
4657 @cindex @code{target("dlmzb")} function attribute, PowerPC
4658 Generate code that uses (does not use) the string-search @samp{dlmzb}
4659 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4660 generated by default when targeting those processors.
4664 @cindex @code{target("fprnd")} function attribute, PowerPC
4665 Generate code that uses (does not use) the FP round to integer
4666 instructions implemented on the POWER5+ processor and other processors
4667 that support the PowerPC V2.03 architecture.
4671 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4672 Generate code that uses (does not use) the decimal floating-point
4673 instructions implemented on some POWER processors.
4677 @cindex @code{target("isel")} function attribute, PowerPC
4678 Generate code that uses (does not use) ISEL instruction.
4682 @cindex @code{target("mfcrf")} function attribute, PowerPC
4683 Generate code that uses (does not use) the move from condition
4684 register field instruction implemented on the POWER4 processor and
4685 other processors that support the PowerPC V2.01 architecture.
4689 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4690 Generate code that uses (does not use) the FP move to/from general
4691 purpose register instructions implemented on the POWER6X processor and
4692 other processors that support the extended PowerPC V2.05 architecture.
4696 @cindex @code{target("mulhw")} function attribute, PowerPC
4697 Generate code that uses (does not use) the half-word multiply and
4698 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4699 These instructions are generated by default when targeting those
4704 @cindex @code{target("multiple")} function attribute, PowerPC
4705 Generate code that uses (does not use) the load multiple word
4706 instructions and the store multiple word instructions.
4710 @cindex @code{target("update")} function attribute, PowerPC
4711 Generate code that uses (does not use) the load or store instructions
4712 that update the base register to the address of the calculated memory
4717 @cindex @code{target("popcntb")} function attribute, PowerPC
4718 Generate code that uses (does not use) the popcount and double-precision
4719 FP reciprocal estimate instruction implemented on the POWER5
4720 processor and other processors that support the PowerPC V2.02
4725 @cindex @code{target("popcntd")} function attribute, PowerPC
4726 Generate code that uses (does not use) the popcount instruction
4727 implemented on the POWER7 processor and other processors that support
4728 the PowerPC V2.06 architecture.
4730 @item powerpc-gfxopt
4731 @itemx no-powerpc-gfxopt
4732 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4733 Generate code that uses (does not use) the optional PowerPC
4734 architecture instructions in the Graphics group, including
4735 floating-point select.
4738 @itemx no-powerpc-gpopt
4739 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4740 Generate code that uses (does not use) the optional PowerPC
4741 architecture instructions in the General Purpose group, including
4742 floating-point square root.
4744 @item recip-precision
4745 @itemx no-recip-precision
4746 @cindex @code{target("recip-precision")} function attribute, PowerPC
4747 Assume (do not assume) that the reciprocal estimate instructions
4748 provide higher-precision estimates than is mandated by the PowerPC
4753 @cindex @code{target("string")} function attribute, PowerPC
4754 Generate code that uses (does not use) the load string instructions
4755 and the store string word instructions to save multiple registers and
4756 do small block moves.
4760 @cindex @code{target("vsx")} function attribute, PowerPC
4761 Generate code that uses (does not use) vector/scalar (VSX)
4762 instructions, and also enable the use of built-in functions that allow
4763 more direct access to the VSX instruction set. In 32-bit code, you
4764 cannot enable VSX or AltiVec instructions unless
4765 @option{-mabi=altivec} is used on the command line.
4769 @cindex @code{target("friz")} function attribute, PowerPC
4770 Generate (do not generate) the @code{friz} instruction when the
4771 @option{-funsafe-math-optimizations} option is used to optimize
4772 rounding a floating-point value to 64-bit integer and back to floating
4773 point. The @code{friz} instruction does not return the same value if
4774 the floating-point number is too large to fit in an integer.
4776 @item avoid-indexed-addresses
4777 @itemx no-avoid-indexed-addresses
4778 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4779 Generate code that tries to avoid (not avoid) the use of indexed load
4780 or store instructions.
4784 @cindex @code{target("paired")} function attribute, PowerPC
4785 Generate code that uses (does not use) the generation of PAIRED simd
4790 @cindex @code{target("longcall")} function attribute, PowerPC
4791 Generate code that assumes (does not assume) that all calls are far
4792 away so that a longer more expensive calling sequence is required.
4795 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4796 Specify the architecture to generate code for when compiling the
4797 function. If you select the @code{target("cpu=power7")} attribute when
4798 generating 32-bit code, VSX and AltiVec instructions are not generated
4799 unless you use the @option{-mabi=altivec} option on the command line.
4801 @item tune=@var{TUNE}
4802 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4803 Specify the architecture to tune for when compiling the function. If
4804 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4805 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4806 compilation tunes for the @var{CPU} architecture, and not the
4807 default tuning specified on the command line.
4810 On the PowerPC, the inliner does not inline a
4811 function that has different target options than the caller, unless the
4812 callee has a subset of the target options of the caller.
4815 @node RL78 Function Attributes
4816 @subsection RL78 Function Attributes
4818 These function attributes are supported by the RL78 back end:
4822 @itemx brk_interrupt
4823 @cindex @code{interrupt} function attribute, RL78
4824 @cindex @code{brk_interrupt} function attribute, RL78
4825 These attributes indicate
4826 that the specified function is an interrupt handler. The compiler generates
4827 function entry and exit sequences suitable for use in an interrupt handler
4828 when this attribute is present.
4830 Use @code{brk_interrupt} instead of @code{interrupt} for
4831 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4832 that must end with @code{RETB} instead of @code{RETI}).
4835 @cindex @code{naked} function attribute, RL78
4836 This attribute allows the compiler to construct the
4837 requisite function declaration, while allowing the body of the
4838 function to be assembly code. The specified function will not have
4839 prologue/epilogue sequences generated by the compiler. Only basic
4840 @code{asm} statements can safely be included in naked functions
4841 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4842 basic @code{asm} and C code may appear to work, they cannot be
4843 depended upon to work reliably and are not supported.
4846 @node RX Function Attributes
4847 @subsection RX Function Attributes
4849 These function attributes are supported by the RX back end:
4852 @item fast_interrupt
4853 @cindex @code{fast_interrupt} function attribute, RX
4854 Use this attribute on the RX port to indicate that the specified
4855 function is a fast interrupt handler. This is just like the
4856 @code{interrupt} attribute, except that @code{freit} is used to return
4857 instead of @code{reit}.
4860 @cindex @code{interrupt} function attribute, RX
4861 Use this attribute to indicate
4862 that the specified function is an interrupt handler. The compiler generates
4863 function entry and exit sequences suitable for use in an interrupt handler
4864 when this attribute is present.
4866 On RX targets, you may specify one or more vector numbers as arguments
4867 to the attribute, as well as naming an alternate table name.
4868 Parameters are handled sequentially, so one handler can be assigned to
4869 multiple entries in multiple tables. One may also pass the magic
4870 string @code{"$default"} which causes the function to be used for any
4871 unfilled slots in the current table.
4873 This example shows a simple assignment of a function to one vector in
4874 the default table (note that preprocessor macros may be used for
4875 chip-specific symbolic vector names):
4877 void __attribute__ ((interrupt (5))) txd1_handler ();
4880 This example assigns a function to two slots in the default table
4881 (using preprocessor macros defined elsewhere) and makes it the default
4882 for the @code{dct} table:
4884 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4889 @cindex @code{naked} function attribute, RX
4890 This attribute allows the compiler to construct the
4891 requisite function declaration, while allowing the body of the
4892 function to be assembly code. The specified function will not have
4893 prologue/epilogue sequences generated by the compiler. Only basic
4894 @code{asm} statements can safely be included in naked functions
4895 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4896 basic @code{asm} and C code may appear to work, they cannot be
4897 depended upon to work reliably and are not supported.
4900 @cindex @code{vector} function attribute, RX
4901 This RX attribute is similar to the @code{interrupt} attribute, including its
4902 parameters, but does not make the function an interrupt-handler type
4903 function (i.e. it retains the normal C function calling ABI). See the
4904 @code{interrupt} attribute for a description of its arguments.
4907 @node S/390 Function Attributes
4908 @subsection S/390 Function Attributes
4910 These function attributes are supported on the S/390:
4913 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4914 @cindex @code{hotpatch} function attribute, S/390
4916 On S/390 System z targets, you can use this function attribute to
4917 make GCC generate a ``hot-patching'' function prologue. If the
4918 @option{-mhotpatch=} command-line option is used at the same time,
4919 the @code{hotpatch} attribute takes precedence. The first of the
4920 two arguments specifies the number of halfwords to be added before
4921 the function label. A second argument can be used to specify the
4922 number of halfwords to be added after the function label. For
4923 both arguments the maximum allowed value is 1000000.
4925 If both arguments are zero, hotpatching is disabled.
4928 @node SH Function Attributes
4929 @subsection SH Function Attributes
4931 These function attributes are supported on the SH family of processors:
4934 @item function_vector
4935 @cindex @code{function_vector} function attribute, SH
4936 @cindex calling functions through the function vector on SH2A
4937 On SH2A targets, this attribute declares a function to be called using the
4938 TBR relative addressing mode. The argument to this attribute is the entry
4939 number of the same function in a vector table containing all the TBR
4940 relative addressable functions. For correct operation the TBR must be setup
4941 accordingly to point to the start of the vector table before any functions with
4942 this attribute are invoked. Usually a good place to do the initialization is
4943 the startup routine. The TBR relative vector table can have at max 256 function
4944 entries. The jumps to these functions are generated using a SH2A specific,
4945 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
4946 from GNU binutils version 2.7 or later for this attribute to work correctly.
4948 In an application, for a function being called once, this attribute
4949 saves at least 8 bytes of code; and if other successive calls are being
4950 made to the same function, it saves 2 bytes of code per each of these
4953 @item interrupt_handler
4954 @cindex @code{interrupt_handler} function attribute, SH
4955 Use this attribute to
4956 indicate that the specified function is an interrupt handler. The compiler
4957 generates function entry and exit sequences suitable for use in an
4958 interrupt handler when this attribute is present.
4960 @item nosave_low_regs
4961 @cindex @code{nosave_low_regs} function attribute, SH
4962 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
4963 function should not save and restore registers R0..R7. This can be used on SH3*
4964 and SH4* targets that have a second R0..R7 register bank for non-reentrant
4968 @cindex @code{renesas} function attribute, SH
4969 On SH targets this attribute specifies that the function or struct follows the
4973 @cindex @code{resbank} function attribute, SH
4974 On the SH2A target, this attribute enables the high-speed register
4975 saving and restoration using a register bank for @code{interrupt_handler}
4976 routines. Saving to the bank is performed automatically after the CPU
4977 accepts an interrupt that uses a register bank.
4979 The nineteen 32-bit registers comprising general register R0 to R14,
4980 control register GBR, and system registers MACH, MACL, and PR and the
4981 vector table address offset are saved into a register bank. Register
4982 banks are stacked in first-in last-out (FILO) sequence. Restoration
4983 from the bank is executed by issuing a RESBANK instruction.
4986 @cindex @code{sp_switch} function attribute, SH
4987 Use this attribute on the SH to indicate an @code{interrupt_handler}
4988 function should switch to an alternate stack. It expects a string
4989 argument that names a global variable holding the address of the
4994 void f () __attribute__ ((interrupt_handler,
4995 sp_switch ("alt_stack")));
4999 @cindex @code{trap_exit} function attribute, SH
5000 Use this attribute on the SH for an @code{interrupt_handler} to return using
5001 @code{trapa} instead of @code{rte}. This attribute expects an integer
5002 argument specifying the trap number to be used.
5005 @cindex @code{trapa_handler} function attribute, SH
5006 On SH targets this function attribute is similar to @code{interrupt_handler}
5007 but it does not save and restore all registers.
5010 @node SPU Function Attributes
5011 @subsection SPU Function Attributes
5013 These function attributes are supported by the SPU back end:
5017 @cindex @code{naked} function attribute, SPU
5018 This attribute allows the compiler to construct the
5019 requisite function declaration, while allowing the body of the
5020 function to be assembly code. The specified function will not have
5021 prologue/epilogue sequences generated by the compiler. Only basic
5022 @code{asm} statements can safely be included in naked functions
5023 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5024 basic @code{asm} and C code may appear to work, they cannot be
5025 depended upon to work reliably and are not supported.
5028 @node Symbian OS Function Attributes
5029 @subsection Symbian OS Function Attributes
5031 @xref{Microsoft Windows Function Attributes}, for discussion of the
5032 @code{dllexport} and @code{dllimport} attributes.
5034 @node Visium Function Attributes
5035 @subsection Visium Function Attributes
5037 These function attributes are supported by the Visium back end:
5041 @cindex @code{interrupt} function attribute, Visium
5042 Use this attribute to indicate
5043 that the specified function is an interrupt handler. The compiler generates
5044 function entry and exit sequences suitable for use in an interrupt handler
5045 when this attribute is present.
5048 @node x86 Function Attributes
5049 @subsection x86 Function Attributes
5051 These function attributes are supported by the x86 back end:
5055 @cindex @code{cdecl} function attribute, x86-32
5056 @cindex functions that pop the argument stack on x86-32
5058 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5059 assume that the calling function pops off the stack space used to
5060 pass arguments. This is
5061 useful to override the effects of the @option{-mrtd} switch.
5064 @cindex @code{fastcall} function attribute, x86-32
5065 @cindex functions that pop the argument stack on x86-32
5066 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5067 pass the first argument (if of integral type) in the register ECX and
5068 the second argument (if of integral type) in the register EDX@. Subsequent
5069 and other typed arguments are passed on the stack. The called function
5070 pops the arguments off the stack. If the number of arguments is variable all
5071 arguments are pushed on the stack.
5074 @cindex @code{thiscall} function attribute, x86-32
5075 @cindex functions that pop the argument stack on x86-32
5076 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5077 pass the first argument (if of integral type) in the register ECX.
5078 Subsequent and other typed arguments are passed on the stack. The called
5079 function pops the arguments off the stack.
5080 If the number of arguments is variable all arguments are pushed on the
5082 The @code{thiscall} attribute is intended for C++ non-static member functions.
5083 As a GCC extension, this calling convention can be used for C functions
5084 and for static member methods.
5088 @cindex @code{ms_abi} function attribute, x86
5089 @cindex @code{sysv_abi} function attribute, x86
5091 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5092 to indicate which calling convention should be used for a function. The
5093 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5094 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5095 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5096 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5098 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5099 requires the @option{-maccumulate-outgoing-args} option.
5101 @item callee_pop_aggregate_return (@var{number})
5102 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5104 On x86-32 targets, you can use this attribute to control how
5105 aggregates are returned in memory. If the caller is responsible for
5106 popping the hidden pointer together with the rest of the arguments, specify
5107 @var{number} equal to zero. If callee is responsible for popping the
5108 hidden pointer, specify @var{number} equal to one.
5110 The default x86-32 ABI assumes that the callee pops the
5111 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5112 the compiler assumes that the
5113 caller pops the stack for hidden pointer.
5115 @item ms_hook_prologue
5116 @cindex @code{ms_hook_prologue} function attribute, x86
5118 On 32-bit and 64-bit x86 targets, you can use
5119 this function attribute to make GCC generate the ``hot-patching'' function
5120 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5123 @item regparm (@var{number})
5124 @cindex @code{regparm} function attribute, x86
5125 @cindex functions that are passed arguments in registers on x86-32
5126 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5127 pass arguments number one to @var{number} if they are of integral type
5128 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5129 take a variable number of arguments continue to be passed all of their
5130 arguments on the stack.
5132 Beware that on some ELF systems this attribute is unsuitable for
5133 global functions in shared libraries with lazy binding (which is the
5134 default). Lazy binding sends the first call via resolving code in
5135 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5136 per the standard calling conventions. Solaris 8 is affected by this.
5137 Systems with the GNU C Library version 2.1 or higher
5138 and FreeBSD are believed to be
5139 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5140 disabled with the linker or the loader if desired, to avoid the
5144 @cindex @code{sseregparm} function attribute, x86
5145 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5146 causes the compiler to pass up to 3 floating-point arguments in
5147 SSE registers instead of on the stack. Functions that take a
5148 variable number of arguments continue to pass all of their
5149 floating-point arguments on the stack.
5151 @item force_align_arg_pointer
5152 @cindex @code{force_align_arg_pointer} function attribute, x86
5153 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5154 applied to individual function definitions, generating an alternate
5155 prologue and epilogue that realigns the run-time stack if necessary.
5156 This supports mixing legacy codes that run with a 4-byte aligned stack
5157 with modern codes that keep a 16-byte stack for SSE compatibility.
5160 @cindex @code{stdcall} function attribute, x86-32
5161 @cindex functions that pop the argument stack on x86-32
5162 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5163 assume that the called function pops off the stack space used to
5164 pass arguments, unless it takes a variable number of arguments.
5166 @item target (@var{options})
5167 @cindex @code{target} function attribute
5168 As discussed in @ref{Common Function Attributes}, this attribute
5169 allows specification of target-specific compilation options.
5171 On the x86, the following options are allowed:
5175 @cindex @code{target("abm")} function attribute, x86
5176 Enable/disable the generation of the advanced bit instructions.
5180 @cindex @code{target("aes")} function attribute, x86
5181 Enable/disable the generation of the AES instructions.
5184 @cindex @code{target("default")} function attribute, x86
5185 @xref{Function Multiversioning}, where it is used to specify the
5186 default function version.
5190 @cindex @code{target("mmx")} function attribute, x86
5191 Enable/disable the generation of the MMX instructions.
5195 @cindex @code{target("pclmul")} function attribute, x86
5196 Enable/disable the generation of the PCLMUL instructions.
5200 @cindex @code{target("popcnt")} function attribute, x86
5201 Enable/disable the generation of the POPCNT instruction.
5205 @cindex @code{target("sse")} function attribute, x86
5206 Enable/disable the generation of the SSE instructions.
5210 @cindex @code{target("sse2")} function attribute, x86
5211 Enable/disable the generation of the SSE2 instructions.
5215 @cindex @code{target("sse3")} function attribute, x86
5216 Enable/disable the generation of the SSE3 instructions.
5220 @cindex @code{target("sse4")} function attribute, x86
5221 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5226 @cindex @code{target("sse4.1")} function attribute, x86
5227 Enable/disable the generation of the sse4.1 instructions.
5231 @cindex @code{target("sse4.2")} function attribute, x86
5232 Enable/disable the generation of the sse4.2 instructions.
5236 @cindex @code{target("sse4a")} function attribute, x86
5237 Enable/disable the generation of the SSE4A instructions.
5241 @cindex @code{target("fma4")} function attribute, x86
5242 Enable/disable the generation of the FMA4 instructions.
5246 @cindex @code{target("xop")} function attribute, x86
5247 Enable/disable the generation of the XOP instructions.
5251 @cindex @code{target("lwp")} function attribute, x86
5252 Enable/disable the generation of the LWP instructions.
5256 @cindex @code{target("ssse3")} function attribute, x86
5257 Enable/disable the generation of the SSSE3 instructions.
5261 @cindex @code{target("cld")} function attribute, x86
5262 Enable/disable the generation of the CLD before string moves.
5264 @item fancy-math-387
5265 @itemx no-fancy-math-387
5266 @cindex @code{target("fancy-math-387")} function attribute, x86
5267 Enable/disable the generation of the @code{sin}, @code{cos}, and
5268 @code{sqrt} instructions on the 387 floating-point unit.
5271 @itemx no-fused-madd
5272 @cindex @code{target("fused-madd")} function attribute, x86
5273 Enable/disable the generation of the fused multiply/add instructions.
5277 @cindex @code{target("ieee-fp")} function attribute, x86
5278 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5280 @item inline-all-stringops
5281 @itemx no-inline-all-stringops
5282 @cindex @code{target("inline-all-stringops")} function attribute, x86
5283 Enable/disable inlining of string operations.
5285 @item inline-stringops-dynamically
5286 @itemx no-inline-stringops-dynamically
5287 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5288 Enable/disable the generation of the inline code to do small string
5289 operations and calling the library routines for large operations.
5291 @item align-stringops
5292 @itemx no-align-stringops
5293 @cindex @code{target("align-stringops")} function attribute, x86
5294 Do/do not align destination of inlined string operations.
5298 @cindex @code{target("recip")} function attribute, x86
5299 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5300 instructions followed an additional Newton-Raphson step instead of
5301 doing a floating-point division.
5303 @item arch=@var{ARCH}
5304 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5305 Specify the architecture to generate code for in compiling the function.
5307 @item tune=@var{TUNE}
5308 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5309 Specify the architecture to tune for in compiling the function.
5311 @item fpmath=@var{FPMATH}
5312 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5313 Specify which floating-point unit to use. You must specify the
5314 @code{target("fpmath=sse,387")} option as
5315 @code{target("fpmath=sse+387")} because the comma would separate
5319 On the x86, the inliner does not inline a
5320 function that has different target options than the caller, unless the
5321 callee has a subset of the target options of the caller. For example
5322 a function declared with @code{target("sse3")} can inline a function
5323 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5326 @node Xstormy16 Function Attributes
5327 @subsection Xstormy16 Function Attributes
5329 These function attributes are supported by the Xstormy16 back end:
5333 @cindex @code{interrupt} function attribute, Xstormy16
5334 Use this attribute to indicate
5335 that the specified function is an interrupt handler. The compiler generates
5336 function entry and exit sequences suitable for use in an interrupt handler
5337 when this attribute is present.
5340 @node Variable Attributes
5341 @section Specifying Attributes of Variables
5342 @cindex attribute of variables
5343 @cindex variable attributes
5345 The keyword @code{__attribute__} allows you to specify special
5346 attributes of variables or structure fields. This keyword is followed
5347 by an attribute specification inside double parentheses. Some
5348 attributes are currently defined generically for variables.
5349 Other attributes are defined for variables on particular target
5350 systems. Other attributes are available for functions
5351 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5352 enumerators (@pxref{Enumerator Attributes}), and for types
5353 (@pxref{Type Attributes}).
5354 Other front ends might define more attributes
5355 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5357 @xref{Attribute Syntax}, for details of the exact syntax for using
5361 * Common Variable Attributes::
5362 * AVR Variable Attributes::
5363 * Blackfin Variable Attributes::
5364 * H8/300 Variable Attributes::
5365 * IA-64 Variable Attributes::
5366 * M32R/D Variable Attributes::
5367 * MeP Variable Attributes::
5368 * Microsoft Windows Variable Attributes::
5369 * MSP430 Variable Attributes::
5370 * PowerPC Variable Attributes::
5371 * SPU Variable Attributes::
5372 * x86 Variable Attributes::
5373 * Xstormy16 Variable Attributes::
5376 @node Common Variable Attributes
5377 @subsection Common Variable Attributes
5379 The following attributes are supported on most targets.
5382 @cindex @code{aligned} variable attribute
5383 @item aligned (@var{alignment})
5384 This attribute specifies a minimum alignment for the variable or
5385 structure field, measured in bytes. For example, the declaration:
5388 int x __attribute__ ((aligned (16))) = 0;
5392 causes the compiler to allocate the global variable @code{x} on a
5393 16-byte boundary. On a 68040, this could be used in conjunction with
5394 an @code{asm} expression to access the @code{move16} instruction which
5395 requires 16-byte aligned operands.
5397 You can also specify the alignment of structure fields. For example, to
5398 create a double-word aligned @code{int} pair, you could write:
5401 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5405 This is an alternative to creating a union with a @code{double} member,
5406 which forces the union to be double-word aligned.
5408 As in the preceding examples, you can explicitly specify the alignment
5409 (in bytes) that you wish the compiler to use for a given variable or
5410 structure field. Alternatively, you can leave out the alignment factor
5411 and just ask the compiler to align a variable or field to the
5412 default alignment for the target architecture you are compiling for.
5413 The default alignment is sufficient for all scalar types, but may not be
5414 enough for all vector types on a target that supports vector operations.
5415 The default alignment is fixed for a particular target ABI.
5417 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5418 which is the largest alignment ever used for any data type on the
5419 target machine you are compiling for. For example, you could write:
5422 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5425 The compiler automatically sets the alignment for the declared
5426 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5427 often make copy operations more efficient, because the compiler can
5428 use whatever instructions copy the biggest chunks of memory when
5429 performing copies to or from the variables or fields that you have
5430 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5431 may change depending on command-line options.
5433 When used on a struct, or struct member, the @code{aligned} attribute can
5434 only increase the alignment; in order to decrease it, the @code{packed}
5435 attribute must be specified as well. When used as part of a typedef, the
5436 @code{aligned} attribute can both increase and decrease alignment, and
5437 specifying the @code{packed} attribute generates a warning.
5439 Note that the effectiveness of @code{aligned} attributes may be limited
5440 by inherent limitations in your linker. On many systems, the linker is
5441 only able to arrange for variables to be aligned up to a certain maximum
5442 alignment. (For some linkers, the maximum supported alignment may
5443 be very very small.) If your linker is only able to align variables
5444 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5445 in an @code{__attribute__} still only provides you with 8-byte
5446 alignment. See your linker documentation for further information.
5448 The @code{aligned} attribute can also be used for functions
5449 (@pxref{Common Function Attributes}.)
5451 @item cleanup (@var{cleanup_function})
5452 @cindex @code{cleanup} variable attribute
5453 The @code{cleanup} attribute runs a function when the variable goes
5454 out of scope. This attribute can only be applied to auto function
5455 scope variables; it may not be applied to parameters or variables
5456 with static storage duration. The function must take one parameter,
5457 a pointer to a type compatible with the variable. The return value
5458 of the function (if any) is ignored.
5460 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5461 is run during the stack unwinding that happens during the
5462 processing of the exception. Note that the @code{cleanup} attribute
5463 does not allow the exception to be caught, only to perform an action.
5464 It is undefined what happens if @var{cleanup_function} does not
5469 @cindex @code{common} variable attribute
5470 @cindex @code{nocommon} variable attribute
5473 The @code{common} attribute requests GCC to place a variable in
5474 ``common'' storage. The @code{nocommon} attribute requests the
5475 opposite---to allocate space for it directly.
5477 These attributes override the default chosen by the
5478 @option{-fno-common} and @option{-fcommon} flags respectively.
5481 @itemx deprecated (@var{msg})
5482 @cindex @code{deprecated} variable attribute
5483 The @code{deprecated} attribute results in a warning if the variable
5484 is used anywhere in the source file. This is useful when identifying
5485 variables that are expected to be removed in a future version of a
5486 program. The warning also includes the location of the declaration
5487 of the deprecated variable, to enable users to easily find further
5488 information about why the variable is deprecated, or what they should
5489 do instead. Note that the warning only occurs for uses:
5492 extern int old_var __attribute__ ((deprecated));
5494 int new_fn () @{ return old_var; @}
5498 results in a warning on line 3 but not line 2. The optional @var{msg}
5499 argument, which must be a string, is printed in the warning if
5502 The @code{deprecated} attribute can also be used for functions and
5503 types (@pxref{Common Function Attributes},
5504 @pxref{Common Type Attributes}).
5506 @item mode (@var{mode})
5507 @cindex @code{mode} variable attribute
5508 This attribute specifies the data type for the declaration---whichever
5509 type corresponds to the mode @var{mode}. This in effect lets you
5510 request an integer or floating-point type according to its width.
5512 You may also specify a mode of @code{byte} or @code{__byte__} to
5513 indicate the mode corresponding to a one-byte integer, @code{word} or
5514 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5515 or @code{__pointer__} for the mode used to represent pointers.
5518 @cindex @code{packed} variable attribute
5519 The @code{packed} attribute specifies that a variable or structure field
5520 should have the smallest possible alignment---one byte for a variable,
5521 and one bit for a field, unless you specify a larger value with the
5522 @code{aligned} attribute.
5524 Here is a structure in which the field @code{x} is packed, so that it
5525 immediately follows @code{a}:
5531 int x[2] __attribute__ ((packed));
5535 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5536 @code{packed} attribute on bit-fields of type @code{char}. This has
5537 been fixed in GCC 4.4 but the change can lead to differences in the
5538 structure layout. See the documentation of
5539 @option{-Wpacked-bitfield-compat} for more information.
5541 @item section ("@var{section-name}")
5542 @cindex @code{section} variable attribute
5543 Normally, the compiler places the objects it generates in sections like
5544 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5545 or you need certain particular variables to appear in special sections,
5546 for example to map to special hardware. The @code{section}
5547 attribute specifies that a variable (or function) lives in a particular
5548 section. For example, this small program uses several specific section names:
5551 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5552 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5553 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5554 int init_data __attribute__ ((section ("INITDATA")));
5558 /* @r{Initialize stack pointer} */
5559 init_sp (stack + sizeof (stack));
5561 /* @r{Initialize initialized data} */
5562 memcpy (&init_data, &data, &edata - &data);
5564 /* @r{Turn on the serial ports} */
5571 Use the @code{section} attribute with
5572 @emph{global} variables and not @emph{local} variables,
5573 as shown in the example.
5575 You may use the @code{section} attribute with initialized or
5576 uninitialized global variables but the linker requires
5577 each object be defined once, with the exception that uninitialized
5578 variables tentatively go in the @code{common} (or @code{bss}) section
5579 and can be multiply ``defined''. Using the @code{section} attribute
5580 changes what section the variable goes into and may cause the
5581 linker to issue an error if an uninitialized variable has multiple
5582 definitions. You can force a variable to be initialized with the
5583 @option{-fno-common} flag or the @code{nocommon} attribute.
5585 Some file formats do not support arbitrary sections so the @code{section}
5586 attribute is not available on all platforms.
5587 If you need to map the entire contents of a module to a particular
5588 section, consider using the facilities of the linker instead.
5590 @item tls_model ("@var{tls_model}")
5591 @cindex @code{tls_model} variable attribute
5592 The @code{tls_model} attribute sets thread-local storage model
5593 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5594 overriding @option{-ftls-model=} command-line switch on a per-variable
5596 The @var{tls_model} argument should be one of @code{global-dynamic},
5597 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5599 Not all targets support this attribute.
5602 @cindex @code{unused} variable attribute
5603 This attribute, attached to a variable, means that the variable is meant
5604 to be possibly unused. GCC does not produce a warning for this
5608 @cindex @code{used} variable attribute
5609 This attribute, attached to a variable with static storage, means that
5610 the variable must be emitted even if it appears that the variable is not
5613 When applied to a static data member of a C++ class template, the
5614 attribute also means that the member is instantiated if the
5615 class itself is instantiated.
5617 @item vector_size (@var{bytes})
5618 @cindex @code{vector_size} variable attribute
5619 This attribute specifies the vector size for the variable, measured in
5620 bytes. For example, the declaration:
5623 int foo __attribute__ ((vector_size (16)));
5627 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5628 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5629 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5631 This attribute is only applicable to integral and float scalars,
5632 although arrays, pointers, and function return values are allowed in
5633 conjunction with this construct.
5635 Aggregates with this attribute are invalid, even if they are of the same
5636 size as a corresponding scalar. For example, the declaration:
5639 struct S @{ int a; @};
5640 struct S __attribute__ ((vector_size (16))) foo;
5644 is invalid even if the size of the structure is the same as the size of
5648 @cindex @code{weak} variable attribute
5649 The @code{weak} attribute is described in
5650 @ref{Common Function Attributes}.
5654 @node AVR Variable Attributes
5655 @subsection AVR Variable Attributes
5659 @cindex @code{progmem} variable attribute, AVR
5660 The @code{progmem} attribute is used on the AVR to place read-only
5661 data in the non-volatile program memory (flash). The @code{progmem}
5662 attribute accomplishes this by putting respective variables into a
5663 section whose name starts with @code{.progmem}.
5665 This attribute works similar to the @code{section} attribute
5666 but adds additional checking. Notice that just like the
5667 @code{section} attribute, @code{progmem} affects the location
5668 of the data but not how this data is accessed.
5670 In order to read data located with the @code{progmem} attribute
5671 (inline) assembler must be used.
5673 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5674 #include <avr/pgmspace.h>
5676 /* Locate var in flash memory */
5677 const int var[2] PROGMEM = @{ 1, 2 @};
5679 int read_var (int i)
5681 /* Access var[] by accessor macro from avr/pgmspace.h */
5682 return (int) pgm_read_word (& var[i]);
5686 AVR is a Harvard architecture processor and data and read-only data
5687 normally resides in the data memory (RAM).
5689 See also the @ref{AVR Named Address Spaces} section for
5690 an alternate way to locate and access data in flash memory.
5693 @itemx io (@var{addr})
5694 @cindex @code{io} variable attribute, AVR
5695 Variables with the @code{io} attribute are used to address
5696 memory-mapped peripherals in the io address range.
5697 If an address is specified, the variable
5698 is assigned that address, and the value is interpreted as an
5699 address in the data address space.
5703 volatile int porta __attribute__((io (0x22)));
5706 The address specified in the address in the data address range.
5708 Otherwise, the variable it is not assigned an address, but the
5709 compiler will still use in/out instructions where applicable,
5710 assuming some other module assigns an address in the io address range.
5714 extern volatile int porta __attribute__((io));
5718 @itemx io_low (@var{addr})
5719 @cindex @code{io_low} variable attribute, AVR
5720 This is like the @code{io} attribute, but additionally it informs the
5721 compiler that the object lies in the lower half of the I/O area,
5722 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5726 @itemx address (@var{addr})
5727 @cindex @code{address} variable attribute, AVR
5728 Variables with the @code{address} attribute are used to address
5729 memory-mapped peripherals that may lie outside the io address range.
5732 volatile int porta __attribute__((address (0x600)));
5737 @node Blackfin Variable Attributes
5738 @subsection Blackfin Variable Attributes
5740 Three attributes are currently defined for the Blackfin.
5746 @cindex @code{l1_data} variable attribute, Blackfin
5747 @cindex @code{l1_data_A} variable attribute, Blackfin
5748 @cindex @code{l1_data_B} variable attribute, Blackfin
5749 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5750 Variables with @code{l1_data} attribute are put into the specific section
5751 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5752 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5753 attribute are put into the specific section named @code{.l1.data.B}.
5756 @cindex @code{l2} variable attribute, Blackfin
5757 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5758 Variables with @code{l2} attribute are put into the specific section
5759 named @code{.l2.data}.
5762 @node H8/300 Variable Attributes
5763 @subsection H8/300 Variable Attributes
5765 These variable attributes are available for H8/300 targets:
5769 @cindex @code{eightbit_data} variable attribute, H8/300
5770 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5771 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5772 variable should be placed into the eight-bit data section.
5773 The compiler generates more efficient code for certain operations
5774 on data in the eight-bit data area. Note the eight-bit data area is limited to
5777 You must use GAS and GLD from GNU binutils version 2.7 or later for
5778 this attribute to work correctly.
5781 @cindex @code{tiny_data} variable attribute, H8/300
5782 @cindex tiny data section on the H8/300H and H8S
5783 Use this attribute on the H8/300H and H8S to indicate that the specified
5784 variable should be placed into the tiny data section.
5785 The compiler generates more efficient code for loads and stores
5786 on data in the tiny data section. Note the tiny data area is limited to
5787 slightly under 32KB of data.
5791 @node IA-64 Variable Attributes
5792 @subsection IA-64 Variable Attributes
5794 The IA-64 back end supports the following variable attribute:
5797 @item model (@var{model-name})
5798 @cindex @code{model} variable attribute, IA-64
5800 On IA-64, use this attribute to set the addressability of an object.
5801 At present, the only supported identifier for @var{model-name} is
5802 @code{small}, indicating addressability via ``small'' (22-bit)
5803 addresses (so that their addresses can be loaded with the @code{addl}
5804 instruction). Caveat: such addressing is by definition not position
5805 independent and hence this attribute must not be used for objects
5806 defined by shared libraries.
5810 @node M32R/D Variable Attributes
5811 @subsection M32R/D Variable Attributes
5813 One attribute is currently defined for the M32R/D@.
5816 @item model (@var{model-name})
5817 @cindex @code{model-name} variable attribute, M32R/D
5818 @cindex variable addressability on the M32R/D
5819 Use this attribute on the M32R/D to set the addressability of an object.
5820 The identifier @var{model-name} is one of @code{small}, @code{medium},
5821 or @code{large}, representing each of the code models.
5823 Small model objects live in the lower 16MB of memory (so that their
5824 addresses can be loaded with the @code{ld24} instruction).
5826 Medium and large model objects may live anywhere in the 32-bit address space
5827 (the compiler generates @code{seth/add3} instructions to load their
5831 @node MeP Variable Attributes
5832 @subsection MeP Variable Attributes
5834 The MeP target has a number of addressing modes and busses. The
5835 @code{near} space spans the standard memory space's first 16 megabytes
5836 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5837 The @code{based} space is a 128-byte region in the memory space that
5838 is addressed relative to the @code{$tp} register. The @code{tiny}
5839 space is a 65536-byte region relative to the @code{$gp} register. In
5840 addition to these memory regions, the MeP target has a separate 16-bit
5841 control bus which is specified with @code{cb} attributes.
5846 @cindex @code{based} variable attribute, MeP
5847 Any variable with the @code{based} attribute is assigned to the
5848 @code{.based} section, and is accessed with relative to the
5849 @code{$tp} register.
5852 @cindex @code{tiny} variable attribute, MeP
5853 Likewise, the @code{tiny} attribute assigned variables to the
5854 @code{.tiny} section, relative to the @code{$gp} register.
5857 @cindex @code{near} variable attribute, MeP
5858 Variables with the @code{near} attribute are assumed to have addresses
5859 that fit in a 24-bit addressing mode. This is the default for large
5860 variables (@code{-mtiny=4} is the default) but this attribute can
5861 override @code{-mtiny=} for small variables, or override @code{-ml}.
5864 @cindex @code{far} variable attribute, MeP
5865 Variables with the @code{far} attribute are addressed using a full
5866 32-bit address. Since this covers the entire memory space, this
5867 allows modules to make no assumptions about where variables might be
5871 @cindex @code{io} variable attribute, MeP
5872 @itemx io (@var{addr})
5873 Variables with the @code{io} attribute are used to address
5874 memory-mapped peripherals. If an address is specified, the variable
5875 is assigned that address, else it is not assigned an address (it is
5876 assumed some other module assigns an address). Example:
5879 int timer_count __attribute__((io(0x123)));
5883 @itemx cb (@var{addr})
5884 @cindex @code{cb} variable attribute, MeP
5885 Variables with the @code{cb} attribute are used to access the control
5886 bus, using special instructions. @code{addr} indicates the control bus
5890 int cpu_clock __attribute__((cb(0x123)));
5895 @node Microsoft Windows Variable Attributes
5896 @subsection Microsoft Windows Variable Attributes
5898 You can use these attributes on Microsoft Windows targets.
5899 @ref{x86 Variable Attributes} for additional Windows compatibility
5900 attributes available on all x86 targets.
5905 @cindex @code{dllimport} variable attribute
5906 @cindex @code{dllexport} variable attribute
5907 The @code{dllimport} and @code{dllexport} attributes are described in
5908 @ref{Microsoft Windows Function Attributes}.
5911 @cindex @code{selectany} variable attribute
5912 The @code{selectany} attribute causes an initialized global variable to
5913 have link-once semantics. When multiple definitions of the variable are
5914 encountered by the linker, the first is selected and the remainder are
5915 discarded. Following usage by the Microsoft compiler, the linker is told
5916 @emph{not} to warn about size or content differences of the multiple
5919 Although the primary usage of this attribute is for POD types, the
5920 attribute can also be applied to global C++ objects that are initialized
5921 by a constructor. In this case, the static initialization and destruction
5922 code for the object is emitted in each translation defining the object,
5923 but the calls to the constructor and destructor are protected by a
5924 link-once guard variable.
5926 The @code{selectany} attribute is only available on Microsoft Windows
5927 targets. You can use @code{__declspec (selectany)} as a synonym for
5928 @code{__attribute__ ((selectany))} for compatibility with other
5932 @cindex @code{shared} variable attribute
5933 On Microsoft Windows, in addition to putting variable definitions in a named
5934 section, the section can also be shared among all running copies of an
5935 executable or DLL@. For example, this small program defines shared data
5936 by putting it in a named section @code{shared} and marking the section
5940 int foo __attribute__((section ("shared"), shared)) = 0;
5945 /* @r{Read and write foo. All running
5946 copies see the same value.} */
5952 You may only use the @code{shared} attribute along with @code{section}
5953 attribute with a fully-initialized global definition because of the way
5954 linkers work. See @code{section} attribute for more information.
5956 The @code{shared} attribute is only available on Microsoft Windows@.
5960 @node MSP430 Variable Attributes
5961 @subsection MSP430 Variable Attributes
5965 @cindex @code{noinit} MSP430 variable attribute
5966 Any data with the @code{noinit} attribute will not be initialised by
5967 the C runtime startup code, or the program loader. Not initialising
5968 data in this way can reduce program startup times.
5971 @cindex @code{persistent} MSP430 variable attribute
5972 Any variable with the @code{persistent} attribute will not be
5973 initialised by the C runtime startup code. Instead its value will be
5974 set once, when the application is loaded, and then never initialised
5975 again, even if the processor is reset or the program restarts.
5976 Persistent data is intended to be placed into FLASH RAM, where its
5977 value will be retained across resets. The linker script being used to
5978 create the application should ensure that persistent data is correctly
5984 @cindex @code{lower} memory region on the MSP430
5985 @cindex @code{upper} memory region on the MSP430
5986 @cindex @code{either} memory region on the MSP430
5987 These attributes are the same as the MSP430 function attributes of the
5988 same name. These attributes can be applied to both functions and
5992 @node PowerPC Variable Attributes
5993 @subsection PowerPC Variable Attributes
5995 Three attributes currently are defined for PowerPC configurations:
5996 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5998 @cindex @code{ms_struct} variable attribute, PowerPC
5999 @cindex @code{gcc_struct} variable attribute, PowerPC
6000 For full documentation of the struct attributes please see the
6001 documentation in @ref{x86 Variable Attributes}.
6003 @cindex @code{altivec} variable attribute, PowerPC
6004 For documentation of @code{altivec} attribute please see the
6005 documentation in @ref{PowerPC Type Attributes}.
6007 @node SPU Variable Attributes
6008 @subsection SPU Variable Attributes
6010 @cindex @code{spu_vector} variable attribute, SPU
6011 The SPU supports the @code{spu_vector} attribute for variables. For
6012 documentation of this attribute please see the documentation in
6013 @ref{SPU Type Attributes}.
6015 @node x86 Variable Attributes
6016 @subsection x86 Variable Attributes
6018 Two attributes are currently defined for x86 configurations:
6019 @code{ms_struct} and @code{gcc_struct}.
6024 @cindex @code{ms_struct} variable attribute, x86
6025 @cindex @code{gcc_struct} variable attribute, x86
6027 If @code{packed} is used on a structure, or if bit-fields are used,
6028 it may be that the Microsoft ABI lays out the structure differently
6029 than the way GCC normally does. Particularly when moving packed
6030 data between functions compiled with GCC and the native Microsoft compiler
6031 (either via function call or as data in a file), it may be necessary to access
6034 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
6035 compilers to match the native Microsoft compiler.
6037 The Microsoft structure layout algorithm is fairly simple with the exception
6038 of the bit-field packing.
6039 The padding and alignment of members of structures and whether a bit-field
6040 can straddle a storage-unit boundary are determine by these rules:
6043 @item Structure members are stored sequentially in the order in which they are
6044 declared: the first member has the lowest memory address and the last member
6047 @item Every data object has an alignment requirement. The alignment requirement
6048 for all data except structures, unions, and arrays is either the size of the
6049 object or the current packing size (specified with either the
6050 @code{aligned} attribute or the @code{pack} pragma),
6051 whichever is less. For structures, unions, and arrays,
6052 the alignment requirement is the largest alignment requirement of its members.
6053 Every object is allocated an offset so that:
6056 offset % alignment_requirement == 0
6059 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
6060 unit if the integral types are the same size and if the next bit-field fits
6061 into the current allocation unit without crossing the boundary imposed by the
6062 common alignment requirements of the bit-fields.
6065 MSVC interprets zero-length bit-fields in the following ways:
6068 @item If a zero-length bit-field is inserted between two bit-fields that
6069 are normally coalesced, the bit-fields are not coalesced.
6076 unsigned long bf_1 : 12;
6078 unsigned long bf_2 : 12;
6083 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
6084 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
6086 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
6087 alignment of the zero-length bit-field is greater than the member that follows it,
6088 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
6109 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
6110 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
6111 bit-field does not affect the alignment of @code{bar} or, as a result, the size
6114 Taking this into account, it is important to note the following:
6117 @item If a zero-length bit-field follows a normal bit-field, the type of the
6118 zero-length bit-field may affect the alignment of the structure as whole. For
6119 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
6120 normal bit-field, and is of type short.
6122 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
6123 still affect the alignment of the structure:
6134 Here, @code{t4} takes up 4 bytes.
6137 @item Zero-length bit-fields following non-bit-field members are ignored:
6149 Here, @code{t5} takes up 2 bytes.
6153 @node Xstormy16 Variable Attributes
6154 @subsection Xstormy16 Variable Attributes
6156 One attribute is currently defined for xstormy16 configurations:
6161 @cindex @code{below100} variable attribute, Xstormy16
6163 If a variable has the @code{below100} attribute (@code{BELOW100} is
6164 allowed also), GCC places the variable in the first 0x100 bytes of
6165 memory and use special opcodes to access it. Such variables are
6166 placed in either the @code{.bss_below100} section or the
6167 @code{.data_below100} section.
6171 @node Type Attributes
6172 @section Specifying Attributes of Types
6173 @cindex attribute of types
6174 @cindex type attributes
6176 The keyword @code{__attribute__} allows you to specify special
6177 attributes of types. Some type attributes apply only to @code{struct}
6178 and @code{union} types, while others can apply to any type defined
6179 via a @code{typedef} declaration. Other attributes are defined for
6180 functions (@pxref{Function Attributes}), labels (@pxref{Label
6181 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6182 variables (@pxref{Variable Attributes}).
6184 The @code{__attribute__} keyword is followed by an attribute specification
6185 inside double parentheses.
6187 You may specify type attributes in an enum, struct or union type
6188 declaration or definition by placing them immediately after the
6189 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6190 syntax is to place them just past the closing curly brace of the
6193 You can also include type attributes in a @code{typedef} declaration.
6194 @xref{Attribute Syntax}, for details of the exact syntax for using
6198 * Common Type Attributes::
6199 * ARM Type Attributes::
6200 * MeP Type Attributes::
6201 * PowerPC Type Attributes::
6202 * SPU Type Attributes::
6203 * x86 Type Attributes::
6206 @node Common Type Attributes
6207 @subsection Common Type Attributes
6209 The following type attributes are supported on most targets.
6212 @cindex @code{aligned} type attribute
6213 @item aligned (@var{alignment})
6214 This attribute specifies a minimum alignment (in bytes) for variables
6215 of the specified type. For example, the declarations:
6218 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6219 typedef int more_aligned_int __attribute__ ((aligned (8)));
6223 force the compiler to ensure (as far as it can) that each variable whose
6224 type is @code{struct S} or @code{more_aligned_int} is allocated and
6225 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6226 variables of type @code{struct S} aligned to 8-byte boundaries allows
6227 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6228 store) instructions when copying one variable of type @code{struct S} to
6229 another, thus improving run-time efficiency.
6231 Note that the alignment of any given @code{struct} or @code{union} type
6232 is required by the ISO C standard to be at least a perfect multiple of
6233 the lowest common multiple of the alignments of all of the members of
6234 the @code{struct} or @code{union} in question. This means that you @emph{can}
6235 effectively adjust the alignment of a @code{struct} or @code{union}
6236 type by attaching an @code{aligned} attribute to any one of the members
6237 of such a type, but the notation illustrated in the example above is a
6238 more obvious, intuitive, and readable way to request the compiler to
6239 adjust the alignment of an entire @code{struct} or @code{union} type.
6241 As in the preceding example, you can explicitly specify the alignment
6242 (in bytes) that you wish the compiler to use for a given @code{struct}
6243 or @code{union} type. Alternatively, you can leave out the alignment factor
6244 and just ask the compiler to align a type to the maximum
6245 useful alignment for the target machine you are compiling for. For
6246 example, you could write:
6249 struct S @{ short f[3]; @} __attribute__ ((aligned));
6252 Whenever you leave out the alignment factor in an @code{aligned}
6253 attribute specification, the compiler automatically sets the alignment
6254 for the type to the largest alignment that is ever used for any data
6255 type on the target machine you are compiling for. Doing this can often
6256 make copy operations more efficient, because the compiler can use
6257 whatever instructions copy the biggest chunks of memory when performing
6258 copies to or from the variables that have types that you have aligned
6261 In the example above, if the size of each @code{short} is 2 bytes, then
6262 the size of the entire @code{struct S} type is 6 bytes. The smallest
6263 power of two that is greater than or equal to that is 8, so the
6264 compiler sets the alignment for the entire @code{struct S} type to 8
6267 Note that although you can ask the compiler to select a time-efficient
6268 alignment for a given type and then declare only individual stand-alone
6269 objects of that type, the compiler's ability to select a time-efficient
6270 alignment is primarily useful only when you plan to create arrays of
6271 variables having the relevant (efficiently aligned) type. If you
6272 declare or use arrays of variables of an efficiently-aligned type, then
6273 it is likely that your program also does pointer arithmetic (or
6274 subscripting, which amounts to the same thing) on pointers to the
6275 relevant type, and the code that the compiler generates for these
6276 pointer arithmetic operations is often more efficient for
6277 efficiently-aligned types than for other types.
6279 The @code{aligned} attribute can only increase the alignment; but you
6280 can decrease it by specifying @code{packed} as well. See below.
6282 Note that the effectiveness of @code{aligned} attributes may be limited
6283 by inherent limitations in your linker. On many systems, the linker is
6284 only able to arrange for variables to be aligned up to a certain maximum
6285 alignment. (For some linkers, the maximum supported alignment may
6286 be very very small.) If your linker is only able to align variables
6287 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6288 in an @code{__attribute__} still only provides you with 8-byte
6289 alignment. See your linker documentation for further information.
6291 @opindex fshort-enums
6292 Specifying this attribute for @code{struct} and @code{union} types is
6293 equivalent to specifying the @code{packed} attribute on each of the
6294 structure or union members. Specifying the @option{-fshort-enums}
6295 flag on the line is equivalent to specifying the @code{packed}
6296 attribute on all @code{enum} definitions.
6298 In the following example @code{struct my_packed_struct}'s members are
6299 packed closely together, but the internal layout of its @code{s} member
6300 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6304 struct my_unpacked_struct
6310 struct __attribute__ ((__packed__)) my_packed_struct
6314 struct my_unpacked_struct s;
6318 You may only specify this attribute on the definition of an @code{enum},
6319 @code{struct} or @code{union}, not on a @code{typedef} that does not
6320 also define the enumerated type, structure or union.
6322 @item bnd_variable_size
6323 @cindex @code{bnd_variable_size} type attribute
6324 @cindex Pointer Bounds Checker attributes
6325 When applied to a structure field, this attribute tells Pointer
6326 Bounds Checker that the size of this field should not be computed
6327 using static type information. It may be used to mark variably-sized
6328 static array fields placed at the end of a structure.
6336 S *p = (S *)malloc (sizeof(S) + 100);
6337 p->data[10] = 0; //Bounds violation
6341 By using an attribute for the field we may avoid unwanted bound
6348 char data[1] __attribute__((bnd_variable_size));
6350 S *p = (S *)malloc (sizeof(S) + 100);
6351 p->data[10] = 0; //OK
6355 @itemx deprecated (@var{msg})
6356 @cindex @code{deprecated} type attribute
6357 The @code{deprecated} attribute results in a warning if the type
6358 is used anywhere in the source file. This is useful when identifying
6359 types that are expected to be removed in a future version of a program.
6360 If possible, the warning also includes the location of the declaration
6361 of the deprecated type, to enable users to easily find further
6362 information about why the type is deprecated, or what they should do
6363 instead. Note that the warnings only occur for uses and then only
6364 if the type is being applied to an identifier that itself is not being
6365 declared as deprecated.
6368 typedef int T1 __attribute__ ((deprecated));
6372 typedef T1 T3 __attribute__ ((deprecated));
6373 T3 z __attribute__ ((deprecated));
6377 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6378 warning is issued for line 4 because T2 is not explicitly
6379 deprecated. Line 5 has no warning because T3 is explicitly
6380 deprecated. Similarly for line 6. The optional @var{msg}
6381 argument, which must be a string, is printed in the warning if
6384 The @code{deprecated} attribute can also be used for functions and
6385 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6387 @item designated_init
6388 @cindex @code{designated_init} type attribute
6389 This attribute may only be applied to structure types. It indicates
6390 that any initialization of an object of this type must use designated
6391 initializers rather than positional initializers. The intent of this
6392 attribute is to allow the programmer to indicate that a structure's
6393 layout may change, and that therefore relying on positional
6394 initialization will result in future breakage.
6396 GCC emits warnings based on this attribute by default; use
6397 @option{-Wno-designated-init} to suppress them.
6400 @cindex @code{may_alias} type attribute
6401 Accesses through pointers to types with this attribute are not subject
6402 to type-based alias analysis, but are instead assumed to be able to alias
6403 any other type of objects.
6404 In the context of section 6.5 paragraph 7 of the C99 standard,
6405 an lvalue expression
6406 dereferencing such a pointer is treated like having a character type.
6407 See @option{-fstrict-aliasing} for more information on aliasing issues.
6408 This extension exists to support some vector APIs, in which pointers to
6409 one vector type are permitted to alias pointers to a different vector type.
6411 Note that an object of a type with this attribute does not have any
6417 typedef short __attribute__((__may_alias__)) short_a;
6423 short_a *b = (short_a *) &a;
6427 if (a == 0x12345678)
6435 If you replaced @code{short_a} with @code{short} in the variable
6436 declaration, the above program would abort when compiled with
6437 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6441 @cindex @code{packed} type attribute
6442 This attribute, attached to @code{struct} or @code{union} type
6443 definition, specifies that each member (other than zero-width bit-fields)
6444 of the structure or union is placed to minimize the memory required. When
6445 attached to an @code{enum} definition, it indicates that the smallest
6446 integral type should be used.
6448 @item scalar_storage_order ("@var{endianness}")
6449 @cindex @code{scalar_storage_order} type attribute
6450 When attached to a @code{union} or a @code{struct}, this attribute sets
6451 the storage order, aka endianness, of the scalar fields of the type, as
6452 well as the array fields whose component is scalar. The supported
6453 endianness are @code{big-endian} and @code{little-endian}. The attribute
6454 has no effects on fields which are themselves a @code{union}, a @code{struct}
6455 or an array whose component is a @code{union} or a @code{struct}, and it is
6456 possible to have fields with a different scalar storage order than the
6459 This attribute is supported only for targets that use a uniform default
6460 scalar storage order (fortunately, most of them), i.e. targets that store
6461 the scalars either all in big-endian or all in little-endian.
6463 Additional restrictions are enforced for types with the reverse scalar
6464 storage order with regard to the scalar storage order of the target:
6467 @item Taking the address of a scalar field of a @code{union} or a
6468 @code{struct} with reverse scalar storage order is not permitted and will
6470 @item Taking the address of an array field, whose component is scalar, of
6471 a @code{union} or a @code{struct} with reverse scalar storage order is
6472 permitted but will yield a warning, unless @option{-Wno-scalar-storage-order}
6474 @item Taking the address of a @code{union} or a @code{struct} with reverse
6475 scalar storage order is permitted.
6478 These restrictions exist because the storage order attribute is lost when
6479 the address of a scalar or the address of an array with scalar component
6480 is taken, so storing indirectly through this address will generally not work.
6481 The second case is nevertheless allowed to be able to perform a block copy
6482 from or to the array.
6484 @item transparent_union
6485 @cindex @code{transparent_union} type attribute
6487 This attribute, attached to a @code{union} type definition, indicates
6488 that any function parameter having that union type causes calls to that
6489 function to be treated in a special way.
6491 First, the argument corresponding to a transparent union type can be of
6492 any type in the union; no cast is required. Also, if the union contains
6493 a pointer type, the corresponding argument can be a null pointer
6494 constant or a void pointer expression; and if the union contains a void
6495 pointer type, the corresponding argument can be any pointer expression.
6496 If the union member type is a pointer, qualifiers like @code{const} on
6497 the referenced type must be respected, just as with normal pointer
6500 Second, the argument is passed to the function using the calling
6501 conventions of the first member of the transparent union, not the calling
6502 conventions of the union itself. All members of the union must have the
6503 same machine representation; this is necessary for this argument passing
6506 Transparent unions are designed for library functions that have multiple
6507 interfaces for compatibility reasons. For example, suppose the
6508 @code{wait} function must accept either a value of type @code{int *} to
6509 comply with POSIX, or a value of type @code{union wait *} to comply with
6510 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6511 @code{wait} would accept both kinds of arguments, but it would also
6512 accept any other pointer type and this would make argument type checking
6513 less useful. Instead, @code{<sys/wait.h>} might define the interface
6517 typedef union __attribute__ ((__transparent_union__))
6521 @} wait_status_ptr_t;
6523 pid_t wait (wait_status_ptr_t);
6527 This interface allows either @code{int *} or @code{union wait *}
6528 arguments to be passed, using the @code{int *} calling convention.
6529 The program can call @code{wait} with arguments of either type:
6532 int w1 () @{ int w; return wait (&w); @}
6533 int w2 () @{ union wait w; return wait (&w); @}
6537 With this interface, @code{wait}'s implementation might look like this:
6540 pid_t wait (wait_status_ptr_t p)
6542 return waitpid (-1, p.__ip, 0);
6547 @cindex @code{unused} type attribute
6548 When attached to a type (including a @code{union} or a @code{struct}),
6549 this attribute means that variables of that type are meant to appear
6550 possibly unused. GCC does not produce a warning for any variables of
6551 that type, even if the variable appears to do nothing. This is often
6552 the case with lock or thread classes, which are usually defined and then
6553 not referenced, but contain constructors and destructors that have
6554 nontrivial bookkeeping functions.
6557 @cindex @code{visibility} type attribute
6558 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6559 applied to class, struct, union and enum types. Unlike other type
6560 attributes, the attribute must appear between the initial keyword and
6561 the name of the type; it cannot appear after the body of the type.
6563 Note that the type visibility is applied to vague linkage entities
6564 associated with the class (vtable, typeinfo node, etc.). In
6565 particular, if a class is thrown as an exception in one shared object
6566 and caught in another, the class must have default visibility.
6567 Otherwise the two shared objects are unable to use the same
6568 typeinfo node and exception handling will break.
6572 To specify multiple attributes, separate them by commas within the
6573 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6576 @node ARM Type Attributes
6577 @subsection ARM Type Attributes
6579 @cindex @code{notshared} type attribute, ARM
6580 On those ARM targets that support @code{dllimport} (such as Symbian
6581 OS), you can use the @code{notshared} attribute to indicate that the
6582 virtual table and other similar data for a class should not be
6583 exported from a DLL@. For example:
6586 class __declspec(notshared) C @{
6588 __declspec(dllimport) C();
6592 __declspec(dllexport)
6597 In this code, @code{C::C} is exported from the current DLL, but the
6598 virtual table for @code{C} is not exported. (You can use
6599 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6600 most Symbian OS code uses @code{__declspec}.)
6602 @node MeP Type Attributes
6603 @subsection MeP Type Attributes
6605 @cindex @code{based} type attribute, MeP
6606 @cindex @code{tiny} type attribute, MeP
6607 @cindex @code{near} type attribute, MeP
6608 @cindex @code{far} type attribute, MeP
6609 Many of the MeP variable attributes may be applied to types as well.
6610 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6611 @code{far} attributes may be applied to either. The @code{io} and
6612 @code{cb} attributes may not be applied to types.
6614 @node PowerPC Type Attributes
6615 @subsection PowerPC Type Attributes
6617 Three attributes currently are defined for PowerPC configurations:
6618 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6620 @cindex @code{ms_struct} type attribute, PowerPC
6621 @cindex @code{gcc_struct} type attribute, PowerPC
6622 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6623 attributes please see the documentation in @ref{x86 Type Attributes}.
6625 @cindex @code{altivec} type attribute, PowerPC
6626 The @code{altivec} attribute allows one to declare AltiVec vector data
6627 types supported by the AltiVec Programming Interface Manual. The
6628 attribute requires an argument to specify one of three vector types:
6629 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6630 and @code{bool__} (always followed by unsigned).
6633 __attribute__((altivec(vector__)))
6634 __attribute__((altivec(pixel__))) unsigned short
6635 __attribute__((altivec(bool__))) unsigned
6638 These attributes mainly are intended to support the @code{__vector},
6639 @code{__pixel}, and @code{__bool} AltiVec keywords.
6641 @node SPU Type Attributes
6642 @subsection SPU Type Attributes
6644 @cindex @code{spu_vector} type attribute, SPU
6645 The SPU supports the @code{spu_vector} attribute for types. This attribute
6646 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6647 Language Extensions Specification. It is intended to support the
6648 @code{__vector} keyword.
6650 @node x86 Type Attributes
6651 @subsection x86 Type Attributes
6653 Two attributes are currently defined for x86 configurations:
6654 @code{ms_struct} and @code{gcc_struct}.
6660 @cindex @code{ms_struct} type attribute, x86
6661 @cindex @code{gcc_struct} type attribute, x86
6663 If @code{packed} is used on a structure, or if bit-fields are used
6664 it may be that the Microsoft ABI packs them differently
6665 than GCC normally packs them. Particularly when moving packed
6666 data between functions compiled with GCC and the native Microsoft compiler
6667 (either via function call or as data in a file), it may be necessary to access
6670 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows x86
6671 compilers to match the native Microsoft compiler.
6674 @node Label Attributes
6675 @section Label Attributes
6676 @cindex Label Attributes
6678 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6679 details of the exact syntax for using attributes. Other attributes are
6680 available for functions (@pxref{Function Attributes}), variables
6681 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6682 and for types (@pxref{Type Attributes}).
6684 This example uses the @code{cold} label attribute to indicate the
6685 @code{ErrorHandling} branch is unlikely to be taken and that the
6686 @code{ErrorHandling} label is unused:
6690 asm goto ("some asm" : : : : NoError);
6692 /* This branch (the fall-through from the asm) is less commonly used */
6694 __attribute__((cold, unused)); /* Semi-colon is required here */
6699 printf("no error\n");
6705 @cindex @code{unused} label attribute
6706 This feature is intended for program-generated code that may contain
6707 unused labels, but which is compiled with @option{-Wall}. It is
6708 not normally appropriate to use in it human-written code, though it
6709 could be useful in cases where the code that jumps to the label is
6710 contained within an @code{#ifdef} conditional.
6713 @cindex @code{hot} label attribute
6714 The @code{hot} attribute on a label is used to inform the compiler that
6715 the path following the label is more likely than paths that are not so
6716 annotated. This attribute is used in cases where @code{__builtin_expect}
6717 cannot be used, for instance with computed goto or @code{asm goto}.
6720 @cindex @code{cold} label attribute
6721 The @code{cold} attribute on labels is used to inform the compiler that
6722 the path following the label is unlikely to be executed. This attribute
6723 is used in cases where @code{__builtin_expect} cannot be used, for instance
6724 with computed goto or @code{asm goto}.
6728 @node Enumerator Attributes
6729 @section Enumerator Attributes
6730 @cindex Enumerator Attributes
6732 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6733 details of the exact syntax for using attributes. Other attributes are
6734 available for functions (@pxref{Function Attributes}), variables
6735 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6736 and for types (@pxref{Type Attributes}).
6738 This example uses the @code{deprecated} enumerator attribute to indicate the
6739 @code{oldval} enumerator is deprecated:
6743 oldval __attribute__((deprecated)),
6756 @cindex @code{deprecated} enumerator attribute
6757 The @code{deprecated} attribute results in a warning if the enumerator
6758 is used anywhere in the source file. This is useful when identifying
6759 enumerators that are expected to be removed in a future version of a
6760 program. The warning also includes the location of the declaration
6761 of the deprecated enumerator, to enable users to easily find further
6762 information about why the enumerator is deprecated, or what they should
6763 do instead. Note that the warnings only occurs for uses.
6767 @node Attribute Syntax
6768 @section Attribute Syntax
6769 @cindex attribute syntax
6771 This section describes the syntax with which @code{__attribute__} may be
6772 used, and the constructs to which attribute specifiers bind, for the C
6773 language. Some details may vary for C++ and Objective-C@. Because of
6774 infelicities in the grammar for attributes, some forms described here
6775 may not be successfully parsed in all cases.
6777 There are some problems with the semantics of attributes in C++. For
6778 example, there are no manglings for attributes, although they may affect
6779 code generation, so problems may arise when attributed types are used in
6780 conjunction with templates or overloading. Similarly, @code{typeid}
6781 does not distinguish between types with different attributes. Support
6782 for attributes in C++ may be restricted in future to attributes on
6783 declarations only, but not on nested declarators.
6785 @xref{Function Attributes}, for details of the semantics of attributes
6786 applying to functions. @xref{Variable Attributes}, for details of the
6787 semantics of attributes applying to variables. @xref{Type Attributes},
6788 for details of the semantics of attributes applying to structure, union
6789 and enumerated types.
6790 @xref{Label Attributes}, for details of the semantics of attributes
6792 @xref{Enumerator Attributes}, for details of the semantics of attributes
6793 applying to enumerators.
6795 An @dfn{attribute specifier} is of the form
6796 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6797 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6798 each attribute is one of the following:
6802 Empty. Empty attributes are ignored.
6806 (which may be an identifier such as @code{unused}, or a reserved
6807 word such as @code{const}).
6810 An attribute name followed by a parenthesized list of
6811 parameters for the attribute.
6812 These parameters take one of the following forms:
6816 An identifier. For example, @code{mode} attributes use this form.
6819 An identifier followed by a comma and a non-empty comma-separated list
6820 of expressions. For example, @code{format} attributes use this form.
6823 A possibly empty comma-separated list of expressions. For example,
6824 @code{format_arg} attributes use this form with the list being a single
6825 integer constant expression, and @code{alias} attributes use this form
6826 with the list being a single string constant.
6830 An @dfn{attribute specifier list} is a sequence of one or more attribute
6831 specifiers, not separated by any other tokens.
6833 You may optionally specify attribute names with @samp{__}
6834 preceding and following the name.
6835 This allows you to use them in header files without
6836 being concerned about a possible macro of the same name. For example,
6837 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6840 @subsubheading Label Attributes
6842 In GNU C, an attribute specifier list may appear after the colon following a
6843 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6844 attributes on labels if the attribute specifier is immediately
6845 followed by a semicolon (i.e., the label applies to an empty
6846 statement). If the semicolon is missing, C++ label attributes are
6847 ambiguous, as it is permissible for a declaration, which could begin
6848 with an attribute list, to be labelled in C++. Declarations cannot be
6849 labelled in C90 or C99, so the ambiguity does not arise there.
6851 @subsubheading Enumerator Attributes
6853 In GNU C, an attribute specifier list may appear as part of an enumerator.
6854 The attribute goes after the enumeration constant, before @code{=}, if
6855 present. The optional attribute in the enumerator appertains to the
6856 enumeration constant. It is not possible to place the attribute after
6857 the constant expression, if present.
6859 @subsubheading Type Attributes
6861 An attribute specifier list may appear as part of a @code{struct},
6862 @code{union} or @code{enum} specifier. It may go either immediately
6863 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6864 the closing brace. The former syntax is preferred.
6865 Where attribute specifiers follow the closing brace, they are considered
6866 to relate to the structure, union or enumerated type defined, not to any
6867 enclosing declaration the type specifier appears in, and the type
6868 defined is not complete until after the attribute specifiers.
6869 @c Otherwise, there would be the following problems: a shift/reduce
6870 @c conflict between attributes binding the struct/union/enum and
6871 @c binding to the list of specifiers/qualifiers; and "aligned"
6872 @c attributes could use sizeof for the structure, but the size could be
6873 @c changed later by "packed" attributes.
6876 @subsubheading All other attributes
6878 Otherwise, an attribute specifier appears as part of a declaration,
6879 counting declarations of unnamed parameters and type names, and relates
6880 to that declaration (which may be nested in another declaration, for
6881 example in the case of a parameter declaration), or to a particular declarator
6882 within a declaration. Where an
6883 attribute specifier is applied to a parameter declared as a function or
6884 an array, it should apply to the function or array rather than the
6885 pointer to which the parameter is implicitly converted, but this is not
6886 yet correctly implemented.
6888 Any list of specifiers and qualifiers at the start of a declaration may
6889 contain attribute specifiers, whether or not such a list may in that
6890 context contain storage class specifiers. (Some attributes, however,
6891 are essentially in the nature of storage class specifiers, and only make
6892 sense where storage class specifiers may be used; for example,
6893 @code{section}.) There is one necessary limitation to this syntax: the
6894 first old-style parameter declaration in a function definition cannot
6895 begin with an attribute specifier, because such an attribute applies to
6896 the function instead by syntax described below (which, however, is not
6897 yet implemented in this case). In some other cases, attribute
6898 specifiers are permitted by this grammar but not yet supported by the
6899 compiler. All attribute specifiers in this place relate to the
6900 declaration as a whole. In the obsolescent usage where a type of
6901 @code{int} is implied by the absence of type specifiers, such a list of
6902 specifiers and qualifiers may be an attribute specifier list with no
6903 other specifiers or qualifiers.
6905 At present, the first parameter in a function prototype must have some
6906 type specifier that is not an attribute specifier; this resolves an
6907 ambiguity in the interpretation of @code{void f(int
6908 (__attribute__((foo)) x))}, but is subject to change. At present, if
6909 the parentheses of a function declarator contain only attributes then
6910 those attributes are ignored, rather than yielding an error or warning
6911 or implying a single parameter of type int, but this is subject to
6914 An attribute specifier list may appear immediately before a declarator
6915 (other than the first) in a comma-separated list of declarators in a
6916 declaration of more than one identifier using a single list of
6917 specifiers and qualifiers. Such attribute specifiers apply
6918 only to the identifier before whose declarator they appear. For
6922 __attribute__((noreturn)) void d0 (void),
6923 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6928 the @code{noreturn} attribute applies to all the functions
6929 declared; the @code{format} attribute only applies to @code{d1}.
6931 An attribute specifier list may appear immediately before the comma,
6932 @code{=} or semicolon terminating the declaration of an identifier other
6933 than a function definition. Such attribute specifiers apply
6934 to the declared object or function. Where an
6935 assembler name for an object or function is specified (@pxref{Asm
6936 Labels}), the attribute must follow the @code{asm}
6939 An attribute specifier list may, in future, be permitted to appear after
6940 the declarator in a function definition (before any old-style parameter
6941 declarations or the function body).
6943 Attribute specifiers may be mixed with type qualifiers appearing inside
6944 the @code{[]} of a parameter array declarator, in the C99 construct by
6945 which such qualifiers are applied to the pointer to which the array is
6946 implicitly converted. Such attribute specifiers apply to the pointer,
6947 not to the array, but at present this is not implemented and they are
6950 An attribute specifier list may appear at the start of a nested
6951 declarator. At present, there are some limitations in this usage: the
6952 attributes correctly apply to the declarator, but for most individual
6953 attributes the semantics this implies are not implemented.
6954 When attribute specifiers follow the @code{*} of a pointer
6955 declarator, they may be mixed with any type qualifiers present.
6956 The following describes the formal semantics of this syntax. It makes the
6957 most sense if you are familiar with the formal specification of
6958 declarators in the ISO C standard.
6960 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
6961 D1}, where @code{T} contains declaration specifiers that specify a type
6962 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
6963 contains an identifier @var{ident}. The type specified for @var{ident}
6964 for derived declarators whose type does not include an attribute
6965 specifier is as in the ISO C standard.
6967 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
6968 and the declaration @code{T D} specifies the type
6969 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6970 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6971 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
6973 If @code{D1} has the form @code{*
6974 @var{type-qualifier-and-attribute-specifier-list} D}, and the
6975 declaration @code{T D} specifies the type
6976 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
6977 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
6978 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
6984 void (__attribute__((noreturn)) ****f) (void);
6988 specifies the type ``pointer to pointer to pointer to pointer to
6989 non-returning function returning @code{void}''. As another example,
6992 char *__attribute__((aligned(8))) *f;
6996 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
6997 Note again that this does not work with most attributes; for example,
6998 the usage of @samp{aligned} and @samp{noreturn} attributes given above
6999 is not yet supported.
7001 For compatibility with existing code written for compiler versions that
7002 did not implement attributes on nested declarators, some laxity is
7003 allowed in the placing of attributes. If an attribute that only applies
7004 to types is applied to a declaration, it is treated as applying to
7005 the type of that declaration. If an attribute that only applies to
7006 declarations is applied to the type of a declaration, it is treated
7007 as applying to that declaration; and, for compatibility with code
7008 placing the attributes immediately before the identifier declared, such
7009 an attribute applied to a function return type is treated as
7010 applying to the function type, and such an attribute applied to an array
7011 element type is treated as applying to the array type. If an
7012 attribute that only applies to function types is applied to a
7013 pointer-to-function type, it is treated as applying to the pointer
7014 target type; if such an attribute is applied to a function return type
7015 that is not a pointer-to-function type, it is treated as applying
7016 to the function type.
7018 @node Function Prototypes
7019 @section Prototypes and Old-Style Function Definitions
7020 @cindex function prototype declarations
7021 @cindex old-style function definitions
7022 @cindex promotion of formal parameters
7024 GNU C extends ISO C to allow a function prototype to override a later
7025 old-style non-prototype definition. Consider the following example:
7028 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7035 /* @r{Prototype function declaration.} */
7036 int isroot P((uid_t));
7038 /* @r{Old-style function definition.} */
7040 isroot (x) /* @r{??? lossage here ???} */
7047 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7048 not allow this example, because subword arguments in old-style
7049 non-prototype definitions are promoted. Therefore in this example the
7050 function definition's argument is really an @code{int}, which does not
7051 match the prototype argument type of @code{short}.
7053 This restriction of ISO C makes it hard to write code that is portable
7054 to traditional C compilers, because the programmer does not know
7055 whether the @code{uid_t} type is @code{short}, @code{int}, or
7056 @code{long}. Therefore, in cases like these GNU C allows a prototype
7057 to override a later old-style definition. More precisely, in GNU C, a
7058 function prototype argument type overrides the argument type specified
7059 by a later old-style definition if the former type is the same as the
7060 latter type before promotion. Thus in GNU C the above example is
7061 equivalent to the following:
7074 GNU C++ does not support old-style function definitions, so this
7075 extension is irrelevant.
7078 @section C++ Style Comments
7080 @cindex C++ comments
7081 @cindex comments, C++ style
7083 In GNU C, you may use C++ style comments, which start with @samp{//} and
7084 continue until the end of the line. Many other C implementations allow
7085 such comments, and they are included in the 1999 C standard. However,
7086 C++ style comments are not recognized if you specify an @option{-std}
7087 option specifying a version of ISO C before C99, or @option{-ansi}
7088 (equivalent to @option{-std=c90}).
7091 @section Dollar Signs in Identifier Names
7093 @cindex dollar signs in identifier names
7094 @cindex identifier names, dollar signs in
7096 In GNU C, you may normally use dollar signs in identifier names.
7097 This is because many traditional C implementations allow such identifiers.
7098 However, dollar signs in identifiers are not supported on a few target
7099 machines, typically because the target assembler does not allow them.
7101 @node Character Escapes
7102 @section The Character @key{ESC} in Constants
7104 You can use the sequence @samp{\e} in a string or character constant to
7105 stand for the ASCII character @key{ESC}.
7108 @section Inquiring on Alignment of Types or Variables
7110 @cindex type alignment
7111 @cindex variable alignment
7113 The keyword @code{__alignof__} allows you to inquire about how an object
7114 is aligned, or the minimum alignment usually required by a type. Its
7115 syntax is just like @code{sizeof}.
7117 For example, if the target machine requires a @code{double} value to be
7118 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7119 This is true on many RISC machines. On more traditional machine
7120 designs, @code{__alignof__ (double)} is 4 or even 2.
7122 Some machines never actually require alignment; they allow reference to any
7123 data type even at an odd address. For these machines, @code{__alignof__}
7124 reports the smallest alignment that GCC gives the data type, usually as
7125 mandated by the target ABI.
7127 If the operand of @code{__alignof__} is an lvalue rather than a type,
7128 its value is the required alignment for its type, taking into account
7129 any minimum alignment specified with GCC's @code{__attribute__}
7130 extension (@pxref{Variable Attributes}). For example, after this
7134 struct foo @{ int x; char y; @} foo1;
7138 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7139 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7141 It is an error to ask for the alignment of an incomplete type.
7145 @section An Inline Function is As Fast As a Macro
7146 @cindex inline functions
7147 @cindex integrating function code
7149 @cindex macros, inline alternative
7151 By declaring a function inline, you can direct GCC to make
7152 calls to that function faster. One way GCC can achieve this is to
7153 integrate that function's code into the code for its callers. This
7154 makes execution faster by eliminating the function-call overhead; in
7155 addition, if any of the actual argument values are constant, their
7156 known values may permit simplifications at compile time so that not
7157 all of the inline function's code needs to be included. The effect on
7158 code size is less predictable; object code may be larger or smaller
7159 with function inlining, depending on the particular case. You can
7160 also direct GCC to try to integrate all ``simple enough'' functions
7161 into their callers with the option @option{-finline-functions}.
7163 GCC implements three different semantics of declaring a function
7164 inline. One is available with @option{-std=gnu89} or
7165 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7166 on all inline declarations, another when
7167 @option{-std=c99}, @option{-std=c11},
7168 @option{-std=gnu99} or @option{-std=gnu11}
7169 (without @option{-fgnu89-inline}), and the third
7170 is used when compiling C++.
7172 To declare a function inline, use the @code{inline} keyword in its
7173 declaration, like this:
7183 If you are writing a header file to be included in ISO C90 programs, write
7184 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7186 The three types of inlining behave similarly in two important cases:
7187 when the @code{inline} keyword is used on a @code{static} function,
7188 like the example above, and when a function is first declared without
7189 using the @code{inline} keyword and then is defined with
7190 @code{inline}, like this:
7193 extern int inc (int *a);
7201 In both of these common cases, the program behaves the same as if you
7202 had not used the @code{inline} keyword, except for its speed.
7204 @cindex inline functions, omission of
7205 @opindex fkeep-inline-functions
7206 When a function is both inline and @code{static}, if all calls to the
7207 function are integrated into the caller, and the function's address is
7208 never used, then the function's own assembler code is never referenced.
7209 In this case, GCC does not actually output assembler code for the
7210 function, unless you specify the option @option{-fkeep-inline-functions}.
7211 If there is a nonintegrated call, then the function is compiled to
7212 assembler code as usual. The function must also be compiled as usual if
7213 the program refers to its address, because that can't be inlined.
7216 Note that certain usages in a function definition can make it unsuitable
7217 for inline substitution. Among these usages are: variadic functions,
7218 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7219 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7220 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7221 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7222 function marked @code{inline} could not be substituted, and gives the
7223 reason for the failure.
7225 @cindex automatic @code{inline} for C++ member fns
7226 @cindex @code{inline} automatic for C++ member fns
7227 @cindex member fns, automatically @code{inline}
7228 @cindex C++ member fns, automatically @code{inline}
7229 @opindex fno-default-inline
7230 As required by ISO C++, GCC considers member functions defined within
7231 the body of a class to be marked inline even if they are
7232 not explicitly declared with the @code{inline} keyword. You can
7233 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7234 Options,,Options Controlling C++ Dialect}.
7236 GCC does not inline any functions when not optimizing unless you specify
7237 the @samp{always_inline} attribute for the function, like this:
7240 /* @r{Prototype.} */
7241 inline void foo (const char) __attribute__((always_inline));
7244 The remainder of this section is specific to GNU C90 inlining.
7246 @cindex non-static inline function
7247 When an inline function is not @code{static}, then the compiler must assume
7248 that there may be calls from other source files; since a global symbol can
7249 be defined only once in any program, the function must not be defined in
7250 the other source files, so the calls therein cannot be integrated.
7251 Therefore, a non-@code{static} inline function is always compiled on its
7252 own in the usual fashion.
7254 If you specify both @code{inline} and @code{extern} in the function
7255 definition, then the definition is used only for inlining. In no case
7256 is the function compiled on its own, not even if you refer to its
7257 address explicitly. Such an address becomes an external reference, as
7258 if you had only declared the function, and had not defined it.
7260 This combination of @code{inline} and @code{extern} has almost the
7261 effect of a macro. The way to use it is to put a function definition in
7262 a header file with these keywords, and put another copy of the
7263 definition (lacking @code{inline} and @code{extern}) in a library file.
7264 The definition in the header file causes most calls to the function
7265 to be inlined. If any uses of the function remain, they refer to
7266 the single copy in the library.
7269 @section When is a Volatile Object Accessed?
7270 @cindex accessing volatiles
7271 @cindex volatile read
7272 @cindex volatile write
7273 @cindex volatile access
7275 C has the concept of volatile objects. These are normally accessed by
7276 pointers and used for accessing hardware or inter-thread
7277 communication. The standard encourages compilers to refrain from
7278 optimizations concerning accesses to volatile objects, but leaves it
7279 implementation defined as to what constitutes a volatile access. The
7280 minimum requirement is that at a sequence point all previous accesses
7281 to volatile objects have stabilized and no subsequent accesses have
7282 occurred. Thus an implementation is free to reorder and combine
7283 volatile accesses that occur between sequence points, but cannot do
7284 so for accesses across a sequence point. The use of volatile does
7285 not allow you to violate the restriction on updating objects multiple
7286 times between two sequence points.
7288 Accesses to non-volatile objects are not ordered with respect to
7289 volatile accesses. You cannot use a volatile object as a memory
7290 barrier to order a sequence of writes to non-volatile memory. For
7294 int *ptr = @var{something};
7296 *ptr = @var{something};
7301 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7302 that the write to @var{*ptr} occurs by the time the update
7303 of @var{vobj} happens. If you need this guarantee, you must use
7304 a stronger memory barrier such as:
7307 int *ptr = @var{something};
7309 *ptr = @var{something};
7310 asm volatile ("" : : : "memory");
7314 A scalar volatile object is read when it is accessed in a void context:
7317 volatile int *src = @var{somevalue};
7321 Such expressions are rvalues, and GCC implements this as a
7322 read of the volatile object being pointed to.
7324 Assignments are also expressions and have an rvalue. However when
7325 assigning to a scalar volatile, the volatile object is not reread,
7326 regardless of whether the assignment expression's rvalue is used or
7327 not. If the assignment's rvalue is used, the value is that assigned
7328 to the volatile object. For instance, there is no read of @var{vobj}
7329 in all the following cases:
7334 vobj = @var{something};
7335 obj = vobj = @var{something};
7336 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7337 obj = (@var{something}, vobj = @var{anotherthing});
7340 If you need to read the volatile object after an assignment has
7341 occurred, you must use a separate expression with an intervening
7344 As bit-fields are not individually addressable, volatile bit-fields may
7345 be implicitly read when written to, or when adjacent bit-fields are
7346 accessed. Bit-field operations may be optimized such that adjacent
7347 bit-fields are only partially accessed, if they straddle a storage unit
7348 boundary. For these reasons it is unwise to use volatile bit-fields to
7351 @node Using Assembly Language with C
7352 @section How to Use Inline Assembly Language in C Code
7353 @cindex @code{asm} keyword
7354 @cindex assembly language in C
7355 @cindex inline assembly language
7356 @cindex mixing assembly language and C
7358 The @code{asm} keyword allows you to embed assembler instructions
7359 within C code. GCC provides two forms of inline @code{asm}
7360 statements. A @dfn{basic @code{asm}} statement is one with no
7361 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7362 statement (@pxref{Extended Asm}) includes one or more operands.
7363 The extended form is preferred for mixing C and assembly language
7364 within a function, but to include assembly language at
7365 top level you must use basic @code{asm}.
7367 You can also use the @code{asm} keyword to override the assembler name
7368 for a C symbol, or to place a C variable in a specific register.
7371 * Basic Asm:: Inline assembler without operands.
7372 * Extended Asm:: Inline assembler with operands.
7373 * Constraints:: Constraints for @code{asm} operands
7374 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7375 * Explicit Register Variables:: Defining variables residing in specified
7377 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7381 @subsection Basic Asm --- Assembler Instructions Without Operands
7382 @cindex basic @code{asm}
7383 @cindex assembly language in C, basic
7385 A basic @code{asm} statement has the following syntax:
7388 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7391 The @code{asm} keyword is a GNU extension.
7392 When writing code that can be compiled with @option{-ansi} and the
7393 various @option{-std} options, use @code{__asm__} instead of
7394 @code{asm} (@pxref{Alternate Keywords}).
7396 @subsubheading Qualifiers
7399 The optional @code{volatile} qualifier has no effect.
7400 All basic @code{asm} blocks are implicitly volatile.
7403 @subsubheading Parameters
7406 @item AssemblerInstructions
7407 This is a literal string that specifies the assembler code. The string can
7408 contain any instructions recognized by the assembler, including directives.
7409 GCC does not parse the assembler instructions themselves and
7410 does not know what they mean or even whether they are valid assembler input.
7412 You may place multiple assembler instructions together in a single @code{asm}
7413 string, separated by the characters normally used in assembly code for the
7414 system. A combination that works in most places is a newline to break the
7415 line, plus a tab character (written as @samp{\n\t}).
7416 Some assemblers allow semicolons as a line separator. However,
7417 note that some assembler dialects use semicolons to start a comment.
7420 @subsubheading Remarks
7421 Using extended @code{asm} typically produces smaller, safer, and more
7422 efficient code, and in most cases it is a better solution than basic
7423 @code{asm}. However, there are two situations where only basic @code{asm}
7428 Extended @code{asm} statements have to be inside a C
7429 function, so to write inline assembly language at file scope (``top-level''),
7430 outside of C functions, you must use basic @code{asm}.
7431 You can use this technique to emit assembler directives,
7432 define assembly language macros that can be invoked elsewhere in the file,
7433 or write entire functions in assembly language.
7437 with the @code{naked} attribute also require basic @code{asm}
7438 (@pxref{Function Attributes}).
7441 Safely accessing C data and calling functions from basic @code{asm} is more
7442 complex than it may appear. To access C data, it is better to use extended
7445 Do not expect a sequence of @code{asm} statements to remain perfectly
7446 consecutive after compilation. If certain instructions need to remain
7447 consecutive in the output, put them in a single multi-instruction @code{asm}
7448 statement. Note that GCC's optimizers can move @code{asm} statements
7449 relative to other code, including across jumps.
7451 @code{asm} statements may not perform jumps into other @code{asm} statements.
7452 GCC does not know about these jumps, and therefore cannot take
7453 account of them when deciding how to optimize. Jumps from @code{asm} to C
7454 labels are only supported in extended @code{asm}.
7456 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7457 assembly code when optimizing. This can lead to unexpected duplicate
7458 symbol errors during compilation if your assembly code defines symbols or
7461 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7462 visibility of any symbols it references. This may result in GCC discarding
7463 those symbols as unreferenced.
7465 The compiler copies the assembler instructions in a basic @code{asm}
7466 verbatim to the assembly language output file, without
7467 processing dialects or any of the @samp{%} operators that are available with
7468 extended @code{asm}. This results in minor differences between basic
7469 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7470 registers you might use @samp{%eax} in basic @code{asm} and
7471 @samp{%%eax} in extended @code{asm}.
7473 On targets such as x86 that support multiple assembler dialects,
7474 all basic @code{asm} blocks use the assembler dialect specified by the
7475 @option{-masm} command-line option (@pxref{x86 Options}).
7476 Basic @code{asm} provides no
7477 mechanism to provide different assembler strings for different dialects.
7479 Here is an example of basic @code{asm} for i386:
7482 /* Note that this code will not compile with -masm=intel */
7483 #define DebugBreak() asm("int $3")
7487 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7488 @cindex extended @code{asm}
7489 @cindex assembly language in C, extended
7491 With extended @code{asm} you can read and write C variables from
7492 assembler and perform jumps from assembler code to C labels.
7493 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7494 the operand parameters after the assembler template:
7497 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7498 : @var{OutputOperands}
7499 @r{[} : @var{InputOperands}
7500 @r{[} : @var{Clobbers} @r{]} @r{]})
7502 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7504 : @var{InputOperands}
7509 The @code{asm} keyword is a GNU extension.
7510 When writing code that can be compiled with @option{-ansi} and the
7511 various @option{-std} options, use @code{__asm__} instead of
7512 @code{asm} (@pxref{Alternate Keywords}).
7514 @subsubheading Qualifiers
7518 The typical use of extended @code{asm} statements is to manipulate input
7519 values to produce output values. However, your @code{asm} statements may
7520 also produce side effects. If so, you may need to use the @code{volatile}
7521 qualifier to disable certain optimizations. @xref{Volatile}.
7524 This qualifier informs the compiler that the @code{asm} statement may
7525 perform a jump to one of the labels listed in the @var{GotoLabels}.
7529 @subsubheading Parameters
7531 @item AssemblerTemplate
7532 This is a literal string that is the template for the assembler code. It is a
7533 combination of fixed text and tokens that refer to the input, output,
7534 and goto parameters. @xref{AssemblerTemplate}.
7536 @item OutputOperands
7537 A comma-separated list of the C variables modified by the instructions in the
7538 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7541 A comma-separated list of C expressions read by the instructions in the
7542 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7545 A comma-separated list of registers or other values changed by the
7546 @var{AssemblerTemplate}, beyond those listed as outputs.
7547 An empty list is permitted. @xref{Clobbers}.
7550 When you are using the @code{goto} form of @code{asm}, this section contains
7551 the list of all C labels to which the code in the
7552 @var{AssemblerTemplate} may jump.
7555 @code{asm} statements may not perform jumps into other @code{asm} statements,
7556 only to the listed @var{GotoLabels}.
7557 GCC's optimizers do not know about other jumps; therefore they cannot take
7558 account of them when deciding how to optimize.
7561 The total number of input + output + goto operands is limited to 30.
7563 @subsubheading Remarks
7564 The @code{asm} statement allows you to include assembly instructions directly
7565 within C code. This may help you to maximize performance in time-sensitive
7566 code or to access assembly instructions that are not readily available to C
7569 Note that extended @code{asm} statements must be inside a function. Only
7570 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7571 Functions declared with the @code{naked} attribute also require basic
7572 @code{asm} (@pxref{Function Attributes}).
7574 While the uses of @code{asm} are many and varied, it may help to think of an
7575 @code{asm} statement as a series of low-level instructions that convert input
7576 parameters to output parameters. So a simple (if not particularly useful)
7577 example for i386 using @code{asm} might look like this:
7583 asm ("mov %1, %0\n\t"
7588 printf("%d\n", dst);
7591 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7594 @subsubsection Volatile
7595 @cindex volatile @code{asm}
7596 @cindex @code{asm} volatile
7598 GCC's optimizers sometimes discard @code{asm} statements if they determine
7599 there is no need for the output variables. Also, the optimizers may move
7600 code out of loops if they believe that the code will always return the same
7601 result (i.e. none of its input values change between calls). Using the
7602 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7603 that have no output operands, including @code{asm goto} statements,
7604 are implicitly volatile.
7606 This i386 code demonstrates a case that does not use (or require) the
7607 @code{volatile} qualifier. If it is performing assertion checking, this code
7608 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7609 unreferenced by any code. As a result, the optimizers can discard the
7610 @code{asm} statement, which in turn removes the need for the entire
7611 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7612 isn't needed you allow the optimizers to produce the most efficient code
7616 void DoCheck(uint32_t dwSomeValue)
7620 // Assumes dwSomeValue is not zero.
7630 The next example shows a case where the optimizers can recognize that the input
7631 (@code{dwSomeValue}) never changes during the execution of the function and can
7632 therefore move the @code{asm} outside the loop to produce more efficient code.
7633 Again, using @code{volatile} disables this type of optimization.
7636 void do_print(uint32_t dwSomeValue)
7640 for (uint32_t x=0; x < 5; x++)
7642 // Assumes dwSomeValue is not zero.
7648 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7653 The following example demonstrates a case where you need to use the
7654 @code{volatile} qualifier.
7655 It uses the x86 @code{rdtsc} instruction, which reads
7656 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7657 the optimizers might assume that the @code{asm} block will always return the
7658 same value and therefore optimize away the second call.
7663 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7664 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7665 "or %%rdx, %0" // 'Or' in the lower bits.
7670 printf("msr: %llx\n", msr);
7674 // Reprint the timestamp
7675 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7676 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7677 "or %%rdx, %0" // 'Or' in the lower bits.
7682 printf("msr: %llx\n", msr);
7685 GCC's optimizers do not treat this code like the non-volatile code in the
7686 earlier examples. They do not move it out of loops or omit it on the
7687 assumption that the result from a previous call is still valid.
7689 Note that the compiler can move even volatile @code{asm} instructions relative
7690 to other code, including across jump instructions. For example, on many
7691 targets there is a system register that controls the rounding mode of
7692 floating-point operations. Setting it with a volatile @code{asm}, as in the
7693 following PowerPC example, does not work reliably.
7696 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7700 The compiler may move the addition back before the volatile @code{asm}. To
7701 make it work as expected, add an artificial dependency to the @code{asm} by
7702 referencing a variable in the subsequent code, for example:
7705 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7709 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7710 assembly code when optimizing. This can lead to unexpected duplicate symbol
7711 errors during compilation if your asm code defines symbols or labels.
7713 (@pxref{AssemblerTemplate}) may help resolve this problem.
7715 @anchor{AssemblerTemplate}
7716 @subsubsection Assembler Template
7717 @cindex @code{asm} assembler template
7719 An assembler template is a literal string containing assembler instructions.
7720 The compiler replaces tokens in the template that refer
7721 to inputs, outputs, and goto labels,
7722 and then outputs the resulting string to the assembler. The
7723 string can contain any instructions recognized by the assembler, including
7724 directives. GCC does not parse the assembler instructions
7725 themselves and does not know what they mean or even whether they are valid
7726 assembler input. However, it does count the statements
7727 (@pxref{Size of an asm}).
7729 You may place multiple assembler instructions together in a single @code{asm}
7730 string, separated by the characters normally used in assembly code for the
7731 system. A combination that works in most places is a newline to break the
7732 line, plus a tab character to move to the instruction field (written as
7734 Some assemblers allow semicolons as a line separator. However, note
7735 that some assembler dialects use semicolons to start a comment.
7737 Do not expect a sequence of @code{asm} statements to remain perfectly
7738 consecutive after compilation, even when you are using the @code{volatile}
7739 qualifier. If certain instructions need to remain consecutive in the output,
7740 put them in a single multi-instruction asm statement.
7742 Accessing data from C programs without using input/output operands (such as
7743 by using global symbols directly from the assembler template) may not work as
7744 expected. Similarly, calling functions directly from an assembler template
7745 requires a detailed understanding of the target assembler and ABI.
7747 Since GCC does not parse the assembler template,
7748 it has no visibility of any
7749 symbols it references. This may result in GCC discarding those symbols as
7750 unreferenced unless they are also listed as input, output, or goto operands.
7752 @subsubheading Special format strings
7754 In addition to the tokens described by the input, output, and goto operands,
7755 these tokens have special meanings in the assembler template:
7759 Outputs a single @samp{%} into the assembler code.
7762 Outputs a number that is unique to each instance of the @code{asm}
7763 statement in the entire compilation. This option is useful when creating local
7764 labels and referring to them multiple times in a single template that
7765 generates multiple assembler instructions.
7770 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7771 into the assembler code. When unescaped, these characters have special
7772 meaning to indicate multiple assembler dialects, as described below.
7775 @subsubheading Multiple assembler dialects in @code{asm} templates
7777 On targets such as x86, GCC supports multiple assembler dialects.
7778 The @option{-masm} option controls which dialect GCC uses as its
7779 default for inline assembler. The target-specific documentation for the
7780 @option{-masm} option contains the list of supported dialects, as well as the
7781 default dialect if the option is not specified. This information may be
7782 important to understand, since assembler code that works correctly when
7783 compiled using one dialect will likely fail if compiled using another.
7786 If your code needs to support multiple assembler dialects (for example, if
7787 you are writing public headers that need to support a variety of compilation
7788 options), use constructs of this form:
7791 @{ dialect0 | dialect1 | dialect2... @}
7794 This construct outputs @code{dialect0}
7795 when using dialect #0 to compile the code,
7796 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7797 braces than the number of dialects the compiler supports, the construct
7800 For example, if an x86 compiler supports two dialects
7801 (@samp{att}, @samp{intel}), an
7802 assembler template such as this:
7805 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7809 is equivalent to one of
7812 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7813 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7816 Using that same compiler, this code:
7819 "xchg@{l@}\t@{%%@}ebx, %1"
7823 corresponds to either
7826 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7827 "xchg\tebx, %1" @r{/* intel dialect */}
7830 There is no support for nesting dialect alternatives.
7832 @anchor{OutputOperands}
7833 @subsubsection Output Operands
7834 @cindex @code{asm} output operands
7836 An @code{asm} statement has zero or more output operands indicating the names
7837 of C variables modified by the assembler code.
7839 In this i386 example, @code{old} (referred to in the template string as
7840 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7841 (@code{%2}) is an input:
7846 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7847 "sbb %0,%0" // Use the CF to calculate old.
7848 : "=r" (old), "+rm" (*Base)
7855 Operands are separated by commas. Each operand has this format:
7858 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7862 @item asmSymbolicName
7863 Specifies a symbolic name for the operand.
7864 Reference the name in the assembler template
7865 by enclosing it in square brackets
7866 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7867 that contains the definition. Any valid C variable name is acceptable,
7868 including names already defined in the surrounding code. No two operands
7869 within the same @code{asm} statement can use the same symbolic name.
7871 When not using an @var{asmSymbolicName}, use the (zero-based) position
7873 in the list of operands in the assembler template. For example if there are
7874 three output operands, use @samp{%0} in the template to refer to the first,
7875 @samp{%1} for the second, and @samp{%2} for the third.
7878 A string constant specifying constraints on the placement of the operand;
7879 @xref{Constraints}, for details.
7881 Output constraints must begin with either @samp{=} (a variable overwriting an
7882 existing value) or @samp{+} (when reading and writing). When using
7883 @samp{=}, do not assume the location contains the existing value
7884 on entry to the @code{asm}, except
7885 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7887 After the prefix, there must be one or more additional constraints
7888 (@pxref{Constraints}) that describe where the value resides. Common
7889 constraints include @samp{r} for register and @samp{m} for memory.
7890 When you list more than one possible location (for example, @code{"=rm"}),
7891 the compiler chooses the most efficient one based on the current context.
7892 If you list as many alternates as the @code{asm} statement allows, you permit
7893 the optimizers to produce the best possible code.
7894 If you must use a specific register, but your Machine Constraints do not
7895 provide sufficient control to select the specific register you want,
7896 local register variables may provide a solution (@pxref{Local Register
7900 Specifies a C lvalue expression to hold the output, typically a variable name.
7901 The enclosing parentheses are a required part of the syntax.
7905 When the compiler selects the registers to use to
7906 represent the output operands, it does not use any of the clobbered registers
7909 Output operand expressions must be lvalues. The compiler cannot check whether
7910 the operands have data types that are reasonable for the instruction being
7911 executed. For output expressions that are not directly addressable (for
7912 example a bit-field), the constraint must allow a register. In that case, GCC
7913 uses the register as the output of the @code{asm}, and then stores that
7914 register into the output.
7916 Operands using the @samp{+} constraint modifier count as two operands
7917 (that is, both as input and output) towards the total maximum of 30 operands
7918 per @code{asm} statement.
7920 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7921 operands that must not overlap an input. Otherwise,
7922 GCC may allocate the output operand in the same register as an unrelated
7923 input operand, on the assumption that the assembler code consumes its
7924 inputs before producing outputs. This assumption may be false if the assembler
7925 code actually consists of more than one instruction.
7927 The same problem can occur if one output parameter (@var{a}) allows a register
7928 constraint and another output parameter (@var{b}) allows a memory constraint.
7929 The code generated by GCC to access the memory address in @var{b} can contain
7930 registers which @emph{might} be shared by @var{a}, and GCC considers those
7931 registers to be inputs to the asm. As above, GCC assumes that such input
7932 registers are consumed before any outputs are written. This assumption may
7933 result in incorrect behavior if the asm writes to @var{a} before using
7934 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7935 ensures that modifying @var{a} does not affect the address referenced by
7936 @var{b}. Otherwise, the location of @var{b}
7937 is undefined if @var{a} is modified before using @var{b}.
7939 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7940 instead of simply @samp{%2}). Typically these qualifiers are hardware
7941 dependent. The list of supported modifiers for x86 is found at
7942 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7944 If the C code that follows the @code{asm} makes no use of any of the output
7945 operands, use @code{volatile} for the @code{asm} statement to prevent the
7946 optimizers from discarding the @code{asm} statement as unneeded
7947 (see @ref{Volatile}).
7949 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7950 references the first output operand as @code{%0} (were there a second, it
7951 would be @code{%1}, etc). The number of the first input operand is one greater
7952 than that of the last output operand. In this i386 example, that makes
7953 @code{Mask} referenced as @code{%1}:
7956 uint32_t Mask = 1234;
7965 That code overwrites the variable @code{Index} (@samp{=}),
7966 placing the value in a register (@samp{r}).
7967 Using the generic @samp{r} constraint instead of a constraint for a specific
7968 register allows the compiler to pick the register to use, which can result
7969 in more efficient code. This may not be possible if an assembler instruction
7970 requires a specific register.
7972 The following i386 example uses the @var{asmSymbolicName} syntax.
7974 same result as the code above, but some may consider it more readable or more
7975 maintainable since reordering index numbers is not necessary when adding or
7976 removing operands. The names @code{aIndex} and @code{aMask}
7977 are only used in this example to emphasize which
7978 names get used where.
7979 It is acceptable to reuse the names @code{Index} and @code{Mask}.
7982 uint32_t Mask = 1234;
7985 asm ("bsfl %[aMask], %[aIndex]"
7986 : [aIndex] "=r" (Index)
7987 : [aMask] "r" (Mask)
7991 Here are some more examples of output operands.
7998 asm ("mov %[e], %[d]"
8003 Here, @code{d} may either be in a register or in memory. Since the compiler
8004 might already have the current value of the @code{uint32_t} location
8005 pointed to by @code{e}
8006 in a register, you can enable it to choose the best location
8007 for @code{d} by specifying both constraints.
8009 @anchor{FlagOutputOperands}
8010 @subsection Flag Output Operands
8011 @cindex @code{asm} flag output operands
8013 Some targets have a special register that holds the ``flags'' for the
8014 result of an operation or comparison. Normally, the contents of that
8015 register are either unmodifed by the asm, or the asm is considered to
8016 clobber the contents.
8018 On some targets, a special form of output operand exists by which
8019 conditions in the flags register may be outputs of the asm. The set of
8020 conditions supported are target specific, but the general rule is that
8021 the output variable must be a scalar integer, and the value will be boolean.
8022 When supported, the target will define the preprocessor symbol
8023 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8025 Because of the special nature of the flag output operands, the constraint
8026 may not include alternatives.
8028 Most often, the target has only one flags register, and thus is an implied
8029 operand of many instructions. In this case, the operand should not be
8030 referenced within the assembler template via @code{%0} etc, as there's
8031 no corresponding text in the assembly language.
8035 The flag output constraints for the x86 family are of the form
8036 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8037 conditions defined in the ISA manual for @code{j@var{cc}} or
8042 ``above'' or unsigned greater than
8044 ``above or equal'' or unsigned greater than or equal
8046 ``below'' or unsigned less than
8048 ``below or equal'' or unsigned less than or equal
8053 ``equal'' or zero flag set
8057 signed greater than or equal
8061 signed less than or equal
8082 ``not'' @var{flag}, or inverted versions of those above
8087 @anchor{InputOperands}
8088 @subsubsection Input Operands
8089 @cindex @code{asm} input operands
8090 @cindex @code{asm} expressions
8092 Input operands make values from C variables and expressions available to the
8095 Operands are separated by commas. Each operand has this format:
8098 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8102 @item asmSymbolicName
8103 Specifies a symbolic name for the operand.
8104 Reference the name in the assembler template
8105 by enclosing it in square brackets
8106 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8107 that contains the definition. Any valid C variable name is acceptable,
8108 including names already defined in the surrounding code. No two operands
8109 within the same @code{asm} statement can use the same symbolic name.
8111 When not using an @var{asmSymbolicName}, use the (zero-based) position
8113 in the list of operands in the assembler template. For example if there are
8114 two output operands and three inputs,
8115 use @samp{%2} in the template to refer to the first input operand,
8116 @samp{%3} for the second, and @samp{%4} for the third.
8119 A string constant specifying constraints on the placement of the operand;
8120 @xref{Constraints}, for details.
8122 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8123 When you list more than one possible location (for example, @samp{"irm"}),
8124 the compiler chooses the most efficient one based on the current context.
8125 If you must use a specific register, but your Machine Constraints do not
8126 provide sufficient control to select the specific register you want,
8127 local register variables may provide a solution (@pxref{Local Register
8130 Input constraints can also be digits (for example, @code{"0"}). This indicates
8131 that the specified input must be in the same place as the output constraint
8132 at the (zero-based) index in the output constraint list.
8133 When using @var{asmSymbolicName} syntax for the output operands,
8134 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8137 This is the C variable or expression being passed to the @code{asm} statement
8138 as input. The enclosing parentheses are a required part of the syntax.
8142 When the compiler selects the registers to use to represent the input
8143 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8145 If there are no output operands but there are input operands, place two
8146 consecutive colons where the output operands would go:
8149 __asm__ ("some instructions"
8151 : "r" (Offset / 8));
8154 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8155 (except for inputs tied to outputs). The compiler assumes that on exit from
8156 the @code{asm} statement these operands contain the same values as they
8157 had before executing the statement.
8158 It is @emph{not} possible to use clobbers
8159 to inform the compiler that the values in these inputs are changing. One
8160 common work-around is to tie the changing input variable to an output variable
8161 that never gets used. Note, however, that if the code that follows the
8162 @code{asm} statement makes no use of any of the output operands, the GCC
8163 optimizers may discard the @code{asm} statement as unneeded
8164 (see @ref{Volatile}).
8166 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8167 instead of simply @samp{%2}). Typically these qualifiers are hardware
8168 dependent. The list of supported modifiers for x86 is found at
8169 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8171 In this example using the fictitious @code{combine} instruction, the
8172 constraint @code{"0"} for input operand 1 says that it must occupy the same
8173 location as output operand 0. Only input operands may use numbers in
8174 constraints, and they must each refer to an output operand. Only a number (or
8175 the symbolic assembler name) in the constraint can guarantee that one operand
8176 is in the same place as another. The mere fact that @code{foo} is the value of
8177 both operands is not enough to guarantee that they are in the same place in
8178 the generated assembler code.
8181 asm ("combine %2, %0"
8183 : "0" (foo), "g" (bar));
8186 Here is an example using symbolic names.
8189 asm ("cmoveq %1, %2, %[result]"
8190 : [result] "=r"(result)
8191 : "r" (test), "r" (new), "[result]" (old));
8195 @subsubsection Clobbers
8196 @cindex @code{asm} clobbers
8198 While the compiler is aware of changes to entries listed in the output
8199 operands, the inline @code{asm} code may modify more than just the outputs. For
8200 example, calculations may require additional registers, or the processor may
8201 overwrite a register as a side effect of a particular assembler instruction.
8202 In order to inform the compiler of these changes, list them in the clobber
8203 list. Clobber list items are either register names or the special clobbers
8204 (listed below). Each clobber list item is a string constant
8205 enclosed in double quotes and separated by commas.
8207 Clobber descriptions may not in any way overlap with an input or output
8208 operand. For example, you may not have an operand describing a register class
8209 with one member when listing that register in the clobber list. Variables
8210 declared to live in specific registers (@pxref{Explicit Register
8211 Variables}) and used
8212 as @code{asm} input or output operands must have no part mentioned in the
8213 clobber description. In particular, there is no way to specify that input
8214 operands get modified without also specifying them as output operands.
8216 When the compiler selects which registers to use to represent input and output
8217 operands, it does not use any of the clobbered registers. As a result,
8218 clobbered registers are available for any use in the assembler code.
8220 Here is a realistic example for the VAX showing the use of clobbered
8224 asm volatile ("movc3 %0, %1, %2"
8226 : "g" (from), "g" (to), "g" (count)
8227 : "r0", "r1", "r2", "r3", "r4", "r5");
8230 Also, there are two special clobber arguments:
8234 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8235 register. On some machines, GCC represents the condition codes as a specific
8236 hardware register; @code{"cc"} serves to name this register.
8237 On other machines, condition code handling is different,
8238 and specifying @code{"cc"} has no effect. But
8239 it is valid no matter what the target.
8242 The @code{"memory"} clobber tells the compiler that the assembly code
8244 reads or writes to items other than those listed in the input and output
8245 operands (for example, accessing the memory pointed to by one of the input
8246 parameters). To ensure memory contains correct values, GCC may need to flush
8247 specific register values to memory before executing the @code{asm}. Further,
8248 the compiler does not assume that any values read from memory before an
8249 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8251 Using the @code{"memory"} clobber effectively forms a read/write
8252 memory barrier for the compiler.
8254 Note that this clobber does not prevent the @emph{processor} from doing
8255 speculative reads past the @code{asm} statement. To prevent that, you need
8256 processor-specific fence instructions.
8258 Flushing registers to memory has performance implications and may be an issue
8259 for time-sensitive code. You can use a trick to avoid this if the size of
8260 the memory being accessed is known at compile time. For example, if accessing
8261 ten bytes of a string, use a memory input like:
8263 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8268 @subsubsection Goto Labels
8269 @cindex @code{asm} goto labels
8271 @code{asm goto} allows assembly code to jump to one or more C labels. The
8272 @var{GotoLabels} section in an @code{asm goto} statement contains
8274 list of all C labels to which the assembler code may jump. GCC assumes that
8275 @code{asm} execution falls through to the next statement (if this is not the
8276 case, consider using the @code{__builtin_unreachable} intrinsic after the
8277 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8278 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8281 An @code{asm goto} statement cannot have outputs.
8282 This is due to an internal restriction of
8283 the compiler: control transfer instructions cannot have outputs.
8284 If the assembler code does modify anything, use the @code{"memory"} clobber
8286 optimizers to flush all register values to memory and reload them if
8287 necessary after the @code{asm} statement.
8289 Also note that an @code{asm goto} statement is always implicitly
8290 considered volatile.
8292 To reference a label in the assembler template,
8293 prefix it with @samp{%l} (lowercase @samp{L}) followed
8294 by its (zero-based) position in @var{GotoLabels} plus the number of input
8295 operands. For example, if the @code{asm} has three inputs and references two
8296 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8298 Alternately, you can reference labels using the actual C label name enclosed
8299 in brackets. For example, to reference a label named @code{carry}, you can
8300 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8301 section when using this approach.
8303 Here is an example of @code{asm goto} for i386:
8310 : "r" (p1), "r" (p2)
8320 The following example shows an @code{asm goto} that uses a memory clobber.
8326 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8337 @anchor{x86Operandmodifiers}
8338 @subsubsection x86 Operand Modifiers
8340 References to input, output, and goto operands in the assembler template
8341 of extended @code{asm} statements can use
8342 modifiers to affect the way the operands are formatted in
8343 the code output to the assembler. For example, the
8344 following code uses the @samp{h} and @samp{b} modifiers for x86:
8348 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8352 These modifiers generate this assembler code:
8358 The rest of this discussion uses the following code for illustrative purposes.
8367 asm volatile goto ("some assembler instructions here"
8369 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8370 : /* No clobbers. */
8375 With no modifiers, this is what the output from the operands would be for the
8376 @samp{att} and @samp{intel} dialects of assembler:
8378 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8379 @headitem Operand @tab masm=att @tab masm=intel
8388 @tab @code{OFFSET FLAT:.L2}
8391 The table below shows the list of supported modifiers and their effects.
8393 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8394 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8396 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8401 @tab Print the QImode name of the register.
8406 @tab Print the QImode name for a ``high'' register.
8411 @tab Print the HImode name of the register.
8416 @tab Print the SImode name of the register.
8421 @tab Print the DImode name of the register.
8426 @tab Print the label name with no punctuation.
8431 @tab Require a constant operand and print the constant expression with no punctuation.
8437 @anchor{x86floatingpointasmoperands}
8438 @subsubsection x86 Floating-Point @code{asm} Operands
8440 On x86 targets, there are several rules on the usage of stack-like registers
8441 in the operands of an @code{asm}. These rules apply only to the operands
8442 that are stack-like registers:
8446 Given a set of input registers that die in an @code{asm}, it is
8447 necessary to know which are implicitly popped by the @code{asm}, and
8448 which must be explicitly popped by GCC@.
8450 An input register that is implicitly popped by the @code{asm} must be
8451 explicitly clobbered, unless it is constrained to match an
8455 For any input register that is implicitly popped by an @code{asm}, it is
8456 necessary to know how to adjust the stack to compensate for the pop.
8457 If any non-popped input is closer to the top of the reg-stack than
8458 the implicitly popped register, it would not be possible to know what the
8459 stack looked like---it's not clear how the rest of the stack ``slides
8462 All implicitly popped input registers must be closer to the top of
8463 the reg-stack than any input that is not implicitly popped.
8465 It is possible that if an input dies in an @code{asm}, the compiler might
8466 use the input register for an output reload. Consider this example:
8469 asm ("foo" : "=t" (a) : "f" (b));
8473 This code says that input @code{b} is not popped by the @code{asm}, and that
8474 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8475 deeper after the @code{asm} than it was before. But, it is possible that
8476 reload may think that it can use the same register for both the input and
8479 To prevent this from happening,
8480 if any input operand uses the @samp{f} constraint, all output register
8481 constraints must use the @samp{&} early-clobber modifier.
8483 The example above is correctly written as:
8486 asm ("foo" : "=&t" (a) : "f" (b));
8490 Some operands need to be in particular places on the stack. All
8491 output operands fall in this category---GCC has no other way to
8492 know which registers the outputs appear in unless you indicate
8493 this in the constraints.
8495 Output operands must specifically indicate which register an output
8496 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8497 constraints must select a class with a single register.
8500 Output operands may not be ``inserted'' between existing stack registers.
8501 Since no 387 opcode uses a read/write operand, all output operands
8502 are dead before the @code{asm}, and are pushed by the @code{asm}.
8503 It makes no sense to push anywhere but the top of the reg-stack.
8505 Output operands must start at the top of the reg-stack: output
8506 operands may not ``skip'' a register.
8509 Some @code{asm} statements may need extra stack space for internal
8510 calculations. This can be guaranteed by clobbering stack registers
8511 unrelated to the inputs and outputs.
8516 takes one input, which is internally popped, and produces two outputs.
8519 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8523 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8524 and replaces them with one output. The @code{st(1)} clobber is necessary
8525 for the compiler to know that @code{fyl2xp1} pops both inputs.
8528 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8536 @subsection Controlling Names Used in Assembler Code
8537 @cindex assembler names for identifiers
8538 @cindex names used in assembler code
8539 @cindex identifiers, names in assembler code
8541 You can specify the name to be used in the assembler code for a C
8542 function or variable by writing the @code{asm} (or @code{__asm__})
8543 keyword after the declarator.
8544 It is up to you to make sure that the assembler names you choose do not
8545 conflict with any other assembler symbols, or reference registers.
8547 @subsubheading Assembler names for data:
8549 This sample shows how to specify the assembler name for data:
8552 int foo asm ("myfoo") = 2;
8556 This specifies that the name to be used for the variable @code{foo} in
8557 the assembler code should be @samp{myfoo} rather than the usual
8560 On systems where an underscore is normally prepended to the name of a C
8561 variable, this feature allows you to define names for the
8562 linker that do not start with an underscore.
8564 GCC does not support using this feature with a non-static local variable
8565 since such variables do not have assembler names. If you are
8566 trying to put the variable in a particular register, see
8567 @ref{Explicit Register Variables}.
8569 @subsubheading Assembler names for functions:
8571 To specify the assembler name for functions, write a declaration for the
8572 function before its definition and put @code{asm} there, like this:
8575 int func (int x, int y) asm ("MYFUNC");
8577 int func (int x, int y)
8583 This specifies that the name to be used for the function @code{func} in
8584 the assembler code should be @code{MYFUNC}.
8586 @node Explicit Register Variables
8587 @subsection Variables in Specified Registers
8588 @anchor{Explicit Reg Vars}
8589 @cindex explicit register variables
8590 @cindex variables in specified registers
8591 @cindex specified registers
8593 GNU C allows you to associate specific hardware registers with C
8594 variables. In almost all cases, allowing the compiler to assign
8595 registers produces the best code. However under certain unusual
8596 circumstances, more precise control over the variable storage is
8599 Both global and local variables can be associated with a register. The
8600 consequences of performing this association are very different between
8601 the two, as explained in the sections below.
8604 * Global Register Variables:: Variables declared at global scope.
8605 * Local Register Variables:: Variables declared within a function.
8608 @node Global Register Variables
8609 @subsubsection Defining Global Register Variables
8610 @anchor{Global Reg Vars}
8611 @cindex global register variables
8612 @cindex registers, global variables in
8613 @cindex registers, global allocation
8615 You can define a global register variable and associate it with a specified
8619 register int *foo asm ("r12");
8623 Here @code{r12} is the name of the register that should be used. Note that
8624 this is the same syntax used for defining local register variables, but for
8625 a global variable the declaration appears outside a function. The
8626 @code{register} keyword is required, and cannot be combined with
8627 @code{static}. The register name must be a valid register name for the
8630 Registers are a scarce resource on most systems and allowing the
8631 compiler to manage their usage usually results in the best code. However,
8632 under special circumstances it can make sense to reserve some globally.
8633 For example this may be useful in programs such as programming language
8634 interpreters that have a couple of global variables that are accessed
8637 After defining a global register variable, for the current compilation
8641 @item The register is reserved entirely for this use, and will not be
8642 allocated for any other purpose.
8643 @item The register is not saved and restored by any functions.
8644 @item Stores into this register are never deleted even if they appear to be
8645 dead, but references may be deleted, moved or simplified.
8648 Note that these points @emph{only} apply to code that is compiled with the
8649 definition. The behavior of code that is merely linked in (for example
8650 code from libraries) is not affected.
8652 If you want to recompile source files that do not actually use your global
8653 register variable so they do not use the specified register for any other
8654 purpose, you need not actually add the global register declaration to
8655 their source code. It suffices to specify the compiler option
8656 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8659 @subsubheading Declaring the variable
8661 Global register variables can not have initial values, because an
8662 executable file has no means to supply initial contents for a register.
8664 When selecting a register, choose one that is normally saved and
8665 restored by function calls on your machine. This ensures that code
8666 which is unaware of this reservation (such as library routines) will
8667 restore it before returning.
8669 On machines with register windows, be sure to choose a global
8670 register that is not affected magically by the function call mechanism.
8672 @subsubheading Using the variable
8674 @cindex @code{qsort}, and global register variables
8675 When calling routines that are not aware of the reservation, be
8676 cautious if those routines call back into code which uses them. As an
8677 example, if you call the system library version of @code{qsort}, it may
8678 clobber your registers during execution, but (if you have selected
8679 appropriate registers) it will restore them before returning. However
8680 it will @emph{not} restore them before calling @code{qsort}'s comparison
8681 function. As a result, global values will not reliably be available to
8682 the comparison function unless the @code{qsort} function itself is rebuilt.
8684 Similarly, it is not safe to access the global register variables from signal
8685 handlers or from more than one thread of control. Unless you recompile
8686 them specially for the task at hand, the system library routines may
8687 temporarily use the register for other things.
8689 @cindex register variable after @code{longjmp}
8690 @cindex global register after @code{longjmp}
8691 @cindex value after @code{longjmp}
8694 On most machines, @code{longjmp} restores to each global register
8695 variable the value it had at the time of the @code{setjmp}. On some
8696 machines, however, @code{longjmp} does not change the value of global
8697 register variables. To be portable, the function that called @code{setjmp}
8698 should make other arrangements to save the values of the global register
8699 variables, and to restore them in a @code{longjmp}. This way, the same
8700 thing happens regardless of what @code{longjmp} does.
8702 Eventually there may be a way of asking the compiler to choose a register
8703 automatically, but first we need to figure out how it should choose and
8704 how to enable you to guide the choice. No solution is evident.
8706 @node Local Register Variables
8707 @subsubsection Specifying Registers for Local Variables
8708 @anchor{Local Reg Vars}
8709 @cindex local variables, specifying registers
8710 @cindex specifying registers for local variables
8711 @cindex registers for local variables
8713 You can define a local register variable and associate it with a specified
8717 register int *foo asm ("r12");
8721 Here @code{r12} is the name of the register that should be used. Note
8722 that this is the same syntax used for defining global register variables,
8723 but for a local variable the declaration appears within a function. The
8724 @code{register} keyword is required, and cannot be combined with
8725 @code{static}. The register name must be a valid register name for the
8728 As with global register variables, it is recommended that you choose
8729 a register that is normally saved and restored by function calls on your
8730 machine, so that calls to library routines will not clobber it.
8732 The only supported use for this feature is to specify registers
8733 for input and output operands when calling Extended @code{asm}
8734 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8735 particular machine don't provide sufficient control to select the desired
8736 register. To force an operand into a register, create a local variable
8737 and specify the register name after the variable's declaration. Then use
8738 the local variable for the @code{asm} operand and specify any constraint
8739 letter that matches the register:
8742 register int *p1 asm ("r0") = @dots{};
8743 register int *p2 asm ("r1") = @dots{};
8744 register int *result asm ("r0");
8745 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8748 @emph{Warning:} In the above example, be aware that a register (for example
8749 @code{r0}) can be call-clobbered by subsequent code, including function
8750 calls and library calls for arithmetic operators on other variables (for
8751 example the initialization of @code{p2}). In this case, use temporary
8752 variables for expressions between the register assignments:
8756 register int *p1 asm ("r0") = @dots{};
8757 register int *p2 asm ("r1") = t1;
8758 register int *result asm ("r0");
8759 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8762 Defining a register variable does not reserve the register. Other than
8763 when invoking the Extended @code{asm}, the contents of the specified
8764 register are not guaranteed. For this reason, the following uses
8765 are explicitly @emph{not} supported. If they appear to work, it is only
8766 happenstance, and may stop working as intended due to (seemingly)
8767 unrelated changes in surrounding code, or even minor changes in the
8768 optimization of a future version of gcc:
8771 @item Passing parameters to or from Basic @code{asm}
8772 @item Passing parameters to or from Extended @code{asm} without using input
8774 @item Passing parameters to or from routines written in assembler (or
8775 other languages) using non-standard calling conventions.
8778 Some developers use Local Register Variables in an attempt to improve
8779 gcc's allocation of registers, especially in large functions. In this
8780 case the register name is essentially a hint to the register allocator.
8781 While in some instances this can generate better code, improvements are
8782 subject to the whims of the allocator/optimizers. Since there are no
8783 guarantees that your improvements won't be lost, this usage of Local
8784 Register Variables is discouraged.
8786 On the MIPS platform, there is related use for local register variables
8787 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8788 Defining coprocessor specifics for MIPS targets, gccint,
8789 GNU Compiler Collection (GCC) Internals}).
8791 @node Size of an asm
8792 @subsection Size of an @code{asm}
8794 Some targets require that GCC track the size of each instruction used
8795 in order to generate correct code. Because the final length of the
8796 code produced by an @code{asm} statement is only known by the
8797 assembler, GCC must make an estimate as to how big it will be. It
8798 does this by counting the number of instructions in the pattern of the
8799 @code{asm} and multiplying that by the length of the longest
8800 instruction supported by that processor. (When working out the number
8801 of instructions, it assumes that any occurrence of a newline or of
8802 whatever statement separator character is supported by the assembler --
8803 typically @samp{;} --- indicates the end of an instruction.)
8805 Normally, GCC's estimate is adequate to ensure that correct
8806 code is generated, but it is possible to confuse the compiler if you use
8807 pseudo instructions or assembler macros that expand into multiple real
8808 instructions, or if you use assembler directives that expand to more
8809 space in the object file than is needed for a single instruction.
8810 If this happens then the assembler may produce a diagnostic saying that
8811 a label is unreachable.
8813 @node Alternate Keywords
8814 @section Alternate Keywords
8815 @cindex alternate keywords
8816 @cindex keywords, alternate
8818 @option{-ansi} and the various @option{-std} options disable certain
8819 keywords. This causes trouble when you want to use GNU C extensions, or
8820 a general-purpose header file that should be usable by all programs,
8821 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8822 @code{inline} are not available in programs compiled with
8823 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8824 program compiled with @option{-std=c99} or @option{-std=c11}). The
8826 @code{restrict} is only available when @option{-std=gnu99} (which will
8827 eventually be the default) or @option{-std=c99} (or the equivalent
8828 @option{-std=iso9899:1999}), or an option for a later standard
8831 The way to solve these problems is to put @samp{__} at the beginning and
8832 end of each problematical keyword. For example, use @code{__asm__}
8833 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8835 Other C compilers won't accept these alternative keywords; if you want to
8836 compile with another compiler, you can define the alternate keywords as
8837 macros to replace them with the customary keywords. It looks like this:
8845 @findex __extension__
8847 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8849 prevent such warnings within one expression by writing
8850 @code{__extension__} before the expression. @code{__extension__} has no
8851 effect aside from this.
8853 @node Incomplete Enums
8854 @section Incomplete @code{enum} Types
8856 You can define an @code{enum} tag without specifying its possible values.
8857 This results in an incomplete type, much like what you get if you write
8858 @code{struct foo} without describing the elements. A later declaration
8859 that does specify the possible values completes the type.
8861 You can't allocate variables or storage using the type while it is
8862 incomplete. However, you can work with pointers to that type.
8864 This extension may not be very useful, but it makes the handling of
8865 @code{enum} more consistent with the way @code{struct} and @code{union}
8868 This extension is not supported by GNU C++.
8870 @node Function Names
8871 @section Function Names as Strings
8872 @cindex @code{__func__} identifier
8873 @cindex @code{__FUNCTION__} identifier
8874 @cindex @code{__PRETTY_FUNCTION__} identifier
8876 GCC provides three magic variables that hold the name of the current
8877 function, as a string. The first of these is @code{__func__}, which
8878 is part of the C99 standard:
8880 The identifier @code{__func__} is implicitly declared by the translator
8881 as if, immediately following the opening brace of each function
8882 definition, the declaration
8885 static const char __func__[] = "function-name";
8889 appeared, where function-name is the name of the lexically-enclosing
8890 function. This name is the unadorned name of the function.
8892 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8893 backward compatibility with old versions of GCC.
8895 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8896 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8897 the type signature of the function as well as its bare name. For
8898 example, this program:
8902 extern int printf (char *, ...);
8909 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8910 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8928 __PRETTY_FUNCTION__ = void a::sub(int)
8931 These identifiers are variables, not preprocessor macros, and may not
8932 be used to initialize @code{char} arrays or be concatenated with other string
8935 @node Return Address
8936 @section Getting the Return or Frame Address of a Function
8938 These functions may be used to get information about the callers of a
8941 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8942 This function returns the return address of the current function, or of
8943 one of its callers. The @var{level} argument is number of frames to
8944 scan up the call stack. A value of @code{0} yields the return address
8945 of the current function, a value of @code{1} yields the return address
8946 of the caller of the current function, and so forth. When inlining
8947 the expected behavior is that the function returns the address of
8948 the function that is returned to. To work around this behavior use
8949 the @code{noinline} function attribute.
8951 The @var{level} argument must be a constant integer.
8953 On some machines it may be impossible to determine the return address of
8954 any function other than the current one; in such cases, or when the top
8955 of the stack has been reached, this function returns @code{0} or a
8956 random value. In addition, @code{__builtin_frame_address} may be used
8957 to determine if the top of the stack has been reached.
8959 Additional post-processing of the returned value may be needed, see
8960 @code{__builtin_extract_return_addr}.
8962 Calling this function with a nonzero argument can have unpredictable
8963 effects, including crashing the calling program. As a result, calls
8964 that are considered unsafe are diagnosed when the @option{-Wframe-address}
8965 option is in effect. Such calls should only be made in debugging
8969 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
8970 The address as returned by @code{__builtin_return_address} may have to be fed
8971 through this function to get the actual encoded address. For example, on the
8972 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
8973 platforms an offset has to be added for the true next instruction to be
8976 If no fixup is needed, this function simply passes through @var{addr}.
8979 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
8980 This function does the reverse of @code{__builtin_extract_return_addr}.
8983 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
8984 This function is similar to @code{__builtin_return_address}, but it
8985 returns the address of the function frame rather than the return address
8986 of the function. Calling @code{__builtin_frame_address} with a value of
8987 @code{0} yields the frame address of the current function, a value of
8988 @code{1} yields the frame address of the caller of the current function,
8991 The frame is the area on the stack that holds local variables and saved
8992 registers. The frame address is normally the address of the first word
8993 pushed on to the stack by the function. However, the exact definition
8994 depends upon the processor and the calling convention. If the processor
8995 has a dedicated frame pointer register, and the function has a frame,
8996 then @code{__builtin_frame_address} returns the value of the frame
8999 On some machines it may be impossible to determine the frame address of
9000 any function other than the current one; in such cases, or when the top
9001 of the stack has been reached, this function returns @code{0} if
9002 the first frame pointer is properly initialized by the startup code.
9004 Calling this function with a nonzero argument can have unpredictable
9005 effects, including crashing the calling program. As a result, calls
9006 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9007 option is in effect. Such calls should only be made in debugging
9011 @node Vector Extensions
9012 @section Using Vector Instructions through Built-in Functions
9014 On some targets, the instruction set contains SIMD vector instructions which
9015 operate on multiple values contained in one large register at the same time.
9016 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9019 The first step in using these extensions is to provide the necessary data
9020 types. This should be done using an appropriate @code{typedef}:
9023 typedef int v4si __attribute__ ((vector_size (16)));
9027 The @code{int} type specifies the base type, while the attribute specifies
9028 the vector size for the variable, measured in bytes. For example, the
9029 declaration above causes the compiler to set the mode for the @code{v4si}
9030 type to be 16 bytes wide and divided into @code{int} sized units. For
9031 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9032 corresponding mode of @code{foo} is @acronym{V4SI}.
9034 The @code{vector_size} attribute is only applicable to integral and
9035 float scalars, although arrays, pointers, and function return values
9036 are allowed in conjunction with this construct. Only sizes that are
9037 a power of two are currently allowed.
9039 All the basic integer types can be used as base types, both as signed
9040 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9041 @code{long long}. In addition, @code{float} and @code{double} can be
9042 used to build floating-point vector types.
9044 Specifying a combination that is not valid for the current architecture
9045 causes GCC to synthesize the instructions using a narrower mode.
9046 For example, if you specify a variable of type @code{V4SI} and your
9047 architecture does not allow for this specific SIMD type, GCC
9048 produces code that uses 4 @code{SIs}.
9050 The types defined in this manner can be used with a subset of normal C
9051 operations. Currently, GCC allows using the following operators
9052 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9054 The operations behave like C++ @code{valarrays}. Addition is defined as
9055 the addition of the corresponding elements of the operands. For
9056 example, in the code below, each of the 4 elements in @var{a} is
9057 added to the corresponding 4 elements in @var{b} and the resulting
9058 vector is stored in @var{c}.
9061 typedef int v4si __attribute__ ((vector_size (16)));
9068 Subtraction, multiplication, division, and the logical operations
9069 operate in a similar manner. Likewise, the result of using the unary
9070 minus or complement operators on a vector type is a vector whose
9071 elements are the negative or complemented values of the corresponding
9072 elements in the operand.
9074 It is possible to use shifting operators @code{<<}, @code{>>} on
9075 integer-type vectors. The operation is defined as following: @code{@{a0,
9076 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9077 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9080 For convenience, it is allowed to use a binary vector operation
9081 where one operand is a scalar. In that case the compiler transforms
9082 the scalar operand into a vector where each element is the scalar from
9083 the operation. The transformation happens only if the scalar could be
9084 safely converted to the vector-element type.
9085 Consider the following code.
9088 typedef int v4si __attribute__ ((vector_size (16)));
9093 a = b + 1; /* a = b + @{1,1,1,1@}; */
9094 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9096 a = l + a; /* Error, cannot convert long to int. */
9099 Vectors can be subscripted as if the vector were an array with
9100 the same number of elements and base type. Out of bound accesses
9101 invoke undefined behavior at run time. Warnings for out of bound
9102 accesses for vector subscription can be enabled with
9103 @option{-Warray-bounds}.
9105 Vector comparison is supported with standard comparison
9106 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9107 vector expressions of integer-type or real-type. Comparison between
9108 integer-type vectors and real-type vectors are not supported. The
9109 result of the comparison is a vector of the same width and number of
9110 elements as the comparison operands with a signed integral element
9113 Vectors are compared element-wise producing 0 when comparison is false
9114 and -1 (constant of the appropriate type where all bits are set)
9115 otherwise. Consider the following example.
9118 typedef int v4si __attribute__ ((vector_size (16)));
9120 v4si a = @{1,2,3,4@};
9121 v4si b = @{3,2,1,4@};
9124 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9125 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9128 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9129 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9130 integer vector with the same number of elements of the same size as @code{b}
9131 and @code{c}, computes all three arguments and creates a vector
9132 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9133 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9134 As in the case of binary operations, this syntax is also accepted when
9135 one of @code{b} or @code{c} is a scalar that is then transformed into a
9136 vector. If both @code{b} and @code{c} are scalars and the type of
9137 @code{true?b:c} has the same size as the element type of @code{a}, then
9138 @code{b} and @code{c} are converted to a vector type whose elements have
9139 this type and with the same number of elements as @code{a}.
9141 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9142 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9143 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9144 For mixed operations between a scalar @code{s} and a vector @code{v},
9145 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9146 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9148 Vector shuffling is available using functions
9149 @code{__builtin_shuffle (vec, mask)} and
9150 @code{__builtin_shuffle (vec0, vec1, mask)}.
9151 Both functions construct a permutation of elements from one or two
9152 vectors and return a vector of the same type as the input vector(s).
9153 The @var{mask} is an integral vector with the same width (@var{W})
9154 and element count (@var{N}) as the output vector.
9156 The elements of the input vectors are numbered in memory ordering of
9157 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9158 elements of @var{mask} are considered modulo @var{N} in the single-operand
9159 case and modulo @math{2*@var{N}} in the two-operand case.
9161 Consider the following example,
9164 typedef int v4si __attribute__ ((vector_size (16)));
9166 v4si a = @{1,2,3,4@};
9167 v4si b = @{5,6,7,8@};
9168 v4si mask1 = @{0,1,1,3@};
9169 v4si mask2 = @{0,4,2,5@};
9172 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9173 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9176 Note that @code{__builtin_shuffle} is intentionally semantically
9177 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9179 You can declare variables and use them in function calls and returns, as
9180 well as in assignments and some casts. You can specify a vector type as
9181 a return type for a function. Vector types can also be used as function
9182 arguments. It is possible to cast from one vector type to another,
9183 provided they are of the same size (in fact, you can also cast vectors
9184 to and from other datatypes of the same size).
9186 You cannot operate between vectors of different lengths or different
9187 signedness without a cast.
9190 @section Support for @code{offsetof}
9191 @findex __builtin_offsetof
9193 GCC implements for both C and C++ a syntactic extension to implement
9194 the @code{offsetof} macro.
9198 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9200 offsetof_member_designator:
9202 | offsetof_member_designator "." @code{identifier}
9203 | offsetof_member_designator "[" @code{expr} "]"
9206 This extension is sufficient such that
9209 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9213 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9214 may be dependent. In either case, @var{member} may consist of a single
9215 identifier, or a sequence of member accesses and array references.
9217 @node __sync Builtins
9218 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9220 The following built-in functions
9221 are intended to be compatible with those described
9222 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9223 section 7.4. As such, they depart from normal GCC practice by not using
9224 the @samp{__builtin_} prefix and also by being overloaded so that they
9225 work on multiple types.
9227 The definition given in the Intel documentation allows only for the use of
9228 the types @code{int}, @code{long}, @code{long long} or their unsigned
9229 counterparts. GCC allows any integral scalar or pointer type that is
9230 1, 2, 4 or 8 bytes in length.
9232 These functions are implemented in terms of the @samp{__atomic}
9233 builtins (@pxref{__atomic Builtins}). They should not be used for new
9234 code which should use the @samp{__atomic} builtins instead.
9236 Not all operations are supported by all target processors. If a particular
9237 operation cannot be implemented on the target processor, a warning is
9238 generated and a call to an external function is generated. The external
9239 function carries the same name as the built-in version,
9240 with an additional suffix
9241 @samp{_@var{n}} where @var{n} is the size of the data type.
9243 @c ??? Should we have a mechanism to suppress this warning? This is almost
9244 @c useful for implementing the operation under the control of an external
9247 In most cases, these built-in functions are considered a @dfn{full barrier}.
9249 no memory operand is moved across the operation, either forward or
9250 backward. Further, instructions are issued as necessary to prevent the
9251 processor from speculating loads across the operation and from queuing stores
9252 after the operation.
9254 All of the routines are described in the Intel documentation to take
9255 ``an optional list of variables protected by the memory barrier''. It's
9256 not clear what is meant by that; it could mean that @emph{only} the
9257 listed variables are protected, or it could mean a list of additional
9258 variables to be protected. The list is ignored by GCC which treats it as
9259 empty. GCC interprets an empty list as meaning that all globally
9260 accessible variables should be protected.
9263 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9264 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9265 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9266 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9267 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9268 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9269 @findex __sync_fetch_and_add
9270 @findex __sync_fetch_and_sub
9271 @findex __sync_fetch_and_or
9272 @findex __sync_fetch_and_and
9273 @findex __sync_fetch_and_xor
9274 @findex __sync_fetch_and_nand
9275 These built-in functions perform the operation suggested by the name, and
9276 returns the value that had previously been in memory. That is,
9279 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9280 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9283 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9284 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9286 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9287 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9288 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9289 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9290 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9291 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9292 @findex __sync_add_and_fetch
9293 @findex __sync_sub_and_fetch
9294 @findex __sync_or_and_fetch
9295 @findex __sync_and_and_fetch
9296 @findex __sync_xor_and_fetch
9297 @findex __sync_nand_and_fetch
9298 These built-in functions perform the operation suggested by the name, and
9299 return the new value. That is,
9302 @{ *ptr @var{op}= value; return *ptr; @}
9303 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9306 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9307 as @code{*ptr = ~(*ptr & value)} instead of
9308 @code{*ptr = ~*ptr & value}.
9310 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9311 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9312 @findex __sync_bool_compare_and_swap
9313 @findex __sync_val_compare_and_swap
9314 These built-in functions perform an atomic compare and swap.
9315 That is, if the current
9316 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9319 The ``bool'' version returns true if the comparison is successful and
9320 @var{newval} is written. The ``val'' version returns the contents
9321 of @code{*@var{ptr}} before the operation.
9323 @item __sync_synchronize (...)
9324 @findex __sync_synchronize
9325 This built-in function issues a full memory barrier.
9327 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9328 @findex __sync_lock_test_and_set
9329 This built-in function, as described by Intel, is not a traditional test-and-set
9330 operation, but rather an atomic exchange operation. It writes @var{value}
9331 into @code{*@var{ptr}}, and returns the previous contents of
9334 Many targets have only minimal support for such locks, and do not support
9335 a full exchange operation. In this case, a target may support reduced
9336 functionality here by which the @emph{only} valid value to store is the
9337 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9338 is implementation defined.
9340 This built-in function is not a full barrier,
9341 but rather an @dfn{acquire barrier}.
9342 This means that references after the operation cannot move to (or be
9343 speculated to) before the operation, but previous memory stores may not
9344 be globally visible yet, and previous memory loads may not yet be
9347 @item void __sync_lock_release (@var{type} *ptr, ...)
9348 @findex __sync_lock_release
9349 This built-in function releases the lock acquired by
9350 @code{__sync_lock_test_and_set}.
9351 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9353 This built-in function is not a full barrier,
9354 but rather a @dfn{release barrier}.
9355 This means that all previous memory stores are globally visible, and all
9356 previous memory loads have been satisfied, but following memory reads
9357 are not prevented from being speculated to before the barrier.
9360 @node __atomic Builtins
9361 @section Built-in Functions for Memory Model Aware Atomic Operations
9363 The following built-in functions approximately match the requirements
9364 for the C++11 memory model. They are all
9365 identified by being prefixed with @samp{__atomic} and most are
9366 overloaded so that they work with multiple types.
9368 These functions are intended to replace the legacy @samp{__sync}
9369 builtins. The main difference is that the memory order that is requested
9370 is a parameter to the functions. New code should always use the
9371 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9373 Note that the @samp{__atomic} builtins assume that programs will
9374 conform to the C++11 memory model. In particular, they assume
9375 that programs are free of data races. See the C++11 standard for
9376 detailed requirements.
9378 The @samp{__atomic} builtins can be used with any integral scalar or
9379 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9380 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9381 supported by the architecture.
9383 The four non-arithmetic functions (load, store, exchange, and
9384 compare_exchange) all have a generic version as well. This generic
9385 version works on any data type. It uses the lock-free built-in function
9386 if the specific data type size makes that possible; otherwise, an
9387 external call is left to be resolved at run time. This external call is
9388 the same format with the addition of a @samp{size_t} parameter inserted
9389 as the first parameter indicating the size of the object being pointed to.
9390 All objects must be the same size.
9392 There are 6 different memory orders that can be specified. These map
9393 to the C++11 memory orders with the same names, see the C++11 standard
9394 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9395 on atomic synchronization} for detailed definitions. Individual
9396 targets may also support additional memory orders for use on specific
9397 architectures. Refer to the target documentation for details of
9400 An atomic operation can both constrain code motion and
9401 be mapped to hardware instructions for synchronization between threads
9402 (e.g., a fence). To which extent this happens is controlled by the
9403 memory orders, which are listed here in approximately ascending order of
9404 strength. The description of each memory order is only meant to roughly
9405 illustrate the effects and is not a specification; see the C++11
9406 memory model for precise semantics.
9409 @item __ATOMIC_RELAXED
9410 Implies no inter-thread ordering constraints.
9411 @item __ATOMIC_CONSUME
9412 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9413 memory order because of a deficiency in C++11's semantics for
9414 @code{memory_order_consume}.
9415 @item __ATOMIC_ACQUIRE
9416 Creates an inter-thread happens-before constraint from the release (or
9417 stronger) semantic store to this acquire load. Can prevent hoisting
9418 of code to before the operation.
9419 @item __ATOMIC_RELEASE
9420 Creates an inter-thread happens-before constraint to acquire (or stronger)
9421 semantic loads that read from this release store. Can prevent sinking
9422 of code to after the operation.
9423 @item __ATOMIC_ACQ_REL
9424 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9425 @code{__ATOMIC_RELEASE}.
9426 @item __ATOMIC_SEQ_CST
9427 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9430 Note that in the C++11 memory model, @emph{fences} (e.g.,
9431 @samp{__atomic_thread_fence}) take effect in combination with other
9432 atomic operations on specific memory locations (e.g., atomic loads);
9433 operations on specific memory locations do not necessarily affect other
9434 operations in the same way.
9436 Target architectures are encouraged to provide their own patterns for
9437 each of the atomic built-in functions. If no target is provided, the original
9438 non-memory model set of @samp{__sync} atomic built-in functions are
9439 used, along with any required synchronization fences surrounding it in
9440 order to achieve the proper behavior. Execution in this case is subject
9441 to the same restrictions as those built-in functions.
9443 If there is no pattern or mechanism to provide a lock-free instruction
9444 sequence, a call is made to an external routine with the same parameters
9445 to be resolved at run time.
9447 When implementing patterns for these built-in functions, the memory order
9448 parameter can be ignored as long as the pattern implements the most
9449 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9450 orders execute correctly with this memory order but they may not execute as
9451 efficiently as they could with a more appropriate implementation of the
9452 relaxed requirements.
9454 Note that the C++11 standard allows for the memory order parameter to be
9455 determined at run time rather than at compile time. These built-in
9456 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9457 than invoke a runtime library call or inline a switch statement. This is
9458 standard compliant, safe, and the simplest approach for now.
9460 The memory order parameter is a signed int, but only the lower 16 bits are
9461 reserved for the memory order. The remainder of the signed int is reserved
9462 for target use and should be 0. Use of the predefined atomic values
9463 ensures proper usage.
9465 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9466 This built-in function implements an atomic load operation. It returns the
9467 contents of @code{*@var{ptr}}.
9469 The valid memory order variants are
9470 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9471 and @code{__ATOMIC_CONSUME}.
9475 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9476 This is the generic version of an atomic load. It returns the
9477 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9481 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9482 This built-in function implements an atomic store operation. It writes
9483 @code{@var{val}} into @code{*@var{ptr}}.
9485 The valid memory order variants are
9486 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9490 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9491 This is the generic version of an atomic store. It stores the value
9492 of @code{*@var{val}} into @code{*@var{ptr}}.
9496 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9497 This built-in function implements an atomic exchange operation. It writes
9498 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9501 The valid memory order variants are
9502 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9503 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9507 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9508 This is the generic version of an atomic exchange. It stores the
9509 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9510 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9514 @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)
9515 This built-in function implements an atomic compare and exchange operation.
9516 This compares the contents of @code{*@var{ptr}} with the contents of
9517 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9518 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9519 equal, the operation is a @emph{read} and the current contents of
9520 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
9521 for weak compare_exchange, and false for the strong variation. Many targets
9522 only offer the strong variation and ignore the parameter. When in doubt, use
9523 the strong variation.
9525 True is returned if @var{desired} is written into
9526 @code{*@var{ptr}} and the operation is considered to conform to the
9527 memory order specified by @var{success_memorder}. There are no
9528 restrictions on what memory order can be used here.
9530 False is returned otherwise, and the operation is considered to conform
9531 to @var{failure_memorder}. This memory order cannot be
9532 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9533 stronger order than that specified by @var{success_memorder}.
9537 @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)
9538 This built-in function implements the generic version of
9539 @code{__atomic_compare_exchange}. The function is virtually identical to
9540 @code{__atomic_compare_exchange_n}, except the desired value is also a
9545 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9546 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9547 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9548 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9549 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9550 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9551 These built-in functions perform the operation suggested by the name, and
9552 return the result of the operation. That is,
9555 @{ *ptr @var{op}= val; return *ptr; @}
9558 All memory orders are valid.
9562 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9563 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9564 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9565 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9566 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9567 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9568 These built-in functions perform the operation suggested by the name, and
9569 return the value that had previously been in @code{*@var{ptr}}. That is,
9572 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9575 All memory orders are valid.
9579 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9581 This built-in function performs an atomic test-and-set operation on
9582 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9583 defined nonzero ``set'' value and the return value is @code{true} if and only
9584 if the previous contents were ``set''.
9585 It should be only used for operands of type @code{bool} or @code{char}. For
9586 other types only part of the value may be set.
9588 All memory orders are valid.
9592 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9594 This built-in function performs an atomic clear operation on
9595 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9596 It should be only used for operands of type @code{bool} or @code{char} and
9597 in conjunction with @code{__atomic_test_and_set}.
9598 For other types it may only clear partially. If the type is not @code{bool}
9599 prefer using @code{__atomic_store}.
9601 The valid memory order variants are
9602 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9603 @code{__ATOMIC_RELEASE}.
9607 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9609 This built-in function acts as a synchronization fence between threads
9610 based on the specified memory order.
9612 All memory orders are valid.
9616 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9618 This built-in function acts as a synchronization fence between a thread
9619 and signal handlers based in the same thread.
9621 All memory orders are valid.
9625 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9627 This built-in function returns true if objects of @var{size} bytes always
9628 generate lock-free atomic instructions for the target architecture.
9629 @var{size} must resolve to a compile-time constant and the result also
9630 resolves to a compile-time constant.
9632 @var{ptr} is an optional pointer to the object that may be used to determine
9633 alignment. A value of 0 indicates typical alignment should be used. The
9634 compiler may also ignore this parameter.
9637 if (_atomic_always_lock_free (sizeof (long long), 0))
9642 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9644 This built-in function returns true if objects of @var{size} bytes always
9645 generate lock-free atomic instructions for the target architecture. If
9646 the built-in function is not known to be lock-free, a call is made to a
9647 runtime routine named @code{__atomic_is_lock_free}.
9649 @var{ptr} is an optional pointer to the object that may be used to determine
9650 alignment. A value of 0 indicates typical alignment should be used. The
9651 compiler may also ignore this parameter.
9654 @node Integer Overflow Builtins
9655 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9657 The following built-in functions allow performing simple arithmetic operations
9658 together with checking whether the operations overflowed.
9660 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9661 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9662 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9663 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9664 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9665 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9666 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9668 These built-in functions promote the first two operands into infinite precision signed
9669 type and perform addition on those promoted operands. The result is then
9670 cast to the type the third pointer argument points to and stored there.
9671 If the stored result is equal to the infinite precision result, the built-in
9672 functions return false, otherwise they return true. As the addition is
9673 performed in infinite signed precision, these built-in functions have fully defined
9674 behavior for all argument values.
9676 The first built-in function allows arbitrary integral types for operands and
9677 the result type must be pointer to some integer type, the rest of the built-in
9678 functions have explicit integer types.
9680 The compiler will attempt to use hardware instructions to implement
9681 these built-in functions where possible, like conditional jump on overflow
9682 after addition, conditional jump on carry etc.
9686 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9687 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9688 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9689 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9690 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9691 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9692 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9694 These built-in functions are similar to the add overflow checking built-in
9695 functions above, except they perform subtraction, subtract the second argument
9696 from the first one, instead of addition.
9700 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9701 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9702 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9703 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9704 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9705 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9706 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9708 These built-in functions are similar to the add overflow checking built-in
9709 functions above, except they perform multiplication, instead of addition.
9713 @node x86 specific memory model extensions for transactional memory
9714 @section x86-Specific Memory Model Extensions for Transactional Memory
9716 The x86 architecture supports additional memory ordering flags
9717 to mark lock critical sections for hardware lock elision.
9718 These must be specified in addition to an existing memory order to
9722 @item __ATOMIC_HLE_ACQUIRE
9723 Start lock elision on a lock variable.
9724 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9725 @item __ATOMIC_HLE_RELEASE
9726 End lock elision on a lock variable.
9727 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9730 When a lock acquire fails, it is required for good performance to abort
9731 the transaction quickly. This can be done with a @code{_mm_pause}.
9734 #include <immintrin.h> // For _mm_pause
9738 /* Acquire lock with lock elision */
9739 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9740 _mm_pause(); /* Abort failed transaction */
9742 /* Free lock with lock elision */
9743 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9746 @node Object Size Checking
9747 @section Object Size Checking Built-in Functions
9748 @findex __builtin_object_size
9749 @findex __builtin___memcpy_chk
9750 @findex __builtin___mempcpy_chk
9751 @findex __builtin___memmove_chk
9752 @findex __builtin___memset_chk
9753 @findex __builtin___strcpy_chk
9754 @findex __builtin___stpcpy_chk
9755 @findex __builtin___strncpy_chk
9756 @findex __builtin___strcat_chk
9757 @findex __builtin___strncat_chk
9758 @findex __builtin___sprintf_chk
9759 @findex __builtin___snprintf_chk
9760 @findex __builtin___vsprintf_chk
9761 @findex __builtin___vsnprintf_chk
9762 @findex __builtin___printf_chk
9763 @findex __builtin___vprintf_chk
9764 @findex __builtin___fprintf_chk
9765 @findex __builtin___vfprintf_chk
9767 GCC implements a limited buffer overflow protection mechanism
9768 that can prevent some buffer overflow attacks.
9770 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9771 is a built-in construct that returns a constant number of bytes from
9772 @var{ptr} to the end of the object @var{ptr} pointer points to
9773 (if known at compile time). @code{__builtin_object_size} never evaluates
9774 its arguments for side-effects. If there are any side-effects in them, it
9775 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9776 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9777 point to and all of them are known at compile time, the returned number
9778 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9779 0 and minimum if nonzero. If it is not possible to determine which objects
9780 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9781 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9782 for @var{type} 2 or 3.
9784 @var{type} is an integer constant from 0 to 3. If the least significant
9785 bit is clear, objects are whole variables, if it is set, a closest
9786 surrounding subobject is considered the object a pointer points to.
9787 The second bit determines if maximum or minimum of remaining bytes
9791 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9792 char *p = &var.buf1[1], *q = &var.b;
9794 /* Here the object p points to is var. */
9795 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9796 /* The subobject p points to is var.buf1. */
9797 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9798 /* The object q points to is var. */
9799 assert (__builtin_object_size (q, 0)
9800 == (char *) (&var + 1) - (char *) &var.b);
9801 /* The subobject q points to is var.b. */
9802 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9806 There are built-in functions added for many common string operation
9807 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9808 built-in is provided. This built-in has an additional last argument,
9809 which is the number of bytes remaining in object the @var{dest}
9810 argument points to or @code{(size_t) -1} if the size is not known.
9812 The built-in functions are optimized into the normal string functions
9813 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9814 it is known at compile time that the destination object will not
9815 be overflown. If the compiler can determine at compile time the
9816 object will be always overflown, it issues a warning.
9818 The intended use can be e.g.@:
9822 #define bos0(dest) __builtin_object_size (dest, 0)
9823 #define memcpy(dest, src, n) \
9824 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9828 /* It is unknown what object p points to, so this is optimized
9829 into plain memcpy - no checking is possible. */
9830 memcpy (p, "abcde", n);
9831 /* Destination is known and length too. It is known at compile
9832 time there will be no overflow. */
9833 memcpy (&buf[5], "abcde", 5);
9834 /* Destination is known, but the length is not known at compile time.
9835 This will result in __memcpy_chk call that can check for overflow
9837 memcpy (&buf[5], "abcde", n);
9838 /* Destination is known and it is known at compile time there will
9839 be overflow. There will be a warning and __memcpy_chk call that
9840 will abort the program at run time. */
9841 memcpy (&buf[6], "abcde", 5);
9844 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9845 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9846 @code{strcat} and @code{strncat}.
9848 There are also checking built-in functions for formatted output functions.
9850 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9851 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9852 const char *fmt, ...);
9853 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9855 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9856 const char *fmt, va_list ap);
9859 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9860 etc.@: functions and can contain implementation specific flags on what
9861 additional security measures the checking function might take, such as
9862 handling @code{%n} differently.
9864 The @var{os} argument is the object size @var{s} points to, like in the
9865 other built-in functions. There is a small difference in the behavior
9866 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9867 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9868 the checking function is called with @var{os} argument set to
9871 In addition to this, there are checking built-in functions
9872 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9873 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9874 These have just one additional argument, @var{flag}, right before
9875 format string @var{fmt}. If the compiler is able to optimize them to
9876 @code{fputc} etc.@: functions, it does, otherwise the checking function
9877 is called and the @var{flag} argument passed to it.
9879 @node Pointer Bounds Checker builtins
9880 @section Pointer Bounds Checker Built-in Functions
9881 @cindex Pointer Bounds Checker builtins
9882 @findex __builtin___bnd_set_ptr_bounds
9883 @findex __builtin___bnd_narrow_ptr_bounds
9884 @findex __builtin___bnd_copy_ptr_bounds
9885 @findex __builtin___bnd_init_ptr_bounds
9886 @findex __builtin___bnd_null_ptr_bounds
9887 @findex __builtin___bnd_store_ptr_bounds
9888 @findex __builtin___bnd_chk_ptr_lbounds
9889 @findex __builtin___bnd_chk_ptr_ubounds
9890 @findex __builtin___bnd_chk_ptr_bounds
9891 @findex __builtin___bnd_get_ptr_lbound
9892 @findex __builtin___bnd_get_ptr_ubound
9894 GCC provides a set of built-in functions to control Pointer Bounds Checker
9895 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9896 even if you compile with Pointer Bounds Checker off
9897 (@option{-fno-check-pointer-bounds}).
9898 The behavior may differ in such case as documented below.
9900 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9902 This built-in function returns a new pointer with the value of @var{q}, and
9903 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9904 Bounds Checker off, the built-in function just returns the first argument.
9907 extern void *__wrap_malloc (size_t n)
9909 void *p = (void *)__real_malloc (n);
9910 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9911 return __builtin___bnd_set_ptr_bounds (p, n);
9917 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9919 This built-in function returns a new pointer with the value of @var{p}
9920 and associates it with the narrowed bounds formed by the intersection
9921 of bounds associated with @var{q} and the bounds
9922 [@var{p}, @var{p} + @var{size} - 1].
9923 With Pointer Bounds Checker off, the built-in function just returns the first
9927 void init_objects (object *objs, size_t size)
9930 /* Initialize objects one-by-one passing pointers with bounds of
9931 an object, not the full array of objects. */
9932 for (i = 0; i < size; i++)
9933 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
9940 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
9942 This built-in function returns a new pointer with the value of @var{q},
9943 and associates it with the bounds already associated with pointer @var{r}.
9944 With Pointer Bounds Checker off, the built-in function just returns the first
9948 /* Here is a way to get pointer to object's field but
9949 still with the full object's bounds. */
9950 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
9956 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
9958 This built-in function returns a new pointer with the value of @var{q}, and
9959 associates it with INIT (allowing full memory access) bounds. With Pointer
9960 Bounds Checker off, the built-in function just returns the first argument.
9964 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
9966 This built-in function returns a new pointer with the value of @var{q}, and
9967 associates it with NULL (allowing no memory access) bounds. With Pointer
9968 Bounds Checker off, the built-in function just returns the first argument.
9972 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
9974 This built-in function stores the bounds associated with pointer @var{ptr_val}
9975 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
9976 bounds from legacy code without touching the associated pointer's memory when
9977 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
9978 function call is ignored.
9982 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
9984 This built-in function checks if the pointer @var{q} is within the lower
9985 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
9986 function call is ignored.
9989 extern void *__wrap_memset (void *dst, int c, size_t len)
9993 __builtin___bnd_chk_ptr_lbounds (dst);
9994 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
9995 __real_memset (dst, c, len);
10003 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10005 This built-in function checks if the pointer @var{q} is within the upper
10006 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10007 function call is ignored.
10011 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10013 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10014 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10015 off, the built-in function call is ignored.
10018 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10022 __bnd_chk_ptr_bounds (dst, n);
10023 __bnd_chk_ptr_bounds (src, n);
10024 __real_memcpy (dst, src, n);
10032 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10034 This built-in function returns the lower bound associated
10035 with the pointer @var{q}, as a pointer value.
10036 This is useful for debugging using @code{printf}.
10037 With Pointer Bounds Checker off, the built-in function returns 0.
10040 void *lb = __builtin___bnd_get_ptr_lbound (q);
10041 void *ub = __builtin___bnd_get_ptr_ubound (q);
10042 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10047 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10049 This built-in function returns the upper bound (which is a pointer) associated
10050 with the pointer @var{q}. With Pointer Bounds Checker off,
10051 the built-in function returns -1.
10055 @node Cilk Plus Builtins
10056 @section Cilk Plus C/C++ Language Extension Built-in Functions
10058 GCC provides support for the following built-in reduction functions if Cilk Plus
10059 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10062 @item @code{__sec_implicit_index}
10063 @item @code{__sec_reduce}
10064 @item @code{__sec_reduce_add}
10065 @item @code{__sec_reduce_all_nonzero}
10066 @item @code{__sec_reduce_all_zero}
10067 @item @code{__sec_reduce_any_nonzero}
10068 @item @code{__sec_reduce_any_zero}
10069 @item @code{__sec_reduce_max}
10070 @item @code{__sec_reduce_min}
10071 @item @code{__sec_reduce_max_ind}
10072 @item @code{__sec_reduce_min_ind}
10073 @item @code{__sec_reduce_mul}
10074 @item @code{__sec_reduce_mutating}
10077 Further details and examples about these built-in functions are described
10078 in the Cilk Plus language manual which can be found at
10079 @uref{http://www.cilkplus.org}.
10081 @node Other Builtins
10082 @section Other Built-in Functions Provided by GCC
10083 @cindex built-in functions
10084 @findex __builtin_call_with_static_chain
10085 @findex __builtin_fpclassify
10086 @findex __builtin_isfinite
10087 @findex __builtin_isnormal
10088 @findex __builtin_isgreater
10089 @findex __builtin_isgreaterequal
10090 @findex __builtin_isinf_sign
10091 @findex __builtin_isless
10092 @findex __builtin_islessequal
10093 @findex __builtin_islessgreater
10094 @findex __builtin_isunordered
10095 @findex __builtin_powi
10096 @findex __builtin_powif
10097 @findex __builtin_powil
10255 @findex fprintf_unlocked
10257 @findex fputs_unlocked
10365 @findex nexttowardf
10366 @findex nexttowardl
10374 @findex printf_unlocked
10404 @findex signbitd128
10405 @findex significand
10406 @findex significandf
10407 @findex significandl
10435 @findex strncasecmp
10478 GCC provides a large number of built-in functions other than the ones
10479 mentioned above. Some of these are for internal use in the processing
10480 of exceptions or variable-length argument lists and are not
10481 documented here because they may change from time to time; we do not
10482 recommend general use of these functions.
10484 The remaining functions are provided for optimization purposes.
10486 With the exception of built-ins that have library equivalents such as
10487 the standard C library functions discussed below, or that expand to
10488 library calls, GCC built-in functions are always expanded inline and
10489 thus do not have corresponding entry points and their address cannot
10490 be obtained. Attempting to use them in an expression other than
10491 a function call results in a compile-time error.
10493 @opindex fno-builtin
10494 GCC includes built-in versions of many of the functions in the standard
10495 C library. These functions come in two forms: one whose names start with
10496 the @code{__builtin_} prefix, and the other without. Both forms have the
10497 same type (including prototype), the same address (when their address is
10498 taken), and the same meaning as the C library functions even if you specify
10499 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10500 functions are only optimized in certain cases; if they are not optimized in
10501 a particular case, a call to the library function is emitted.
10505 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10506 @option{-std=c99} or @option{-std=c11}), the functions
10507 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10508 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10509 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10510 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10511 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10512 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10513 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10514 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10515 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10516 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10517 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10518 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10519 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10520 @code{significandl}, @code{significand}, @code{sincosf},
10521 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10522 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10523 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10524 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10526 may be handled as built-in functions.
10527 All these functions have corresponding versions
10528 prefixed with @code{__builtin_}, which may be used even in strict C90
10531 The ISO C99 functions
10532 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10533 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10534 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10535 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10536 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10537 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10538 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10539 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10540 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10541 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10542 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10543 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10544 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10545 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10546 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10547 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10548 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10549 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10550 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10551 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10552 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10553 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10554 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10555 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10556 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10557 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10558 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10559 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10560 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10561 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10562 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10563 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10564 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10565 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10566 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10567 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10568 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10569 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10570 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10571 are handled as built-in functions
10572 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10574 There are also built-in versions of the ISO C99 functions
10575 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10576 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10577 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10578 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10579 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10580 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10581 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10582 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10583 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10584 that are recognized in any mode since ISO C90 reserves these names for
10585 the purpose to which ISO C99 puts them. All these functions have
10586 corresponding versions prefixed with @code{__builtin_}.
10588 The ISO C94 functions
10589 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10590 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10591 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10593 are handled as built-in functions
10594 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10596 The ISO C90 functions
10597 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10598 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10599 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10600 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10601 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10602 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10603 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10604 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10605 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10606 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10607 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10608 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10609 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10610 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10611 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10612 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10613 are all recognized as built-in functions unless
10614 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10615 is specified for an individual function). All of these functions have
10616 corresponding versions prefixed with @code{__builtin_}.
10618 GCC provides built-in versions of the ISO C99 floating-point comparison
10619 macros that avoid raising exceptions for unordered operands. They have
10620 the same names as the standard macros ( @code{isgreater},
10621 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10622 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10623 prefixed. We intend for a library implementor to be able to simply
10624 @code{#define} each standard macro to its built-in equivalent.
10625 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10626 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10627 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10628 built-in functions appear both with and without the @code{__builtin_} prefix.
10630 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10632 You can use the built-in function @code{__builtin_types_compatible_p} to
10633 determine whether two types are the same.
10635 This built-in function returns 1 if the unqualified versions of the
10636 types @var{type1} and @var{type2} (which are types, not expressions) are
10637 compatible, 0 otherwise. The result of this built-in function can be
10638 used in integer constant expressions.
10640 This built-in function ignores top level qualifiers (e.g., @code{const},
10641 @code{volatile}). For example, @code{int} is equivalent to @code{const
10644 The type @code{int[]} and @code{int[5]} are compatible. On the other
10645 hand, @code{int} and @code{char *} are not compatible, even if the size
10646 of their types, on the particular architecture are the same. Also, the
10647 amount of pointer indirection is taken into account when determining
10648 similarity. Consequently, @code{short *} is not similar to
10649 @code{short **}. Furthermore, two types that are typedefed are
10650 considered compatible if their underlying types are compatible.
10652 An @code{enum} type is not considered to be compatible with another
10653 @code{enum} type even if both are compatible with the same integer
10654 type; this is what the C standard specifies.
10655 For example, @code{enum @{foo, bar@}} is not similar to
10656 @code{enum @{hot, dog@}}.
10658 You typically use this function in code whose execution varies
10659 depending on the arguments' types. For example:
10664 typeof (x) tmp = (x); \
10665 if (__builtin_types_compatible_p (typeof (x), long double)) \
10666 tmp = foo_long_double (tmp); \
10667 else if (__builtin_types_compatible_p (typeof (x), double)) \
10668 tmp = foo_double (tmp); \
10669 else if (__builtin_types_compatible_p (typeof (x), float)) \
10670 tmp = foo_float (tmp); \
10677 @emph{Note:} This construct is only available for C@.
10681 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10683 The @var{call_exp} expression must be a function call, and the
10684 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10685 is passed to the function call in the target's static chain location.
10686 The result of builtin is the result of the function call.
10688 @emph{Note:} This builtin is only available for C@.
10689 This builtin can be used to call Go closures from C.
10693 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10695 You can use the built-in function @code{__builtin_choose_expr} to
10696 evaluate code depending on the value of a constant expression. This
10697 built-in function returns @var{exp1} if @var{const_exp}, which is an
10698 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10700 This built-in function is analogous to the @samp{? :} operator in C,
10701 except that the expression returned has its type unaltered by promotion
10702 rules. Also, the built-in function does not evaluate the expression
10703 that is not chosen. For example, if @var{const_exp} evaluates to true,
10704 @var{exp2} is not evaluated even if it has side-effects.
10706 This built-in function can return an lvalue if the chosen argument is an
10709 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10710 type. Similarly, if @var{exp2} is returned, its return type is the same
10717 __builtin_choose_expr ( \
10718 __builtin_types_compatible_p (typeof (x), double), \
10720 __builtin_choose_expr ( \
10721 __builtin_types_compatible_p (typeof (x), float), \
10723 /* @r{The void expression results in a compile-time error} \
10724 @r{when assigning the result to something.} */ \
10728 @emph{Note:} This construct is only available for C@. Furthermore, the
10729 unused expression (@var{exp1} or @var{exp2} depending on the value of
10730 @var{const_exp}) may still generate syntax errors. This may change in
10735 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10737 The built-in function @code{__builtin_complex} is provided for use in
10738 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10739 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10740 real binary floating-point type, and the result has the corresponding
10741 complex type with real and imaginary parts @var{real} and @var{imag}.
10742 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10743 infinities, NaNs and negative zeros are involved.
10747 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10748 You can use the built-in function @code{__builtin_constant_p} to
10749 determine if a value is known to be constant at compile time and hence
10750 that GCC can perform constant-folding on expressions involving that
10751 value. The argument of the function is the value to test. The function
10752 returns the integer 1 if the argument is known to be a compile-time
10753 constant and 0 if it is not known to be a compile-time constant. A
10754 return of 0 does not indicate that the value is @emph{not} a constant,
10755 but merely that GCC cannot prove it is a constant with the specified
10756 value of the @option{-O} option.
10758 You typically use this function in an embedded application where
10759 memory is a critical resource. If you have some complex calculation,
10760 you may want it to be folded if it involves constants, but need to call
10761 a function if it does not. For example:
10764 #define Scale_Value(X) \
10765 (__builtin_constant_p (X) \
10766 ? ((X) * SCALE + OFFSET) : Scale (X))
10769 You may use this built-in function in either a macro or an inline
10770 function. However, if you use it in an inlined function and pass an
10771 argument of the function as the argument to the built-in, GCC
10772 never returns 1 when you call the inline function with a string constant
10773 or compound literal (@pxref{Compound Literals}) and does not return 1
10774 when you pass a constant numeric value to the inline function unless you
10775 specify the @option{-O} option.
10777 You may also use @code{__builtin_constant_p} in initializers for static
10778 data. For instance, you can write
10781 static const int table[] = @{
10782 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10788 This is an acceptable initializer even if @var{EXPRESSION} is not a
10789 constant expression, including the case where
10790 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10791 folded to a constant but @var{EXPRESSION} contains operands that are
10792 not otherwise permitted in a static initializer (for example,
10793 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10794 built-in in this case, because it has no opportunity to perform
10798 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10799 @opindex fprofile-arcs
10800 You may use @code{__builtin_expect} to provide the compiler with
10801 branch prediction information. In general, you should prefer to
10802 use actual profile feedback for this (@option{-fprofile-arcs}), as
10803 programmers are notoriously bad at predicting how their programs
10804 actually perform. However, there are applications in which this
10805 data is hard to collect.
10807 The return value is the value of @var{exp}, which should be an integral
10808 expression. The semantics of the built-in are that it is expected that
10809 @var{exp} == @var{c}. For example:
10812 if (__builtin_expect (x, 0))
10817 indicates that we do not expect to call @code{foo}, since
10818 we expect @code{x} to be zero. Since you are limited to integral
10819 expressions for @var{exp}, you should use constructions such as
10822 if (__builtin_expect (ptr != NULL, 1))
10827 when testing pointer or floating-point values.
10830 @deftypefn {Built-in Function} void __builtin_trap (void)
10831 This function causes the program to exit abnormally. GCC implements
10832 this function by using a target-dependent mechanism (such as
10833 intentionally executing an illegal instruction) or by calling
10834 @code{abort}. The mechanism used may vary from release to release so
10835 you should not rely on any particular implementation.
10838 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10839 If control flow reaches the point of the @code{__builtin_unreachable},
10840 the program is undefined. It is useful in situations where the
10841 compiler cannot deduce the unreachability of the code.
10843 One such case is immediately following an @code{asm} statement that
10844 either never terminates, or one that transfers control elsewhere
10845 and never returns. In this example, without the
10846 @code{__builtin_unreachable}, GCC issues a warning that control
10847 reaches the end of a non-void function. It also generates code
10848 to return after the @code{asm}.
10851 int f (int c, int v)
10859 asm("jmp error_handler");
10860 __builtin_unreachable ();
10866 Because the @code{asm} statement unconditionally transfers control out
10867 of the function, control never reaches the end of the function
10868 body. The @code{__builtin_unreachable} is in fact unreachable and
10869 communicates this fact to the compiler.
10871 Another use for @code{__builtin_unreachable} is following a call a
10872 function that never returns but that is not declared
10873 @code{__attribute__((noreturn))}, as in this example:
10876 void function_that_never_returns (void);
10886 function_that_never_returns ();
10887 __builtin_unreachable ();
10894 @deftypefn {Built-in Function} void *__builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
10895 This function returns its first argument, and allows the compiler
10896 to assume that the returned pointer is at least @var{align} bytes
10897 aligned. This built-in can have either two or three arguments,
10898 if it has three, the third argument should have integer type, and
10899 if it is nonzero means misalignment offset. For example:
10902 void *x = __builtin_assume_aligned (arg, 16);
10906 means that the compiler can assume @code{x}, set to @code{arg}, is at least
10907 16-byte aligned, while:
10910 void *x = __builtin_assume_aligned (arg, 32, 8);
10914 means that the compiler can assume for @code{x}, set to @code{arg}, that
10915 @code{(char *) x - 8} is 32-byte aligned.
10918 @deftypefn {Built-in Function} int __builtin_LINE ()
10919 This function is the equivalent to the preprocessor @code{__LINE__}
10920 macro and returns the line number of the invocation of the built-in.
10921 In a C++ default argument for a function @var{F}, it gets the line number of
10922 the call to @var{F}.
10925 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
10926 This function is the equivalent to the preprocessor @code{__FUNCTION__}
10927 macro and returns the function name the invocation of the built-in is in.
10930 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
10931 This function is the equivalent to the preprocessor @code{__FILE__}
10932 macro and returns the file name the invocation of the built-in is in.
10933 In a C++ default argument for a function @var{F}, it gets the file name of
10934 the call to @var{F}.
10937 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
10938 This function is used to flush the processor's instruction cache for
10939 the region of memory between @var{begin} inclusive and @var{end}
10940 exclusive. Some targets require that the instruction cache be
10941 flushed, after modifying memory containing code, in order to obtain
10942 deterministic behavior.
10944 If the target does not require instruction cache flushes,
10945 @code{__builtin___clear_cache} has no effect. Otherwise either
10946 instructions are emitted in-line to clear the instruction cache or a
10947 call to the @code{__clear_cache} function in libgcc is made.
10950 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
10951 This function is used to minimize cache-miss latency by moving data into
10952 a cache before it is accessed.
10953 You can insert calls to @code{__builtin_prefetch} into code for which
10954 you know addresses of data in memory that is likely to be accessed soon.
10955 If the target supports them, data prefetch instructions are generated.
10956 If the prefetch is done early enough before the access then the data will
10957 be in the cache by the time it is accessed.
10959 The value of @var{addr} is the address of the memory to prefetch.
10960 There are two optional arguments, @var{rw} and @var{locality}.
10961 The value of @var{rw} is a compile-time constant one or zero; one
10962 means that the prefetch is preparing for a write to the memory address
10963 and zero, the default, means that the prefetch is preparing for a read.
10964 The value @var{locality} must be a compile-time constant integer between
10965 zero and three. A value of zero means that the data has no temporal
10966 locality, so it need not be left in the cache after the access. A value
10967 of three means that the data has a high degree of temporal locality and
10968 should be left in all levels of cache possible. Values of one and two
10969 mean, respectively, a low or moderate degree of temporal locality. The
10973 for (i = 0; i < n; i++)
10975 a[i] = a[i] + b[i];
10976 __builtin_prefetch (&a[i+j], 1, 1);
10977 __builtin_prefetch (&b[i+j], 0, 1);
10982 Data prefetch does not generate faults if @var{addr} is invalid, but
10983 the address expression itself must be valid. For example, a prefetch
10984 of @code{p->next} does not fault if @code{p->next} is not a valid
10985 address, but evaluation faults if @code{p} is not a valid address.
10987 If the target does not support data prefetch, the address expression
10988 is evaluated if it includes side effects but no other code is generated
10989 and GCC does not issue a warning.
10992 @deftypefn {Built-in Function} double __builtin_huge_val (void)
10993 Returns a positive infinity, if supported by the floating-point format,
10994 else @code{DBL_MAX}. This function is suitable for implementing the
10995 ISO C macro @code{HUGE_VAL}.
10998 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
10999 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11002 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11003 Similar to @code{__builtin_huge_val}, except the return
11004 type is @code{long double}.
11007 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11008 This built-in implements the C99 fpclassify functionality. The first
11009 five int arguments should be the target library's notion of the
11010 possible FP classes and are used for return values. They must be
11011 constant values and they must appear in this order: @code{FP_NAN},
11012 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11013 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11014 to classify. GCC treats the last argument as type-generic, which
11015 means it does not do default promotion from float to double.
11018 @deftypefn {Built-in Function} double __builtin_inf (void)
11019 Similar to @code{__builtin_huge_val}, except a warning is generated
11020 if the target floating-point format does not support infinities.
11023 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11024 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11027 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11028 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11031 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11032 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11035 @deftypefn {Built-in Function} float __builtin_inff (void)
11036 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11037 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11040 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11041 Similar to @code{__builtin_inf}, except the return
11042 type is @code{long double}.
11045 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11046 Similar to @code{isinf}, except the return value is -1 for
11047 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11048 Note while the parameter list is an
11049 ellipsis, this function only accepts exactly one floating-point
11050 argument. GCC treats this parameter as type-generic, which means it
11051 does not do default promotion from float to double.
11054 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11055 This is an implementation of the ISO C99 function @code{nan}.
11057 Since ISO C99 defines this function in terms of @code{strtod}, which we
11058 do not implement, a description of the parsing is in order. The string
11059 is parsed as by @code{strtol}; that is, the base is recognized by
11060 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11061 in the significand such that the least significant bit of the number
11062 is at the least significant bit of the significand. The number is
11063 truncated to fit the significand field provided. The significand is
11064 forced to be a quiet NaN@.
11066 This function, if given a string literal all of which would have been
11067 consumed by @code{strtol}, is evaluated early enough that it is considered a
11068 compile-time constant.
11071 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11072 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11075 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11076 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11079 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11080 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11083 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11084 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11087 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11088 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11091 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11092 Similar to @code{__builtin_nan}, except the significand is forced
11093 to be a signaling NaN@. The @code{nans} function is proposed by
11094 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11097 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11098 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11101 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11102 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11105 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11106 Returns one plus the index of the least significant 1-bit of @var{x}, or
11107 if @var{x} is zero, returns zero.
11110 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11111 Returns the number of leading 0-bits in @var{x}, starting at the most
11112 significant bit position. If @var{x} is 0, the result is undefined.
11115 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11116 Returns the number of trailing 0-bits in @var{x}, starting at the least
11117 significant bit position. If @var{x} is 0, the result is undefined.
11120 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11121 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11122 number of bits following the most significant bit that are identical
11123 to it. There are no special cases for 0 or other values.
11126 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11127 Returns the number of 1-bits in @var{x}.
11130 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11131 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11135 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11136 Similar to @code{__builtin_ffs}, except the argument type is
11140 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11141 Similar to @code{__builtin_clz}, except the argument type is
11142 @code{unsigned long}.
11145 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11146 Similar to @code{__builtin_ctz}, except the argument type is
11147 @code{unsigned long}.
11150 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11151 Similar to @code{__builtin_clrsb}, except the argument type is
11155 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11156 Similar to @code{__builtin_popcount}, except the argument type is
11157 @code{unsigned long}.
11160 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11161 Similar to @code{__builtin_parity}, except the argument type is
11162 @code{unsigned long}.
11165 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11166 Similar to @code{__builtin_ffs}, except the argument type is
11170 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11171 Similar to @code{__builtin_clz}, except the argument type is
11172 @code{unsigned long long}.
11175 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11176 Similar to @code{__builtin_ctz}, except the argument type is
11177 @code{unsigned long long}.
11180 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11181 Similar to @code{__builtin_clrsb}, except the argument type is
11185 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11186 Similar to @code{__builtin_popcount}, except the argument type is
11187 @code{unsigned long long}.
11190 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11191 Similar to @code{__builtin_parity}, except the argument type is
11192 @code{unsigned long long}.
11195 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11196 Returns the first argument raised to the power of the second. Unlike the
11197 @code{pow} function no guarantees about precision and rounding are made.
11200 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11201 Similar to @code{__builtin_powi}, except the argument and return types
11205 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11206 Similar to @code{__builtin_powi}, except the argument and return types
11207 are @code{long double}.
11210 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11211 Returns @var{x} with the order of the bytes reversed; for example,
11212 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11216 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11217 Similar to @code{__builtin_bswap16}, except the argument and return types
11221 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11222 Similar to @code{__builtin_bswap32}, except the argument and return types
11226 @node Target Builtins
11227 @section Built-in Functions Specific to Particular Target Machines
11229 On some target machines, GCC supports many built-in functions specific
11230 to those machines. Generally these generate calls to specific machine
11231 instructions, but allow the compiler to schedule those calls.
11234 * AArch64 Built-in Functions::
11235 * Alpha Built-in Functions::
11236 * Altera Nios II Built-in Functions::
11237 * ARC Built-in Functions::
11238 * ARC SIMD Built-in Functions::
11239 * ARM iWMMXt Built-in Functions::
11240 * ARM C Language Extensions (ACLE)::
11241 * ARM Floating Point Status and Control Intrinsics::
11242 * AVR Built-in Functions::
11243 * Blackfin Built-in Functions::
11244 * FR-V Built-in Functions::
11245 * MIPS DSP Built-in Functions::
11246 * MIPS Paired-Single Support::
11247 * MIPS Loongson Built-in Functions::
11248 * Other MIPS Built-in Functions::
11249 * MSP430 Built-in Functions::
11250 * NDS32 Built-in Functions::
11251 * picoChip Built-in Functions::
11252 * PowerPC Built-in Functions::
11253 * PowerPC AltiVec/VSX Built-in Functions::
11254 * PowerPC Hardware Transactional Memory Built-in Functions::
11255 * RX Built-in Functions::
11256 * S/390 System z Built-in Functions::
11257 * SH Built-in Functions::
11258 * SPARC VIS Built-in Functions::
11259 * SPU Built-in Functions::
11260 * TI C6X Built-in Functions::
11261 * TILE-Gx Built-in Functions::
11262 * TILEPro Built-in Functions::
11263 * x86 Built-in Functions::
11264 * x86 transactional memory intrinsics::
11267 @node AArch64 Built-in Functions
11268 @subsection AArch64 Built-in Functions
11270 These built-in functions are available for the AArch64 family of
11273 unsigned int __builtin_aarch64_get_fpcr ()
11274 void __builtin_aarch64_set_fpcr (unsigned int)
11275 unsigned int __builtin_aarch64_get_fpsr ()
11276 void __builtin_aarch64_set_fpsr (unsigned int)
11279 @node Alpha Built-in Functions
11280 @subsection Alpha Built-in Functions
11282 These built-in functions are available for the Alpha family of
11283 processors, depending on the command-line switches used.
11285 The following built-in functions are always available. They
11286 all generate the machine instruction that is part of the name.
11289 long __builtin_alpha_implver (void)
11290 long __builtin_alpha_rpcc (void)
11291 long __builtin_alpha_amask (long)
11292 long __builtin_alpha_cmpbge (long, long)
11293 long __builtin_alpha_extbl (long, long)
11294 long __builtin_alpha_extwl (long, long)
11295 long __builtin_alpha_extll (long, long)
11296 long __builtin_alpha_extql (long, long)
11297 long __builtin_alpha_extwh (long, long)
11298 long __builtin_alpha_extlh (long, long)
11299 long __builtin_alpha_extqh (long, long)
11300 long __builtin_alpha_insbl (long, long)
11301 long __builtin_alpha_inswl (long, long)
11302 long __builtin_alpha_insll (long, long)
11303 long __builtin_alpha_insql (long, long)
11304 long __builtin_alpha_inswh (long, long)
11305 long __builtin_alpha_inslh (long, long)
11306 long __builtin_alpha_insqh (long, long)
11307 long __builtin_alpha_mskbl (long, long)
11308 long __builtin_alpha_mskwl (long, long)
11309 long __builtin_alpha_mskll (long, long)
11310 long __builtin_alpha_mskql (long, long)
11311 long __builtin_alpha_mskwh (long, long)
11312 long __builtin_alpha_msklh (long, long)
11313 long __builtin_alpha_mskqh (long, long)
11314 long __builtin_alpha_umulh (long, long)
11315 long __builtin_alpha_zap (long, long)
11316 long __builtin_alpha_zapnot (long, long)
11319 The following built-in functions are always with @option{-mmax}
11320 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11321 later. They all generate the machine instruction that is part
11325 long __builtin_alpha_pklb (long)
11326 long __builtin_alpha_pkwb (long)
11327 long __builtin_alpha_unpkbl (long)
11328 long __builtin_alpha_unpkbw (long)
11329 long __builtin_alpha_minub8 (long, long)
11330 long __builtin_alpha_minsb8 (long, long)
11331 long __builtin_alpha_minuw4 (long, long)
11332 long __builtin_alpha_minsw4 (long, long)
11333 long __builtin_alpha_maxub8 (long, long)
11334 long __builtin_alpha_maxsb8 (long, long)
11335 long __builtin_alpha_maxuw4 (long, long)
11336 long __builtin_alpha_maxsw4 (long, long)
11337 long __builtin_alpha_perr (long, long)
11340 The following built-in functions are always with @option{-mcix}
11341 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11342 later. They all generate the machine instruction that is part
11346 long __builtin_alpha_cttz (long)
11347 long __builtin_alpha_ctlz (long)
11348 long __builtin_alpha_ctpop (long)
11351 The following built-in functions are available on systems that use the OSF/1
11352 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11353 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11354 @code{rdval} and @code{wrval}.
11357 void *__builtin_thread_pointer (void)
11358 void __builtin_set_thread_pointer (void *)
11361 @node Altera Nios II Built-in Functions
11362 @subsection Altera Nios II Built-in Functions
11364 These built-in functions are available for the Altera Nios II
11365 family of processors.
11367 The following built-in functions are always available. They
11368 all generate the machine instruction that is part of the name.
11371 int __builtin_ldbio (volatile const void *)
11372 int __builtin_ldbuio (volatile const void *)
11373 int __builtin_ldhio (volatile const void *)
11374 int __builtin_ldhuio (volatile const void *)
11375 int __builtin_ldwio (volatile const void *)
11376 void __builtin_stbio (volatile void *, int)
11377 void __builtin_sthio (volatile void *, int)
11378 void __builtin_stwio (volatile void *, int)
11379 void __builtin_sync (void)
11380 int __builtin_rdctl (int)
11381 int __builtin_rdprs (int, int)
11382 void __builtin_wrctl (int, int)
11383 void __builtin_flushd (volatile void *)
11384 void __builtin_flushda (volatile void *)
11385 int __builtin_wrpie (int);
11386 void __builtin_eni (int);
11387 int __builtin_ldex (volatile const void *)
11388 int __builtin_stex (volatile void *, int)
11389 int __builtin_ldsex (volatile const void *)
11390 int __builtin_stsex (volatile void *, int)
11393 The following built-in functions are always available. They
11394 all generate a Nios II Custom Instruction. The name of the
11395 function represents the types that the function takes and
11396 returns. The letter before the @code{n} is the return type
11397 or void if absent. The @code{n} represents the first parameter
11398 to all the custom instructions, the custom instruction number.
11399 The two letters after the @code{n} represent the up to two
11400 parameters to the function.
11402 The letters represent the following data types:
11405 @code{void} for return type and no parameter for parameter types.
11408 @code{int} for return type and parameter type
11411 @code{float} for return type and parameter type
11414 @code{void *} for return type and parameter type
11418 And the function names are:
11420 void __builtin_custom_n (void)
11421 void __builtin_custom_ni (int)
11422 void __builtin_custom_nf (float)
11423 void __builtin_custom_np (void *)
11424 void __builtin_custom_nii (int, int)
11425 void __builtin_custom_nif (int, float)
11426 void __builtin_custom_nip (int, void *)
11427 void __builtin_custom_nfi (float, int)
11428 void __builtin_custom_nff (float, float)
11429 void __builtin_custom_nfp (float, void *)
11430 void __builtin_custom_npi (void *, int)
11431 void __builtin_custom_npf (void *, float)
11432 void __builtin_custom_npp (void *, void *)
11433 int __builtin_custom_in (void)
11434 int __builtin_custom_ini (int)
11435 int __builtin_custom_inf (float)
11436 int __builtin_custom_inp (void *)
11437 int __builtin_custom_inii (int, int)
11438 int __builtin_custom_inif (int, float)
11439 int __builtin_custom_inip (int, void *)
11440 int __builtin_custom_infi (float, int)
11441 int __builtin_custom_inff (float, float)
11442 int __builtin_custom_infp (float, void *)
11443 int __builtin_custom_inpi (void *, int)
11444 int __builtin_custom_inpf (void *, float)
11445 int __builtin_custom_inpp (void *, void *)
11446 float __builtin_custom_fn (void)
11447 float __builtin_custom_fni (int)
11448 float __builtin_custom_fnf (float)
11449 float __builtin_custom_fnp (void *)
11450 float __builtin_custom_fnii (int, int)
11451 float __builtin_custom_fnif (int, float)
11452 float __builtin_custom_fnip (int, void *)
11453 float __builtin_custom_fnfi (float, int)
11454 float __builtin_custom_fnff (float, float)
11455 float __builtin_custom_fnfp (float, void *)
11456 float __builtin_custom_fnpi (void *, int)
11457 float __builtin_custom_fnpf (void *, float)
11458 float __builtin_custom_fnpp (void *, void *)
11459 void * __builtin_custom_pn (void)
11460 void * __builtin_custom_pni (int)
11461 void * __builtin_custom_pnf (float)
11462 void * __builtin_custom_pnp (void *)
11463 void * __builtin_custom_pnii (int, int)
11464 void * __builtin_custom_pnif (int, float)
11465 void * __builtin_custom_pnip (int, void *)
11466 void * __builtin_custom_pnfi (float, int)
11467 void * __builtin_custom_pnff (float, float)
11468 void * __builtin_custom_pnfp (float, void *)
11469 void * __builtin_custom_pnpi (void *, int)
11470 void * __builtin_custom_pnpf (void *, float)
11471 void * __builtin_custom_pnpp (void *, void *)
11474 @node ARC Built-in Functions
11475 @subsection ARC Built-in Functions
11477 The following built-in functions are provided for ARC targets. The
11478 built-ins generate the corresponding assembly instructions. In the
11479 examples given below, the generated code often requires an operand or
11480 result to be in a register. Where necessary further code will be
11481 generated to ensure this is true, but for brevity this is not
11482 described in each case.
11484 @emph{Note:} Using a built-in to generate an instruction not supported
11485 by a target may cause problems. At present the compiler is not
11486 guaranteed to detect such misuse, and as a result an internal compiler
11487 error may be generated.
11489 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11490 Return 1 if @var{val} is known to have the byte alignment given
11491 by @var{alignval}, otherwise return 0.
11492 Note that this is different from
11494 __alignof__(*(char *)@var{val}) >= alignval
11496 because __alignof__ sees only the type of the dereference, whereas
11497 __builtin_arc_align uses alignment information from the pointer
11498 as well as from the pointed-to type.
11499 The information available will depend on optimization level.
11502 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11509 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11510 The operand is the number of a register to be read. Generates:
11512 mov @var{dest}, r@var{regno}
11514 where the value in @var{dest} will be the result returned from the
11518 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11519 The first operand is the number of a register to be written, the
11520 second operand is a compile time constant to write into that
11521 register. Generates:
11523 mov r@var{regno}, @var{val}
11527 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11528 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11531 divaw @var{dest}, @var{a}, @var{b}
11533 where the value in @var{dest} will be the result returned from the
11537 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11544 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11545 The operand, @var{auxv}, is the address of an auxiliary register and
11546 must be a compile time constant. Generates:
11548 lr @var{dest}, [@var{auxr}]
11550 Where the value in @var{dest} will be the result returned from the
11554 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11555 Only available with @option{-mmul64}. Generates:
11557 mul64 @var{a}, @var{b}
11561 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11562 Only available with @option{-mmul64}. Generates:
11564 mulu64 @var{a}, @var{b}
11568 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11575 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11576 Only valid if the @samp{norm} instruction is available through the
11577 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11580 norm @var{dest}, @var{src}
11582 Where the value in @var{dest} will be the result returned from the
11586 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11587 Only valid if the @samp{normw} instruction is available through the
11588 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11591 normw @var{dest}, @var{src}
11593 Where the value in @var{dest} will be the result returned from the
11597 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11604 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11611 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11612 The first argument, @var{auxv}, is the address of an auxiliary
11613 register, the second argument, @var{val}, is a compile time constant
11614 to be written to the register. Generates:
11616 sr @var{auxr}, [@var{val}]
11620 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11621 Only valid with @option{-mswap}. Generates:
11623 swap @var{dest}, @var{src}
11625 Where the value in @var{dest} will be the result returned from the
11629 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11636 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11637 Only available with @option{-mcpu=ARC700}. Generates:
11643 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11644 Only available with @option{-mcpu=ARC700}. Generates:
11650 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11651 Only available with @option{-mcpu=ARC700}. Generates:
11657 The instructions generated by the following builtins are not
11658 considered as candidates for scheduling. They are not moved around by
11659 the compiler during scheduling, and thus can be expected to appear
11660 where they are put in the C code:
11662 __builtin_arc_brk()
11663 __builtin_arc_core_read()
11664 __builtin_arc_core_write()
11665 __builtin_arc_flag()
11667 __builtin_arc_sleep()
11669 __builtin_arc_swi()
11672 @node ARC SIMD Built-in Functions
11673 @subsection ARC SIMD Built-in Functions
11675 SIMD builtins provided by the compiler can be used to generate the
11676 vector instructions. This section describes the available builtins
11677 and their usage in programs. With the @option{-msimd} option, the
11678 compiler provides 128-bit vector types, which can be specified using
11679 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11680 can be included to use the following predefined types:
11682 typedef int __v4si __attribute__((vector_size(16)));
11683 typedef short __v8hi __attribute__((vector_size(16)));
11686 These types can be used to define 128-bit variables. The built-in
11687 functions listed in the following section can be used on these
11688 variables to generate the vector operations.
11690 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11691 @file{arc-simd.h} also provides equivalent macros called
11692 @code{_@var{someinsn}} that can be used for programming ease and
11693 improved readability. The following macros for DMA control are also
11696 #define _setup_dma_in_channel_reg _vdiwr
11697 #define _setup_dma_out_channel_reg _vdowr
11700 The following is a complete list of all the SIMD built-ins provided
11701 for ARC, grouped by calling signature.
11703 The following take two @code{__v8hi} arguments and return a
11704 @code{__v8hi} result:
11706 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11707 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11708 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11709 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11710 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11711 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11712 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11713 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11714 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11715 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11716 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11717 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11718 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11719 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11720 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11721 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11722 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11723 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11724 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11725 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11726 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11727 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11728 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11729 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11730 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11731 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11732 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11733 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11734 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11735 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11736 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11737 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11738 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11739 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11740 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11741 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11742 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11743 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11744 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11745 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11746 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11747 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11748 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11749 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11750 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11751 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11752 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11753 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11756 The following take one @code{__v8hi} and one @code{int} argument and return a
11757 @code{__v8hi} result:
11760 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11761 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11762 __v8hi __builtin_arc_vbminw (__v8hi, int)
11763 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11764 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11765 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11766 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11767 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11770 The following take one @code{__v8hi} argument and one @code{int} argument which
11771 must be a 3-bit compile time constant indicating a register number
11772 I0-I7. They return a @code{__v8hi} result.
11774 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11775 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11776 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11779 The following take one @code{__v8hi} argument and one @code{int}
11780 argument which must be a 6-bit compile time constant. They return a
11781 @code{__v8hi} result.
11783 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11784 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11785 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11786 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11787 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11788 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11789 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11792 The following take one @code{__v8hi} argument and one @code{int} argument which
11793 must be a 8-bit compile time constant. They return a @code{__v8hi}
11796 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11797 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11798 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11799 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11802 The following take two @code{int} arguments, the second of which which
11803 must be a 8-bit compile time constant. They return a @code{__v8hi}
11806 __v8hi __builtin_arc_vmovaw (int, const int)
11807 __v8hi __builtin_arc_vmovw (int, const int)
11808 __v8hi __builtin_arc_vmovzw (int, const int)
11811 The following take a single @code{__v8hi} argument and return a
11812 @code{__v8hi} result:
11814 __v8hi __builtin_arc_vabsaw (__v8hi)
11815 __v8hi __builtin_arc_vabsw (__v8hi)
11816 __v8hi __builtin_arc_vaddsuw (__v8hi)
11817 __v8hi __builtin_arc_vexch1 (__v8hi)
11818 __v8hi __builtin_arc_vexch2 (__v8hi)
11819 __v8hi __builtin_arc_vexch4 (__v8hi)
11820 __v8hi __builtin_arc_vsignw (__v8hi)
11821 __v8hi __builtin_arc_vupbaw (__v8hi)
11822 __v8hi __builtin_arc_vupbw (__v8hi)
11823 __v8hi __builtin_arc_vupsbaw (__v8hi)
11824 __v8hi __builtin_arc_vupsbw (__v8hi)
11827 The following take two @code{int} arguments and return no result:
11829 void __builtin_arc_vdirun (int, int)
11830 void __builtin_arc_vdorun (int, int)
11833 The following take two @code{int} arguments and return no result. The
11834 first argument must a 3-bit compile time constant indicating one of
11835 the DR0-DR7 DMA setup channels:
11837 void __builtin_arc_vdiwr (const int, int)
11838 void __builtin_arc_vdowr (const int, int)
11841 The following take an @code{int} argument and return no result:
11843 void __builtin_arc_vendrec (int)
11844 void __builtin_arc_vrec (int)
11845 void __builtin_arc_vrecrun (int)
11846 void __builtin_arc_vrun (int)
11849 The following take a @code{__v8hi} argument and two @code{int}
11850 arguments and return a @code{__v8hi} result. The second argument must
11851 be a 3-bit compile time constants, indicating one the registers I0-I7,
11852 and the third argument must be an 8-bit compile time constant.
11854 @emph{Note:} Although the equivalent hardware instructions do not take
11855 an SIMD register as an operand, these builtins overwrite the relevant
11856 bits of the @code{__v8hi} register provided as the first argument with
11857 the value loaded from the @code{[Ib, u8]} location in the SDM.
11860 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
11861 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
11862 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
11863 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
11866 The following take two @code{int} arguments and return a @code{__v8hi}
11867 result. The first argument must be a 3-bit compile time constants,
11868 indicating one the registers I0-I7, and the second argument must be an
11869 8-bit compile time constant.
11872 __v8hi __builtin_arc_vld128 (const int, const int)
11873 __v8hi __builtin_arc_vld64w (const int, const int)
11876 The following take a @code{__v8hi} argument and two @code{int}
11877 arguments and return no result. The second argument must be a 3-bit
11878 compile time constants, indicating one the registers I0-I7, and the
11879 third argument must be an 8-bit compile time constant.
11882 void __builtin_arc_vst128 (__v8hi, const int, const int)
11883 void __builtin_arc_vst64 (__v8hi, const int, const int)
11886 The following take a @code{__v8hi} argument and three @code{int}
11887 arguments and return no result. The second argument must be a 3-bit
11888 compile-time constant, identifying the 16-bit sub-register to be
11889 stored, the third argument must be a 3-bit compile time constants,
11890 indicating one the registers I0-I7, and the fourth argument must be an
11891 8-bit compile time constant.
11894 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
11895 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
11898 @node ARM iWMMXt Built-in Functions
11899 @subsection ARM iWMMXt Built-in Functions
11901 These built-in functions are available for the ARM family of
11902 processors when the @option{-mcpu=iwmmxt} switch is used:
11905 typedef int v2si __attribute__ ((vector_size (8)));
11906 typedef short v4hi __attribute__ ((vector_size (8)));
11907 typedef char v8qi __attribute__ ((vector_size (8)));
11909 int __builtin_arm_getwcgr0 (void)
11910 void __builtin_arm_setwcgr0 (int)
11911 int __builtin_arm_getwcgr1 (void)
11912 void __builtin_arm_setwcgr1 (int)
11913 int __builtin_arm_getwcgr2 (void)
11914 void __builtin_arm_setwcgr2 (int)
11915 int __builtin_arm_getwcgr3 (void)
11916 void __builtin_arm_setwcgr3 (int)
11917 int __builtin_arm_textrmsb (v8qi, int)
11918 int __builtin_arm_textrmsh (v4hi, int)
11919 int __builtin_arm_textrmsw (v2si, int)
11920 int __builtin_arm_textrmub (v8qi, int)
11921 int __builtin_arm_textrmuh (v4hi, int)
11922 int __builtin_arm_textrmuw (v2si, int)
11923 v8qi __builtin_arm_tinsrb (v8qi, int, int)
11924 v4hi __builtin_arm_tinsrh (v4hi, int, int)
11925 v2si __builtin_arm_tinsrw (v2si, int, int)
11926 long long __builtin_arm_tmia (long long, int, int)
11927 long long __builtin_arm_tmiabb (long long, int, int)
11928 long long __builtin_arm_tmiabt (long long, int, int)
11929 long long __builtin_arm_tmiaph (long long, int, int)
11930 long long __builtin_arm_tmiatb (long long, int, int)
11931 long long __builtin_arm_tmiatt (long long, int, int)
11932 int __builtin_arm_tmovmskb (v8qi)
11933 int __builtin_arm_tmovmskh (v4hi)
11934 int __builtin_arm_tmovmskw (v2si)
11935 long long __builtin_arm_waccb (v8qi)
11936 long long __builtin_arm_wacch (v4hi)
11937 long long __builtin_arm_waccw (v2si)
11938 v8qi __builtin_arm_waddb (v8qi, v8qi)
11939 v8qi __builtin_arm_waddbss (v8qi, v8qi)
11940 v8qi __builtin_arm_waddbus (v8qi, v8qi)
11941 v4hi __builtin_arm_waddh (v4hi, v4hi)
11942 v4hi __builtin_arm_waddhss (v4hi, v4hi)
11943 v4hi __builtin_arm_waddhus (v4hi, v4hi)
11944 v2si __builtin_arm_waddw (v2si, v2si)
11945 v2si __builtin_arm_waddwss (v2si, v2si)
11946 v2si __builtin_arm_waddwus (v2si, v2si)
11947 v8qi __builtin_arm_walign (v8qi, v8qi, int)
11948 long long __builtin_arm_wand(long long, long long)
11949 long long __builtin_arm_wandn (long long, long long)
11950 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
11951 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
11952 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
11953 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
11954 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
11955 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
11956 v2si __builtin_arm_wcmpeqw (v2si, v2si)
11957 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
11958 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
11959 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
11960 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
11961 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
11962 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
11963 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
11964 long long __builtin_arm_wmacsz (v4hi, v4hi)
11965 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
11966 long long __builtin_arm_wmacuz (v4hi, v4hi)
11967 v4hi __builtin_arm_wmadds (v4hi, v4hi)
11968 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
11969 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
11970 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
11971 v2si __builtin_arm_wmaxsw (v2si, v2si)
11972 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
11973 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
11974 v2si __builtin_arm_wmaxuw (v2si, v2si)
11975 v8qi __builtin_arm_wminsb (v8qi, v8qi)
11976 v4hi __builtin_arm_wminsh (v4hi, v4hi)
11977 v2si __builtin_arm_wminsw (v2si, v2si)
11978 v8qi __builtin_arm_wminub (v8qi, v8qi)
11979 v4hi __builtin_arm_wminuh (v4hi, v4hi)
11980 v2si __builtin_arm_wminuw (v2si, v2si)
11981 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
11982 v4hi __builtin_arm_wmulul (v4hi, v4hi)
11983 v4hi __builtin_arm_wmulum (v4hi, v4hi)
11984 long long __builtin_arm_wor (long long, long long)
11985 v2si __builtin_arm_wpackdss (long long, long long)
11986 v2si __builtin_arm_wpackdus (long long, long long)
11987 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
11988 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
11989 v4hi __builtin_arm_wpackwss (v2si, v2si)
11990 v4hi __builtin_arm_wpackwus (v2si, v2si)
11991 long long __builtin_arm_wrord (long long, long long)
11992 long long __builtin_arm_wrordi (long long, int)
11993 v4hi __builtin_arm_wrorh (v4hi, long long)
11994 v4hi __builtin_arm_wrorhi (v4hi, int)
11995 v2si __builtin_arm_wrorw (v2si, long long)
11996 v2si __builtin_arm_wrorwi (v2si, int)
11997 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
11998 v2si __builtin_arm_wsadbz (v8qi, v8qi)
11999 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12000 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12001 v4hi __builtin_arm_wshufh (v4hi, int)
12002 long long __builtin_arm_wslld (long long, long long)
12003 long long __builtin_arm_wslldi (long long, int)
12004 v4hi __builtin_arm_wsllh (v4hi, long long)
12005 v4hi __builtin_arm_wsllhi (v4hi, int)
12006 v2si __builtin_arm_wsllw (v2si, long long)
12007 v2si __builtin_arm_wsllwi (v2si, int)
12008 long long __builtin_arm_wsrad (long long, long long)
12009 long long __builtin_arm_wsradi (long long, int)
12010 v4hi __builtin_arm_wsrah (v4hi, long long)
12011 v4hi __builtin_arm_wsrahi (v4hi, int)
12012 v2si __builtin_arm_wsraw (v2si, long long)
12013 v2si __builtin_arm_wsrawi (v2si, int)
12014 long long __builtin_arm_wsrld (long long, long long)
12015 long long __builtin_arm_wsrldi (long long, int)
12016 v4hi __builtin_arm_wsrlh (v4hi, long long)
12017 v4hi __builtin_arm_wsrlhi (v4hi, int)
12018 v2si __builtin_arm_wsrlw (v2si, long long)
12019 v2si __builtin_arm_wsrlwi (v2si, int)
12020 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12021 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12022 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12023 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12024 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12025 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12026 v2si __builtin_arm_wsubw (v2si, v2si)
12027 v2si __builtin_arm_wsubwss (v2si, v2si)
12028 v2si __builtin_arm_wsubwus (v2si, v2si)
12029 v4hi __builtin_arm_wunpckehsb (v8qi)
12030 v2si __builtin_arm_wunpckehsh (v4hi)
12031 long long __builtin_arm_wunpckehsw (v2si)
12032 v4hi __builtin_arm_wunpckehub (v8qi)
12033 v2si __builtin_arm_wunpckehuh (v4hi)
12034 long long __builtin_arm_wunpckehuw (v2si)
12035 v4hi __builtin_arm_wunpckelsb (v8qi)
12036 v2si __builtin_arm_wunpckelsh (v4hi)
12037 long long __builtin_arm_wunpckelsw (v2si)
12038 v4hi __builtin_arm_wunpckelub (v8qi)
12039 v2si __builtin_arm_wunpckeluh (v4hi)
12040 long long __builtin_arm_wunpckeluw (v2si)
12041 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12042 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12043 v2si __builtin_arm_wunpckihw (v2si, v2si)
12044 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12045 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12046 v2si __builtin_arm_wunpckilw (v2si, v2si)
12047 long long __builtin_arm_wxor (long long, long long)
12048 long long __builtin_arm_wzero ()
12052 @node ARM C Language Extensions (ACLE)
12053 @subsection ARM C Language Extensions (ACLE)
12055 GCC implements extensions for C as described in the ARM C Language
12056 Extensions (ACLE) specification, which can be found at
12057 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12059 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12060 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12061 intrinsics can be found at
12062 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12063 The built-in intrinsics for the Advanced SIMD extension are available when
12066 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12067 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12068 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12069 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12072 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12073 availability of extensions.
12075 @node ARM Floating Point Status and Control Intrinsics
12076 @subsection ARM Floating Point Status and Control Intrinsics
12078 These built-in functions are available for the ARM family of
12079 processors with floating-point unit.
12082 unsigned int __builtin_arm_get_fpscr ()
12083 void __builtin_arm_set_fpscr (unsigned int)
12086 @node AVR Built-in Functions
12087 @subsection AVR Built-in Functions
12089 For each built-in function for AVR, there is an equally named,
12090 uppercase built-in macro defined. That way users can easily query if
12091 or if not a specific built-in is implemented or not. For example, if
12092 @code{__builtin_avr_nop} is available the macro
12093 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12095 The following built-in functions map to the respective machine
12096 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12097 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12098 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12099 as library call if no hardware multiplier is available.
12102 void __builtin_avr_nop (void)
12103 void __builtin_avr_sei (void)
12104 void __builtin_avr_cli (void)
12105 void __builtin_avr_sleep (void)
12106 void __builtin_avr_wdr (void)
12107 unsigned char __builtin_avr_swap (unsigned char)
12108 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12109 int __builtin_avr_fmuls (char, char)
12110 int __builtin_avr_fmulsu (char, unsigned char)
12113 In order to delay execution for a specific number of cycles, GCC
12116 void __builtin_avr_delay_cycles (unsigned long ticks)
12120 @code{ticks} is the number of ticks to delay execution. Note that this
12121 built-in does not take into account the effect of interrupts that
12122 might increase delay time. @code{ticks} must be a compile-time
12123 integer constant; delays with a variable number of cycles are not supported.
12126 char __builtin_avr_flash_segment (const __memx void*)
12130 This built-in takes a byte address to the 24-bit
12131 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12132 the number of the flash segment (the 64 KiB chunk) where the address
12133 points to. Counting starts at @code{0}.
12134 If the address does not point to flash memory, return @code{-1}.
12137 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12141 Insert bits from @var{bits} into @var{val} and return the resulting
12142 value. The nibbles of @var{map} determine how the insertion is
12143 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12145 @item If @var{X} is @code{0xf},
12146 then the @var{n}-th bit of @var{val} is returned unaltered.
12148 @item If X is in the range 0@dots{}7,
12149 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12151 @item If X is in the range 8@dots{}@code{0xe},
12152 then the @var{n}-th result bit is undefined.
12156 One typical use case for this built-in is adjusting input and
12157 output values to non-contiguous port layouts. Some examples:
12160 // same as val, bits is unused
12161 __builtin_avr_insert_bits (0xffffffff, bits, val)
12165 // same as bits, val is unused
12166 __builtin_avr_insert_bits (0x76543210, bits, val)
12170 // same as rotating bits by 4
12171 __builtin_avr_insert_bits (0x32107654, bits, 0)
12175 // high nibble of result is the high nibble of val
12176 // low nibble of result is the low nibble of bits
12177 __builtin_avr_insert_bits (0xffff3210, bits, val)
12181 // reverse the bit order of bits
12182 __builtin_avr_insert_bits (0x01234567, bits, 0)
12185 @node Blackfin Built-in Functions
12186 @subsection Blackfin Built-in Functions
12188 Currently, there are two Blackfin-specific built-in functions. These are
12189 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12190 using inline assembly; by using these built-in functions the compiler can
12191 automatically add workarounds for hardware errata involving these
12192 instructions. These functions are named as follows:
12195 void __builtin_bfin_csync (void)
12196 void __builtin_bfin_ssync (void)
12199 @node FR-V Built-in Functions
12200 @subsection FR-V Built-in Functions
12202 GCC provides many FR-V-specific built-in functions. In general,
12203 these functions are intended to be compatible with those described
12204 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12205 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12206 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12207 pointer rather than by value.
12209 Most of the functions are named after specific FR-V instructions.
12210 Such functions are said to be ``directly mapped'' and are summarized
12211 here in tabular form.
12215 * Directly-mapped Integer Functions::
12216 * Directly-mapped Media Functions::
12217 * Raw read/write Functions::
12218 * Other Built-in Functions::
12221 @node Argument Types
12222 @subsubsection Argument Types
12224 The arguments to the built-in functions can be divided into three groups:
12225 register numbers, compile-time constants and run-time values. In order
12226 to make this classification clear at a glance, the arguments and return
12227 values are given the following pseudo types:
12229 @multitable @columnfractions .20 .30 .15 .35
12230 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12231 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12232 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12233 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12234 @item @code{uw2} @tab @code{unsigned long long} @tab No
12235 @tab an unsigned doubleword
12236 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12237 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12238 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12239 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12242 These pseudo types are not defined by GCC, they are simply a notational
12243 convenience used in this manual.
12245 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12246 and @code{sw2} are evaluated at run time. They correspond to
12247 register operands in the underlying FR-V instructions.
12249 @code{const} arguments represent immediate operands in the underlying
12250 FR-V instructions. They must be compile-time constants.
12252 @code{acc} arguments are evaluated at compile time and specify the number
12253 of an accumulator register. For example, an @code{acc} argument of 2
12254 selects the ACC2 register.
12256 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12257 number of an IACC register. See @pxref{Other Built-in Functions}
12260 @node Directly-mapped Integer Functions
12261 @subsubsection Directly-Mapped Integer Functions
12263 The functions listed below map directly to FR-V I-type instructions.
12265 @multitable @columnfractions .45 .32 .23
12266 @item Function prototype @tab Example usage @tab Assembly output
12267 @item @code{sw1 __ADDSS (sw1, sw1)}
12268 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12269 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12270 @item @code{sw1 __SCAN (sw1, sw1)}
12271 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12272 @tab @code{SCAN @var{a},@var{b},@var{c}}
12273 @item @code{sw1 __SCUTSS (sw1)}
12274 @tab @code{@var{b} = __SCUTSS (@var{a})}
12275 @tab @code{SCUTSS @var{a},@var{b}}
12276 @item @code{sw1 __SLASS (sw1, sw1)}
12277 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12278 @tab @code{SLASS @var{a},@var{b},@var{c}}
12279 @item @code{void __SMASS (sw1, sw1)}
12280 @tab @code{__SMASS (@var{a}, @var{b})}
12281 @tab @code{SMASS @var{a},@var{b}}
12282 @item @code{void __SMSSS (sw1, sw1)}
12283 @tab @code{__SMSSS (@var{a}, @var{b})}
12284 @tab @code{SMSSS @var{a},@var{b}}
12285 @item @code{void __SMU (sw1, sw1)}
12286 @tab @code{__SMU (@var{a}, @var{b})}
12287 @tab @code{SMU @var{a},@var{b}}
12288 @item @code{sw2 __SMUL (sw1, sw1)}
12289 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12290 @tab @code{SMUL @var{a},@var{b},@var{c}}
12291 @item @code{sw1 __SUBSS (sw1, sw1)}
12292 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12293 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12294 @item @code{uw2 __UMUL (uw1, uw1)}
12295 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12296 @tab @code{UMUL @var{a},@var{b},@var{c}}
12299 @node Directly-mapped Media Functions
12300 @subsubsection Directly-Mapped Media Functions
12302 The functions listed below map directly to FR-V M-type instructions.
12304 @multitable @columnfractions .45 .32 .23
12305 @item Function prototype @tab Example usage @tab Assembly output
12306 @item @code{uw1 __MABSHS (sw1)}
12307 @tab @code{@var{b} = __MABSHS (@var{a})}
12308 @tab @code{MABSHS @var{a},@var{b}}
12309 @item @code{void __MADDACCS (acc, acc)}
12310 @tab @code{__MADDACCS (@var{b}, @var{a})}
12311 @tab @code{MADDACCS @var{a},@var{b}}
12312 @item @code{sw1 __MADDHSS (sw1, sw1)}
12313 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12314 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12315 @item @code{uw1 __MADDHUS (uw1, uw1)}
12316 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12317 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12318 @item @code{uw1 __MAND (uw1, uw1)}
12319 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12320 @tab @code{MAND @var{a},@var{b},@var{c}}
12321 @item @code{void __MASACCS (acc, acc)}
12322 @tab @code{__MASACCS (@var{b}, @var{a})}
12323 @tab @code{MASACCS @var{a},@var{b}}
12324 @item @code{uw1 __MAVEH (uw1, uw1)}
12325 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12326 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12327 @item @code{uw2 __MBTOH (uw1)}
12328 @tab @code{@var{b} = __MBTOH (@var{a})}
12329 @tab @code{MBTOH @var{a},@var{b}}
12330 @item @code{void __MBTOHE (uw1 *, uw1)}
12331 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12332 @tab @code{MBTOHE @var{a},@var{b}}
12333 @item @code{void __MCLRACC (acc)}
12334 @tab @code{__MCLRACC (@var{a})}
12335 @tab @code{MCLRACC @var{a}}
12336 @item @code{void __MCLRACCA (void)}
12337 @tab @code{__MCLRACCA ()}
12338 @tab @code{MCLRACCA}
12339 @item @code{uw1 __Mcop1 (uw1, uw1)}
12340 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12341 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12342 @item @code{uw1 __Mcop2 (uw1, uw1)}
12343 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12344 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12345 @item @code{uw1 __MCPLHI (uw2, const)}
12346 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12347 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12348 @item @code{uw1 __MCPLI (uw2, const)}
12349 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12350 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12351 @item @code{void __MCPXIS (acc, sw1, sw1)}
12352 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12353 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12354 @item @code{void __MCPXIU (acc, uw1, uw1)}
12355 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12356 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12357 @item @code{void __MCPXRS (acc, sw1, sw1)}
12358 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12359 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12360 @item @code{void __MCPXRU (acc, uw1, uw1)}
12361 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12362 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12363 @item @code{uw1 __MCUT (acc, uw1)}
12364 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12365 @tab @code{MCUT @var{a},@var{b},@var{c}}
12366 @item @code{uw1 __MCUTSS (acc, sw1)}
12367 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12368 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12369 @item @code{void __MDADDACCS (acc, acc)}
12370 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12371 @tab @code{MDADDACCS @var{a},@var{b}}
12372 @item @code{void __MDASACCS (acc, acc)}
12373 @tab @code{__MDASACCS (@var{b}, @var{a})}
12374 @tab @code{MDASACCS @var{a},@var{b}}
12375 @item @code{uw2 __MDCUTSSI (acc, const)}
12376 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12377 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12378 @item @code{uw2 __MDPACKH (uw2, uw2)}
12379 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12380 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12381 @item @code{uw2 __MDROTLI (uw2, const)}
12382 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12383 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12384 @item @code{void __MDSUBACCS (acc, acc)}
12385 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12386 @tab @code{MDSUBACCS @var{a},@var{b}}
12387 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12388 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12389 @tab @code{MDUNPACKH @var{a},@var{b}}
12390 @item @code{uw2 __MEXPDHD (uw1, const)}
12391 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12392 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12393 @item @code{uw1 __MEXPDHW (uw1, const)}
12394 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12395 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12396 @item @code{uw1 __MHDSETH (uw1, const)}
12397 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12398 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12399 @item @code{sw1 __MHDSETS (const)}
12400 @tab @code{@var{b} = __MHDSETS (@var{a})}
12401 @tab @code{MHDSETS #@var{a},@var{b}}
12402 @item @code{uw1 __MHSETHIH (uw1, const)}
12403 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12404 @tab @code{MHSETHIH #@var{a},@var{b}}
12405 @item @code{sw1 __MHSETHIS (sw1, const)}
12406 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12407 @tab @code{MHSETHIS #@var{a},@var{b}}
12408 @item @code{uw1 __MHSETLOH (uw1, const)}
12409 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12410 @tab @code{MHSETLOH #@var{a},@var{b}}
12411 @item @code{sw1 __MHSETLOS (sw1, const)}
12412 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12413 @tab @code{MHSETLOS #@var{a},@var{b}}
12414 @item @code{uw1 __MHTOB (uw2)}
12415 @tab @code{@var{b} = __MHTOB (@var{a})}
12416 @tab @code{MHTOB @var{a},@var{b}}
12417 @item @code{void __MMACHS (acc, sw1, sw1)}
12418 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12419 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12420 @item @code{void __MMACHU (acc, uw1, uw1)}
12421 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12422 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12423 @item @code{void __MMRDHS (acc, sw1, sw1)}
12424 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12425 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12426 @item @code{void __MMRDHU (acc, uw1, uw1)}
12427 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12428 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12429 @item @code{void __MMULHS (acc, sw1, sw1)}
12430 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12431 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12432 @item @code{void __MMULHU (acc, uw1, uw1)}
12433 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12434 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12435 @item @code{void __MMULXHS (acc, sw1, sw1)}
12436 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12437 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12438 @item @code{void __MMULXHU (acc, uw1, uw1)}
12439 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12440 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12441 @item @code{uw1 __MNOT (uw1)}
12442 @tab @code{@var{b} = __MNOT (@var{a})}
12443 @tab @code{MNOT @var{a},@var{b}}
12444 @item @code{uw1 __MOR (uw1, uw1)}
12445 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12446 @tab @code{MOR @var{a},@var{b},@var{c}}
12447 @item @code{uw1 __MPACKH (uh, uh)}
12448 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12449 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12450 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12451 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12452 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12453 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12454 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12455 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12456 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12457 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12458 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12459 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12460 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12461 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12462 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12463 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12464 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12465 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12466 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12467 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12468 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12469 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12470 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12471 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12472 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12473 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12474 @item @code{void __MQMACHS (acc, sw2, sw2)}
12475 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12476 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12477 @item @code{void __MQMACHU (acc, uw2, uw2)}
12478 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12479 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12480 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12481 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12482 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12483 @item @code{void __MQMULHS (acc, sw2, sw2)}
12484 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12485 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12486 @item @code{void __MQMULHU (acc, uw2, uw2)}
12487 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12488 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12489 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12490 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12491 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12492 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12493 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12494 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12495 @item @code{sw2 __MQSATHS (sw2, sw2)}
12496 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12497 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12498 @item @code{uw2 __MQSLLHI (uw2, int)}
12499 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12500 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12501 @item @code{sw2 __MQSRAHI (sw2, int)}
12502 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12503 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12504 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12505 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12506 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12507 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12508 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12509 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12510 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12511 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12512 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12513 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12514 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12515 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12516 @item @code{uw1 __MRDACC (acc)}
12517 @tab @code{@var{b} = __MRDACC (@var{a})}
12518 @tab @code{MRDACC @var{a},@var{b}}
12519 @item @code{uw1 __MRDACCG (acc)}
12520 @tab @code{@var{b} = __MRDACCG (@var{a})}
12521 @tab @code{MRDACCG @var{a},@var{b}}
12522 @item @code{uw1 __MROTLI (uw1, const)}
12523 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12524 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12525 @item @code{uw1 __MROTRI (uw1, const)}
12526 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12527 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12528 @item @code{sw1 __MSATHS (sw1, sw1)}
12529 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12530 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12531 @item @code{uw1 __MSATHU (uw1, uw1)}
12532 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12533 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12534 @item @code{uw1 __MSLLHI (uw1, const)}
12535 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12536 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12537 @item @code{sw1 __MSRAHI (sw1, const)}
12538 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12539 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12540 @item @code{uw1 __MSRLHI (uw1, const)}
12541 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12542 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12543 @item @code{void __MSUBACCS (acc, acc)}
12544 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12545 @tab @code{MSUBACCS @var{a},@var{b}}
12546 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12547 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12548 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12549 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12550 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12551 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12552 @item @code{void __MTRAP (void)}
12553 @tab @code{__MTRAP ()}
12555 @item @code{uw2 __MUNPACKH (uw1)}
12556 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12557 @tab @code{MUNPACKH @var{a},@var{b}}
12558 @item @code{uw1 __MWCUT (uw2, uw1)}
12559 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12560 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12561 @item @code{void __MWTACC (acc, uw1)}
12562 @tab @code{__MWTACC (@var{b}, @var{a})}
12563 @tab @code{MWTACC @var{a},@var{b}}
12564 @item @code{void __MWTACCG (acc, uw1)}
12565 @tab @code{__MWTACCG (@var{b}, @var{a})}
12566 @tab @code{MWTACCG @var{a},@var{b}}
12567 @item @code{uw1 __MXOR (uw1, uw1)}
12568 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12569 @tab @code{MXOR @var{a},@var{b},@var{c}}
12572 @node Raw read/write Functions
12573 @subsubsection Raw Read/Write Functions
12575 This sections describes built-in functions related to read and write
12576 instructions to access memory. These functions generate
12577 @code{membar} instructions to flush the I/O load and stores where
12578 appropriate, as described in Fujitsu's manual described above.
12582 @item unsigned char __builtin_read8 (void *@var{data})
12583 @item unsigned short __builtin_read16 (void *@var{data})
12584 @item unsigned long __builtin_read32 (void *@var{data})
12585 @item unsigned long long __builtin_read64 (void *@var{data})
12587 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12588 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12589 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12590 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12593 @node Other Built-in Functions
12594 @subsubsection Other Built-in Functions
12596 This section describes built-in functions that are not named after
12597 a specific FR-V instruction.
12600 @item sw2 __IACCreadll (iacc @var{reg})
12601 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12602 for future expansion and must be 0.
12604 @item sw1 __IACCreadl (iacc @var{reg})
12605 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12606 Other values of @var{reg} are rejected as invalid.
12608 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12609 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12610 is reserved for future expansion and must be 0.
12612 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12613 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12614 is 1. Other values of @var{reg} are rejected as invalid.
12616 @item void __data_prefetch0 (const void *@var{x})
12617 Use the @code{dcpl} instruction to load the contents of address @var{x}
12618 into the data cache.
12620 @item void __data_prefetch (const void *@var{x})
12621 Use the @code{nldub} instruction to load the contents of address @var{x}
12622 into the data cache. The instruction is issued in slot I1@.
12625 @node MIPS DSP Built-in Functions
12626 @subsection MIPS DSP Built-in Functions
12628 The MIPS DSP Application-Specific Extension (ASE) includes new
12629 instructions that are designed to improve the performance of DSP and
12630 media applications. It provides instructions that operate on packed
12631 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12633 GCC supports MIPS DSP operations using both the generic
12634 vector extensions (@pxref{Vector Extensions}) and a collection of
12635 MIPS-specific built-in functions. Both kinds of support are
12636 enabled by the @option{-mdsp} command-line option.
12638 Revision 2 of the ASE was introduced in the second half of 2006.
12639 This revision adds extra instructions to the original ASE, but is
12640 otherwise backwards-compatible with it. You can select revision 2
12641 using the command-line option @option{-mdspr2}; this option implies
12644 The SCOUNT and POS bits of the DSP control register are global. The
12645 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12646 POS bits. During optimization, the compiler does not delete these
12647 instructions and it does not delete calls to functions containing
12648 these instructions.
12650 At present, GCC only provides support for operations on 32-bit
12651 vectors. The vector type associated with 8-bit integer data is
12652 usually called @code{v4i8}, the vector type associated with Q7
12653 is usually called @code{v4q7}, the vector type associated with 16-bit
12654 integer data is usually called @code{v2i16}, and the vector type
12655 associated with Q15 is usually called @code{v2q15}. They can be
12656 defined in C as follows:
12659 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12660 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12661 typedef short v2i16 __attribute__ ((vector_size(4)));
12662 typedef short v2q15 __attribute__ ((vector_size(4)));
12665 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12666 initialized in the same way as aggregates. For example:
12669 v4i8 a = @{1, 2, 3, 4@};
12671 b = (v4i8) @{5, 6, 7, 8@};
12673 v2q15 c = @{0x0fcb, 0x3a75@};
12675 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12678 @emph{Note:} The CPU's endianness determines the order in which values
12679 are packed. On little-endian targets, the first value is the least
12680 significant and the last value is the most significant. The opposite
12681 order applies to big-endian targets. For example, the code above
12682 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12683 and @code{4} on big-endian targets.
12685 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12686 representation. As shown in this example, the integer representation
12687 of a Q7 value can be obtained by multiplying the fractional value by
12688 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12689 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12692 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12693 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12694 and @code{c} and @code{d} are @code{v2q15} values.
12696 @multitable @columnfractions .50 .50
12697 @item C code @tab MIPS instruction
12698 @item @code{a + b} @tab @code{addu.qb}
12699 @item @code{c + d} @tab @code{addq.ph}
12700 @item @code{a - b} @tab @code{subu.qb}
12701 @item @code{c - d} @tab @code{subq.ph}
12704 The table below lists the @code{v2i16} operation for which
12705 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12706 @code{v2i16} values.
12708 @multitable @columnfractions .50 .50
12709 @item C code @tab MIPS instruction
12710 @item @code{e * f} @tab @code{mul.ph}
12713 It is easier to describe the DSP built-in functions if we first define
12714 the following types:
12719 typedef unsigned int ui32;
12720 typedef long long a64;
12723 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12724 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12725 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12726 @code{long long}, but we use @code{a64} to indicate values that are
12727 placed in one of the four DSP accumulators (@code{$ac0},
12728 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12730 Also, some built-in functions prefer or require immediate numbers as
12731 parameters, because the corresponding DSP instructions accept both immediate
12732 numbers and register operands, or accept immediate numbers only. The
12733 immediate parameters are listed as follows.
12741 imm0_255: 0 to 255.
12742 imm_n32_31: -32 to 31.
12743 imm_n512_511: -512 to 511.
12746 The following built-in functions map directly to a particular MIPS DSP
12747 instruction. Please refer to the architecture specification
12748 for details on what each instruction does.
12751 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12752 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12753 q31 __builtin_mips_addq_s_w (q31, q31)
12754 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12755 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12756 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12757 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12758 q31 __builtin_mips_subq_s_w (q31, q31)
12759 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12760 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12761 i32 __builtin_mips_addsc (i32, i32)
12762 i32 __builtin_mips_addwc (i32, i32)
12763 i32 __builtin_mips_modsub (i32, i32)
12764 i32 __builtin_mips_raddu_w_qb (v4i8)
12765 v2q15 __builtin_mips_absq_s_ph (v2q15)
12766 q31 __builtin_mips_absq_s_w (q31)
12767 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12768 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12769 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12770 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12771 q31 __builtin_mips_preceq_w_phl (v2q15)
12772 q31 __builtin_mips_preceq_w_phr (v2q15)
12773 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12774 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12775 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12776 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12777 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12778 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12779 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12780 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12781 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12782 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12783 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12784 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12785 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12786 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12787 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12788 q31 __builtin_mips_shll_s_w (q31, i32)
12789 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12790 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12791 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12792 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12793 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12794 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12795 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12796 q31 __builtin_mips_shra_r_w (q31, i32)
12797 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12798 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12799 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12800 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12801 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12802 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12803 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12804 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12805 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12806 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12807 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12808 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12809 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12810 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12811 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12812 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12813 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12814 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12815 i32 __builtin_mips_bitrev (i32)
12816 i32 __builtin_mips_insv (i32, i32)
12817 v4i8 __builtin_mips_repl_qb (imm0_255)
12818 v4i8 __builtin_mips_repl_qb (i32)
12819 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12820 v2q15 __builtin_mips_repl_ph (i32)
12821 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12822 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12823 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12824 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12825 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12826 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12827 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12828 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12829 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12830 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12831 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12832 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12833 i32 __builtin_mips_extr_w (a64, imm0_31)
12834 i32 __builtin_mips_extr_w (a64, i32)
12835 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12836 i32 __builtin_mips_extr_s_h (a64, i32)
12837 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12838 i32 __builtin_mips_extr_rs_w (a64, i32)
12839 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12840 i32 __builtin_mips_extr_r_w (a64, i32)
12841 i32 __builtin_mips_extp (a64, imm0_31)
12842 i32 __builtin_mips_extp (a64, i32)
12843 i32 __builtin_mips_extpdp (a64, imm0_31)
12844 i32 __builtin_mips_extpdp (a64, i32)
12845 a64 __builtin_mips_shilo (a64, imm_n32_31)
12846 a64 __builtin_mips_shilo (a64, i32)
12847 a64 __builtin_mips_mthlip (a64, i32)
12848 void __builtin_mips_wrdsp (i32, imm0_63)
12849 i32 __builtin_mips_rddsp (imm0_63)
12850 i32 __builtin_mips_lbux (void *, i32)
12851 i32 __builtin_mips_lhx (void *, i32)
12852 i32 __builtin_mips_lwx (void *, i32)
12853 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12854 i32 __builtin_mips_bposge32 (void)
12855 a64 __builtin_mips_madd (a64, i32, i32);
12856 a64 __builtin_mips_maddu (a64, ui32, ui32);
12857 a64 __builtin_mips_msub (a64, i32, i32);
12858 a64 __builtin_mips_msubu (a64, ui32, ui32);
12859 a64 __builtin_mips_mult (i32, i32);
12860 a64 __builtin_mips_multu (ui32, ui32);
12863 The following built-in functions map directly to a particular MIPS DSP REV 2
12864 instruction. Please refer to the architecture specification
12865 for details on what each instruction does.
12868 v4q7 __builtin_mips_absq_s_qb (v4q7);
12869 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12870 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12871 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12872 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12873 i32 __builtin_mips_append (i32, i32, imm0_31);
12874 i32 __builtin_mips_balign (i32, i32, imm0_3);
12875 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12876 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12877 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12878 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12879 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12880 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12881 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12882 q31 __builtin_mips_mulq_rs_w (q31, q31);
12883 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12884 q31 __builtin_mips_mulq_s_w (q31, q31);
12885 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12886 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12887 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12888 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12889 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12890 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12891 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12892 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12893 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12894 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12895 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12896 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12897 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12898 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12899 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12900 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12901 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12902 q31 __builtin_mips_addqh_w (q31, q31);
12903 q31 __builtin_mips_addqh_r_w (q31, q31);
12904 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12905 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12906 q31 __builtin_mips_subqh_w (q31, q31);
12907 q31 __builtin_mips_subqh_r_w (q31, q31);
12908 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12909 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12910 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12911 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12912 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12913 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12917 @node MIPS Paired-Single Support
12918 @subsection MIPS Paired-Single Support
12920 The MIPS64 architecture includes a number of instructions that
12921 operate on pairs of single-precision floating-point values.
12922 Each pair is packed into a 64-bit floating-point register,
12923 with one element being designated the ``upper half'' and
12924 the other being designated the ``lower half''.
12926 GCC supports paired-single operations using both the generic
12927 vector extensions (@pxref{Vector Extensions}) and a collection of
12928 MIPS-specific built-in functions. Both kinds of support are
12929 enabled by the @option{-mpaired-single} command-line option.
12931 The vector type associated with paired-single values is usually
12932 called @code{v2sf}. It can be defined in C as follows:
12935 typedef float v2sf __attribute__ ((vector_size (8)));
12938 @code{v2sf} values are initialized in the same way as aggregates.
12942 v2sf a = @{1.5, 9.1@};
12945 b = (v2sf) @{e, f@};
12948 @emph{Note:} The CPU's endianness determines which value is stored in
12949 the upper half of a register and which value is stored in the lower half.
12950 On little-endian targets, the first value is the lower one and the second
12951 value is the upper one. The opposite order applies to big-endian targets.
12952 For example, the code above sets the lower half of @code{a} to
12953 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
12955 @node MIPS Loongson Built-in Functions
12956 @subsection MIPS Loongson Built-in Functions
12958 GCC provides intrinsics to access the SIMD instructions provided by the
12959 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
12960 available after inclusion of the @code{loongson.h} header file,
12961 operate on the following 64-bit vector types:
12964 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
12965 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
12966 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
12967 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
12968 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
12969 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
12972 The intrinsics provided are listed below; each is named after the
12973 machine instruction to which it corresponds, with suffixes added as
12974 appropriate to distinguish intrinsics that expand to the same machine
12975 instruction yet have different argument types. Refer to the architecture
12976 documentation for a description of the functionality of each
12980 int16x4_t packsswh (int32x2_t s, int32x2_t t);
12981 int8x8_t packsshb (int16x4_t s, int16x4_t t);
12982 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
12983 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
12984 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
12985 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
12986 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
12987 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
12988 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
12989 uint64_t paddd_u (uint64_t s, uint64_t t);
12990 int64_t paddd_s (int64_t s, int64_t t);
12991 int16x4_t paddsh (int16x4_t s, int16x4_t t);
12992 int8x8_t paddsb (int8x8_t s, int8x8_t t);
12993 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
12994 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
12995 uint64_t pandn_ud (uint64_t s, uint64_t t);
12996 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
12997 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
12998 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
12999 int64_t pandn_sd (int64_t s, int64_t t);
13000 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13001 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13002 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13003 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13004 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13005 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13006 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13007 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13008 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13009 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13010 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13011 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13012 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13013 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13014 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13015 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13016 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13017 uint16x4_t pextrh_u (uint16x4_t s, int field);
13018 int16x4_t pextrh_s (int16x4_t s, int field);
13019 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13020 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13021 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13022 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13023 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13024 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13025 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13026 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13027 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13028 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13029 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13030 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13031 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13032 uint8x8_t pmovmskb_u (uint8x8_t s);
13033 int8x8_t pmovmskb_s (int8x8_t s);
13034 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13035 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13036 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13037 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13038 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13039 uint16x4_t biadd (uint8x8_t s);
13040 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13041 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13042 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13043 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13044 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13045 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13046 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13047 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13048 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13049 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13050 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13051 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13052 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13053 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13054 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13055 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13056 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13057 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13058 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13059 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13060 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13061 uint64_t psubd_u (uint64_t s, uint64_t t);
13062 int64_t psubd_s (int64_t s, int64_t t);
13063 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13064 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13065 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13066 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13067 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13068 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13069 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13070 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13071 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13072 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13073 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13074 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13075 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13076 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13077 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13078 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13082 * Paired-Single Arithmetic::
13083 * Paired-Single Built-in Functions::
13084 * MIPS-3D Built-in Functions::
13087 @node Paired-Single Arithmetic
13088 @subsubsection Paired-Single Arithmetic
13090 The table below lists the @code{v2sf} operations for which hardware
13091 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13092 values and @code{x} is an integral value.
13094 @multitable @columnfractions .50 .50
13095 @item C code @tab MIPS instruction
13096 @item @code{a + b} @tab @code{add.ps}
13097 @item @code{a - b} @tab @code{sub.ps}
13098 @item @code{-a} @tab @code{neg.ps}
13099 @item @code{a * b} @tab @code{mul.ps}
13100 @item @code{a * b + c} @tab @code{madd.ps}
13101 @item @code{a * b - c} @tab @code{msub.ps}
13102 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13103 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13104 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13107 Note that the multiply-accumulate instructions can be disabled
13108 using the command-line option @code{-mno-fused-madd}.
13110 @node Paired-Single Built-in Functions
13111 @subsubsection Paired-Single Built-in Functions
13113 The following paired-single functions map directly to a particular
13114 MIPS instruction. Please refer to the architecture specification
13115 for details on what each instruction does.
13118 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13119 Pair lower lower (@code{pll.ps}).
13121 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13122 Pair upper lower (@code{pul.ps}).
13124 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13125 Pair lower upper (@code{plu.ps}).
13127 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13128 Pair upper upper (@code{puu.ps}).
13130 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13131 Convert pair to paired single (@code{cvt.ps.s}).
13133 @item float __builtin_mips_cvt_s_pl (v2sf)
13134 Convert pair lower to single (@code{cvt.s.pl}).
13136 @item float __builtin_mips_cvt_s_pu (v2sf)
13137 Convert pair upper to single (@code{cvt.s.pu}).
13139 @item v2sf __builtin_mips_abs_ps (v2sf)
13140 Absolute value (@code{abs.ps}).
13142 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13143 Align variable (@code{alnv.ps}).
13145 @emph{Note:} The value of the third parameter must be 0 or 4
13146 modulo 8, otherwise the result is unpredictable. Please read the
13147 instruction description for details.
13150 The following multi-instruction functions are also available.
13151 In each case, @var{cond} can be any of the 16 floating-point conditions:
13152 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13153 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13154 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13157 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13158 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13159 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13160 @code{movt.ps}/@code{movf.ps}).
13162 The @code{movt} functions return the value @var{x} computed by:
13165 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13166 mov.ps @var{x},@var{c}
13167 movt.ps @var{x},@var{d},@var{cc}
13170 The @code{movf} functions are similar but use @code{movf.ps} instead
13173 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13174 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13175 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13176 @code{bc1t}/@code{bc1f}).
13178 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13179 and return either the upper or lower half of the result. For example:
13183 if (__builtin_mips_upper_c_eq_ps (a, b))
13184 upper_halves_are_equal ();
13186 upper_halves_are_unequal ();
13188 if (__builtin_mips_lower_c_eq_ps (a, b))
13189 lower_halves_are_equal ();
13191 lower_halves_are_unequal ();
13195 @node MIPS-3D Built-in Functions
13196 @subsubsection MIPS-3D Built-in Functions
13198 The MIPS-3D Application-Specific Extension (ASE) includes additional
13199 paired-single instructions that are designed to improve the performance
13200 of 3D graphics operations. Support for these instructions is controlled
13201 by the @option{-mips3d} command-line option.
13203 The functions listed below map directly to a particular MIPS-3D
13204 instruction. Please refer to the architecture specification for
13205 more details on what each instruction does.
13208 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13209 Reduction add (@code{addr.ps}).
13211 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13212 Reduction multiply (@code{mulr.ps}).
13214 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13215 Convert paired single to paired word (@code{cvt.pw.ps}).
13217 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13218 Convert paired word to paired single (@code{cvt.ps.pw}).
13220 @item float __builtin_mips_recip1_s (float)
13221 @itemx double __builtin_mips_recip1_d (double)
13222 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13223 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13225 @item float __builtin_mips_recip2_s (float, float)
13226 @itemx double __builtin_mips_recip2_d (double, double)
13227 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13228 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13230 @item float __builtin_mips_rsqrt1_s (float)
13231 @itemx double __builtin_mips_rsqrt1_d (double)
13232 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13233 Reduced-precision reciprocal square root (sequence step 1)
13234 (@code{rsqrt1.@var{fmt}}).
13236 @item float __builtin_mips_rsqrt2_s (float, float)
13237 @itemx double __builtin_mips_rsqrt2_d (double, double)
13238 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13239 Reduced-precision reciprocal square root (sequence step 2)
13240 (@code{rsqrt2.@var{fmt}}).
13243 The following multi-instruction functions are also available.
13244 In each case, @var{cond} can be any of the 16 floating-point conditions:
13245 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13246 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13247 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13250 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13251 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13252 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13253 @code{bc1t}/@code{bc1f}).
13255 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13256 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13261 if (__builtin_mips_cabs_eq_s (a, b))
13267 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13268 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13269 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13270 @code{bc1t}/@code{bc1f}).
13272 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13273 and return either the upper or lower half of the result. For example:
13277 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13278 upper_halves_are_equal ();
13280 upper_halves_are_unequal ();
13282 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13283 lower_halves_are_equal ();
13285 lower_halves_are_unequal ();
13288 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13289 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13290 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13291 @code{movt.ps}/@code{movf.ps}).
13293 The @code{movt} functions return the value @var{x} computed by:
13296 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13297 mov.ps @var{x},@var{c}
13298 movt.ps @var{x},@var{d},@var{cc}
13301 The @code{movf} functions are similar but use @code{movf.ps} instead
13304 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13305 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13306 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13307 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13308 Comparison of two paired-single values
13309 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13310 @code{bc1any2t}/@code{bc1any2f}).
13312 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13313 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13314 result is true and the @code{all} forms return true if both results are true.
13319 if (__builtin_mips_any_c_eq_ps (a, b))
13324 if (__builtin_mips_all_c_eq_ps (a, b))
13330 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13331 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13332 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13333 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13334 Comparison of four paired-single values
13335 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13336 @code{bc1any4t}/@code{bc1any4f}).
13338 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13339 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13340 The @code{any} forms return true if any of the four results are true
13341 and the @code{all} forms return true if all four results are true.
13346 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13351 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13358 @node Other MIPS Built-in Functions
13359 @subsection Other MIPS Built-in Functions
13361 GCC provides other MIPS-specific built-in functions:
13364 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13365 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13366 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13367 when this function is available.
13369 @item unsigned int __builtin_mips_get_fcsr (void)
13370 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13371 Get and set the contents of the floating-point control and status register
13372 (FPU control register 31). These functions are only available in hard-float
13373 code but can be called in both MIPS16 and non-MIPS16 contexts.
13375 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13376 register except the condition codes, which GCC assumes are preserved.
13379 @node MSP430 Built-in Functions
13380 @subsection MSP430 Built-in Functions
13382 GCC provides a couple of special builtin functions to aid in the
13383 writing of interrupt handlers in C.
13386 @item __bic_SR_register_on_exit (int @var{mask})
13387 This clears the indicated bits in the saved copy of the status register
13388 currently residing on the stack. This only works inside interrupt
13389 handlers and the changes to the status register will only take affect
13390 once the handler returns.
13392 @item __bis_SR_register_on_exit (int @var{mask})
13393 This sets the indicated bits in the saved copy of the status register
13394 currently residing on the stack. This only works inside interrupt
13395 handlers and the changes to the status register will only take affect
13396 once the handler returns.
13398 @item __delay_cycles (long long @var{cycles})
13399 This inserts an instruction sequence that takes exactly @var{cycles}
13400 cycles (between 0 and about 17E9) to complete. The inserted sequence
13401 may use jumps, loops, or no-ops, and does not interfere with any other
13402 instructions. Note that @var{cycles} must be a compile-time constant
13403 integer - that is, you must pass a number, not a variable that may be
13404 optimized to a constant later. The number of cycles delayed by this
13408 @node NDS32 Built-in Functions
13409 @subsection NDS32 Built-in Functions
13411 These built-in functions are available for the NDS32 target:
13413 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13414 Insert an ISYNC instruction into the instruction stream where
13415 @var{addr} is an instruction address for serialization.
13418 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13419 Insert an ISB instruction into the instruction stream.
13422 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13423 Return the content of a system register which is mapped by @var{sr}.
13426 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13427 Return the content of a user space register which is mapped by @var{usr}.
13430 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13431 Move the @var{value} to a system register which is mapped by @var{sr}.
13434 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13435 Move the @var{value} to a user space register which is mapped by @var{usr}.
13438 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13439 Enable global interrupt.
13442 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13443 Disable global interrupt.
13446 @node picoChip Built-in Functions
13447 @subsection picoChip Built-in Functions
13449 GCC provides an interface to selected machine instructions from the
13450 picoChip instruction set.
13453 @item int __builtin_sbc (int @var{value})
13454 Sign bit count. Return the number of consecutive bits in @var{value}
13455 that have the same value as the sign bit. The result is the number of
13456 leading sign bits minus one, giving the number of redundant sign bits in
13459 @item int __builtin_byteswap (int @var{value})
13460 Byte swap. Return the result of swapping the upper and lower bytes of
13463 @item int __builtin_brev (int @var{value})
13464 Bit reversal. Return the result of reversing the bits in
13465 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13468 @item int __builtin_adds (int @var{x}, int @var{y})
13469 Saturating addition. Return the result of adding @var{x} and @var{y},
13470 storing the value 32767 if the result overflows.
13472 @item int __builtin_subs (int @var{x}, int @var{y})
13473 Saturating subtraction. Return the result of subtracting @var{y} from
13474 @var{x}, storing the value @minus{}32768 if the result overflows.
13476 @item void __builtin_halt (void)
13477 Halt. The processor stops execution. This built-in is useful for
13478 implementing assertions.
13482 @node PowerPC Built-in Functions
13483 @subsection PowerPC Built-in Functions
13485 These built-in functions are available for the PowerPC family of
13488 float __builtin_recipdivf (float, float);
13489 float __builtin_rsqrtf (float);
13490 double __builtin_recipdiv (double, double);
13491 double __builtin_rsqrt (double);
13492 uint64_t __builtin_ppc_get_timebase ();
13493 unsigned long __builtin_ppc_mftb ();
13494 double __builtin_unpack_longdouble (long double, int);
13495 long double __builtin_pack_longdouble (double, double);
13498 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13499 @code{__builtin_rsqrtf} functions generate multiple instructions to
13500 implement the reciprocal sqrt functionality using reciprocal sqrt
13501 estimate instructions.
13503 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13504 functions generate multiple instructions to implement division using
13505 the reciprocal estimate instructions.
13507 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13508 functions generate instructions to read the Time Base Register. The
13509 @code{__builtin_ppc_get_timebase} function may generate multiple
13510 instructions and always returns the 64 bits of the Time Base Register.
13511 The @code{__builtin_ppc_mftb} function always generates one instruction and
13512 returns the Time Base Register value as an unsigned long, throwing away
13513 the most significant word on 32-bit environments.
13515 The following built-in functions are available for the PowerPC family
13516 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13517 or @option{-mpopcntd}):
13519 long __builtin_bpermd (long, long);
13520 int __builtin_divwe (int, int);
13521 int __builtin_divweo (int, int);
13522 unsigned int __builtin_divweu (unsigned int, unsigned int);
13523 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13524 long __builtin_divde (long, long);
13525 long __builtin_divdeo (long, long);
13526 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13527 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13528 unsigned int cdtbcd (unsigned int);
13529 unsigned int cbcdtd (unsigned int);
13530 unsigned int addg6s (unsigned int, unsigned int);
13533 The @code{__builtin_divde}, @code{__builtin_divdeo},
13534 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13535 64-bit environment support ISA 2.06 or later.
13537 The following built-in functions are available for the PowerPC family
13538 of processors when hardware decimal floating point
13539 (@option{-mhard-dfp}) is available:
13541 _Decimal64 __builtin_dxex (_Decimal64);
13542 _Decimal128 __builtin_dxexq (_Decimal128);
13543 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13544 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13545 _Decimal64 __builtin_denbcd (int, _Decimal64);
13546 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13547 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13548 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13549 _Decimal64 __builtin_dscli (_Decimal64, int);
13550 _Decimal128 __builtin_dscliq (_Decimal128, int);
13551 _Decimal64 __builtin_dscri (_Decimal64, int);
13552 _Decimal128 __builtin_dscriq (_Decimal128, int);
13553 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13554 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13557 The following built-in functions are available for the PowerPC family
13558 of processors when the Vector Scalar (vsx) instruction set is
13561 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13562 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13563 unsigned long long);
13566 @node PowerPC AltiVec/VSX Built-in Functions
13567 @subsection PowerPC AltiVec Built-in Functions
13569 GCC provides an interface for the PowerPC family of processors to access
13570 the AltiVec operations described in Motorola's AltiVec Programming
13571 Interface Manual. The interface is made available by including
13572 @code{<altivec.h>} and using @option{-maltivec} and
13573 @option{-mabi=altivec}. The interface supports the following vector
13577 vector unsigned char
13581 vector unsigned short
13582 vector signed short
13586 vector unsigned int
13592 If @option{-mvsx} is used the following additional vector types are
13596 vector unsigned long
13601 The long types are only implemented for 64-bit code generation, and
13602 the long type is only used in the floating point/integer conversion
13605 GCC's implementation of the high-level language interface available from
13606 C and C++ code differs from Motorola's documentation in several ways.
13611 A vector constant is a list of constant expressions within curly braces.
13614 A vector initializer requires no cast if the vector constant is of the
13615 same type as the variable it is initializing.
13618 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13619 vector type is the default signedness of the base type. The default
13620 varies depending on the operating system, so a portable program should
13621 always specify the signedness.
13624 Compiling with @option{-maltivec} adds keywords @code{__vector},
13625 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13626 @code{bool}. When compiling ISO C, the context-sensitive substitution
13627 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13628 disabled. To use them, you must include @code{<altivec.h>} instead.
13631 GCC allows using a @code{typedef} name as the type specifier for a
13635 For C, overloaded functions are implemented with macros so the following
13639 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13643 Since @code{vec_add} is a macro, the vector constant in the example
13644 is treated as four separate arguments. Wrap the entire argument in
13645 parentheses for this to work.
13648 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13649 Internally, GCC uses built-in functions to achieve the functionality in
13650 the aforementioned header file, but they are not supported and are
13651 subject to change without notice.
13653 The following interfaces are supported for the generic and specific
13654 AltiVec operations and the AltiVec predicates. In cases where there
13655 is a direct mapping between generic and specific operations, only the
13656 generic names are shown here, although the specific operations can also
13659 Arguments that are documented as @code{const int} require literal
13660 integral values within the range required for that operation.
13663 vector signed char vec_abs (vector signed char);
13664 vector signed short vec_abs (vector signed short);
13665 vector signed int vec_abs (vector signed int);
13666 vector float vec_abs (vector float);
13668 vector signed char vec_abss (vector signed char);
13669 vector signed short vec_abss (vector signed short);
13670 vector signed int vec_abss (vector signed int);
13672 vector signed char vec_add (vector bool char, vector signed char);
13673 vector signed char vec_add (vector signed char, vector bool char);
13674 vector signed char vec_add (vector signed char, vector signed char);
13675 vector unsigned char vec_add (vector bool char, vector unsigned char);
13676 vector unsigned char vec_add (vector unsigned char, vector bool char);
13677 vector unsigned char vec_add (vector unsigned char,
13678 vector unsigned char);
13679 vector signed short vec_add (vector bool short, vector signed short);
13680 vector signed short vec_add (vector signed short, vector bool short);
13681 vector signed short vec_add (vector signed short, vector signed short);
13682 vector unsigned short vec_add (vector bool short,
13683 vector unsigned short);
13684 vector unsigned short vec_add (vector unsigned short,
13685 vector bool short);
13686 vector unsigned short vec_add (vector unsigned short,
13687 vector unsigned short);
13688 vector signed int vec_add (vector bool int, vector signed int);
13689 vector signed int vec_add (vector signed int, vector bool int);
13690 vector signed int vec_add (vector signed int, vector signed int);
13691 vector unsigned int vec_add (vector bool int, vector unsigned int);
13692 vector unsigned int vec_add (vector unsigned int, vector bool int);
13693 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
13694 vector float vec_add (vector float, vector float);
13696 vector float vec_vaddfp (vector float, vector float);
13698 vector signed int vec_vadduwm (vector bool int, vector signed int);
13699 vector signed int vec_vadduwm (vector signed int, vector bool int);
13700 vector signed int vec_vadduwm (vector signed int, vector signed int);
13701 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
13702 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
13703 vector unsigned int vec_vadduwm (vector unsigned int,
13704 vector unsigned int);
13706 vector signed short vec_vadduhm (vector bool short,
13707 vector signed short);
13708 vector signed short vec_vadduhm (vector signed short,
13709 vector bool short);
13710 vector signed short vec_vadduhm (vector signed short,
13711 vector signed short);
13712 vector unsigned short vec_vadduhm (vector bool short,
13713 vector unsigned short);
13714 vector unsigned short vec_vadduhm (vector unsigned short,
13715 vector bool short);
13716 vector unsigned short vec_vadduhm (vector unsigned short,
13717 vector unsigned short);
13719 vector signed char vec_vaddubm (vector bool char, vector signed char);
13720 vector signed char vec_vaddubm (vector signed char, vector bool char);
13721 vector signed char vec_vaddubm (vector signed char, vector signed char);
13722 vector unsigned char vec_vaddubm (vector bool char,
13723 vector unsigned char);
13724 vector unsigned char vec_vaddubm (vector unsigned char,
13726 vector unsigned char vec_vaddubm (vector unsigned char,
13727 vector unsigned char);
13729 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
13731 vector unsigned char vec_adds (vector bool char, vector unsigned char);
13732 vector unsigned char vec_adds (vector unsigned char, vector bool char);
13733 vector unsigned char vec_adds (vector unsigned char,
13734 vector unsigned char);
13735 vector signed char vec_adds (vector bool char, vector signed char);
13736 vector signed char vec_adds (vector signed char, vector bool char);
13737 vector signed char vec_adds (vector signed char, vector signed char);
13738 vector unsigned short vec_adds (vector bool short,
13739 vector unsigned short);
13740 vector unsigned short vec_adds (vector unsigned short,
13741 vector bool short);
13742 vector unsigned short vec_adds (vector unsigned short,
13743 vector unsigned short);
13744 vector signed short vec_adds (vector bool short, vector signed short);
13745 vector signed short vec_adds (vector signed short, vector bool short);
13746 vector signed short vec_adds (vector signed short, vector signed short);
13747 vector unsigned int vec_adds (vector bool int, vector unsigned int);
13748 vector unsigned int vec_adds (vector unsigned int, vector bool int);
13749 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
13750 vector signed int vec_adds (vector bool int, vector signed int);
13751 vector signed int vec_adds (vector signed int, vector bool int);
13752 vector signed int vec_adds (vector signed int, vector signed int);
13754 vector signed int vec_vaddsws (vector bool int, vector signed int);
13755 vector signed int vec_vaddsws (vector signed int, vector bool int);
13756 vector signed int vec_vaddsws (vector signed int, vector signed int);
13758 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
13759 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
13760 vector unsigned int vec_vadduws (vector unsigned int,
13761 vector unsigned int);
13763 vector signed short vec_vaddshs (vector bool short,
13764 vector signed short);
13765 vector signed short vec_vaddshs (vector signed short,
13766 vector bool short);
13767 vector signed short vec_vaddshs (vector signed short,
13768 vector signed short);
13770 vector unsigned short vec_vadduhs (vector bool short,
13771 vector unsigned short);
13772 vector unsigned short vec_vadduhs (vector unsigned short,
13773 vector bool short);
13774 vector unsigned short vec_vadduhs (vector unsigned short,
13775 vector unsigned short);
13777 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13778 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13779 vector signed char vec_vaddsbs (vector signed char, vector signed char);
13781 vector unsigned char vec_vaddubs (vector bool char,
13782 vector unsigned char);
13783 vector unsigned char vec_vaddubs (vector unsigned char,
13785 vector unsigned char vec_vaddubs (vector unsigned char,
13786 vector unsigned char);
13788 vector float vec_and (vector float, vector float);
13789 vector float vec_and (vector float, vector bool int);
13790 vector float vec_and (vector bool int, vector float);
13791 vector bool int vec_and (vector bool int, vector bool int);
13792 vector signed int vec_and (vector bool int, vector signed int);
13793 vector signed int vec_and (vector signed int, vector bool int);
13794 vector signed int vec_and (vector signed int, vector signed int);
13795 vector unsigned int vec_and (vector bool int, vector unsigned int);
13796 vector unsigned int vec_and (vector unsigned int, vector bool int);
13797 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
13798 vector bool short vec_and (vector bool short, vector bool short);
13799 vector signed short vec_and (vector bool short, vector signed short);
13800 vector signed short vec_and (vector signed short, vector bool short);
13801 vector signed short vec_and (vector signed short, vector signed short);
13802 vector unsigned short vec_and (vector bool short,
13803 vector unsigned short);
13804 vector unsigned short vec_and (vector unsigned short,
13805 vector bool short);
13806 vector unsigned short vec_and (vector unsigned short,
13807 vector unsigned short);
13808 vector signed char vec_and (vector bool char, vector signed char);
13809 vector bool char vec_and (vector bool char, vector bool char);
13810 vector signed char vec_and (vector signed char, vector bool char);
13811 vector signed char vec_and (vector signed char, vector signed char);
13812 vector unsigned char vec_and (vector bool char, vector unsigned char);
13813 vector unsigned char vec_and (vector unsigned char, vector bool char);
13814 vector unsigned char vec_and (vector unsigned char,
13815 vector unsigned char);
13817 vector float vec_andc (vector float, vector float);
13818 vector float vec_andc (vector float, vector bool int);
13819 vector float vec_andc (vector bool int, vector float);
13820 vector bool int vec_andc (vector bool int, vector bool int);
13821 vector signed int vec_andc (vector bool int, vector signed int);
13822 vector signed int vec_andc (vector signed int, vector bool int);
13823 vector signed int vec_andc (vector signed int, vector signed int);
13824 vector unsigned int vec_andc (vector bool int, vector unsigned int);
13825 vector unsigned int vec_andc (vector unsigned int, vector bool int);
13826 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
13827 vector bool short vec_andc (vector bool short, vector bool short);
13828 vector signed short vec_andc (vector bool short, vector signed short);
13829 vector signed short vec_andc (vector signed short, vector bool short);
13830 vector signed short vec_andc (vector signed short, vector signed short);
13831 vector unsigned short vec_andc (vector bool short,
13832 vector unsigned short);
13833 vector unsigned short vec_andc (vector unsigned short,
13834 vector bool short);
13835 vector unsigned short vec_andc (vector unsigned short,
13836 vector unsigned short);
13837 vector signed char vec_andc (vector bool char, vector signed char);
13838 vector bool char vec_andc (vector bool char, vector bool char);
13839 vector signed char vec_andc (vector signed char, vector bool char);
13840 vector signed char vec_andc (vector signed char, vector signed char);
13841 vector unsigned char vec_andc (vector bool char, vector unsigned char);
13842 vector unsigned char vec_andc (vector unsigned char, vector bool char);
13843 vector unsigned char vec_andc (vector unsigned char,
13844 vector unsigned char);
13846 vector unsigned char vec_avg (vector unsigned char,
13847 vector unsigned char);
13848 vector signed char vec_avg (vector signed char, vector signed char);
13849 vector unsigned short vec_avg (vector unsigned short,
13850 vector unsigned short);
13851 vector signed short vec_avg (vector signed short, vector signed short);
13852 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
13853 vector signed int vec_avg (vector signed int, vector signed int);
13855 vector signed int vec_vavgsw (vector signed int, vector signed int);
13857 vector unsigned int vec_vavguw (vector unsigned int,
13858 vector unsigned int);
13860 vector signed short vec_vavgsh (vector signed short,
13861 vector signed short);
13863 vector unsigned short vec_vavguh (vector unsigned short,
13864 vector unsigned short);
13866 vector signed char vec_vavgsb (vector signed char, vector signed char);
13868 vector unsigned char vec_vavgub (vector unsigned char,
13869 vector unsigned char);
13871 vector float vec_copysign (vector float);
13873 vector float vec_ceil (vector float);
13875 vector signed int vec_cmpb (vector float, vector float);
13877 vector bool char vec_cmpeq (vector signed char, vector signed char);
13878 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
13879 vector bool short vec_cmpeq (vector signed short, vector signed short);
13880 vector bool short vec_cmpeq (vector unsigned short,
13881 vector unsigned short);
13882 vector bool int vec_cmpeq (vector signed int, vector signed int);
13883 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
13884 vector bool int vec_cmpeq (vector float, vector float);
13886 vector bool int vec_vcmpeqfp (vector float, vector float);
13888 vector bool int vec_vcmpequw (vector signed int, vector signed int);
13889 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
13891 vector bool short vec_vcmpequh (vector signed short,
13892 vector signed short);
13893 vector bool short vec_vcmpequh (vector unsigned short,
13894 vector unsigned short);
13896 vector bool char vec_vcmpequb (vector signed char, vector signed char);
13897 vector bool char vec_vcmpequb (vector unsigned char,
13898 vector unsigned char);
13900 vector bool int vec_cmpge (vector float, vector float);
13902 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
13903 vector bool char vec_cmpgt (vector signed char, vector signed char);
13904 vector bool short vec_cmpgt (vector unsigned short,
13905 vector unsigned short);
13906 vector bool short vec_cmpgt (vector signed short, vector signed short);
13907 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
13908 vector bool int vec_cmpgt (vector signed int, vector signed int);
13909 vector bool int vec_cmpgt (vector float, vector float);
13911 vector bool int vec_vcmpgtfp (vector float, vector float);
13913 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
13915 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
13917 vector bool short vec_vcmpgtsh (vector signed short,
13918 vector signed short);
13920 vector bool short vec_vcmpgtuh (vector unsigned short,
13921 vector unsigned short);
13923 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
13925 vector bool char vec_vcmpgtub (vector unsigned char,
13926 vector unsigned char);
13928 vector bool int vec_cmple (vector float, vector float);
13930 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
13931 vector bool char vec_cmplt (vector signed char, vector signed char);
13932 vector bool short vec_cmplt (vector unsigned short,
13933 vector unsigned short);
13934 vector bool short vec_cmplt (vector signed short, vector signed short);
13935 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
13936 vector bool int vec_cmplt (vector signed int, vector signed int);
13937 vector bool int vec_cmplt (vector float, vector float);
13939 vector float vec_cpsgn (vector float, vector float);
13941 vector float vec_ctf (vector unsigned int, const int);
13942 vector float vec_ctf (vector signed int, const int);
13943 vector double vec_ctf (vector unsigned long, const int);
13944 vector double vec_ctf (vector signed long, const int);
13946 vector float vec_vcfsx (vector signed int, const int);
13948 vector float vec_vcfux (vector unsigned int, const int);
13950 vector signed int vec_cts (vector float, const int);
13951 vector signed long vec_cts (vector double, const int);
13953 vector unsigned int vec_ctu (vector float, const int);
13954 vector unsigned long vec_ctu (vector double, const int);
13956 void vec_dss (const int);
13958 void vec_dssall (void);
13960 void vec_dst (const vector unsigned char *, int, const int);
13961 void vec_dst (const vector signed char *, int, const int);
13962 void vec_dst (const vector bool char *, int, const int);
13963 void vec_dst (const vector unsigned short *, int, const int);
13964 void vec_dst (const vector signed short *, int, const int);
13965 void vec_dst (const vector bool short *, int, const int);
13966 void vec_dst (const vector pixel *, int, const int);
13967 void vec_dst (const vector unsigned int *, int, const int);
13968 void vec_dst (const vector signed int *, int, const int);
13969 void vec_dst (const vector bool int *, int, const int);
13970 void vec_dst (const vector float *, int, const int);
13971 void vec_dst (const unsigned char *, int, const int);
13972 void vec_dst (const signed char *, int, const int);
13973 void vec_dst (const unsigned short *, int, const int);
13974 void vec_dst (const short *, int, const int);
13975 void vec_dst (const unsigned int *, int, const int);
13976 void vec_dst (const int *, int, const int);
13977 void vec_dst (const unsigned long *, int, const int);
13978 void vec_dst (const long *, int, const int);
13979 void vec_dst (const float *, int, const int);
13981 void vec_dstst (const vector unsigned char *, int, const int);
13982 void vec_dstst (const vector signed char *, int, const int);
13983 void vec_dstst (const vector bool char *, int, const int);
13984 void vec_dstst (const vector unsigned short *, int, const int);
13985 void vec_dstst (const vector signed short *, int, const int);
13986 void vec_dstst (const vector bool short *, int, const int);
13987 void vec_dstst (const vector pixel *, int, const int);
13988 void vec_dstst (const vector unsigned int *, int, const int);
13989 void vec_dstst (const vector signed int *, int, const int);
13990 void vec_dstst (const vector bool int *, int, const int);
13991 void vec_dstst (const vector float *, int, const int);
13992 void vec_dstst (const unsigned char *, int, const int);
13993 void vec_dstst (const signed char *, int, const int);
13994 void vec_dstst (const unsigned short *, int, const int);
13995 void vec_dstst (const short *, int, const int);
13996 void vec_dstst (const unsigned int *, int, const int);
13997 void vec_dstst (const int *, int, const int);
13998 void vec_dstst (const unsigned long *, int, const int);
13999 void vec_dstst (const long *, int, const int);
14000 void vec_dstst (const float *, int, const int);
14002 void vec_dststt (const vector unsigned char *, int, const int);
14003 void vec_dststt (const vector signed char *, int, const int);
14004 void vec_dststt (const vector bool char *, int, const int);
14005 void vec_dststt (const vector unsigned short *, int, const int);
14006 void vec_dststt (const vector signed short *, int, const int);
14007 void vec_dststt (const vector bool short *, int, const int);
14008 void vec_dststt (const vector pixel *, int, const int);
14009 void vec_dststt (const vector unsigned int *, int, const int);
14010 void vec_dststt (const vector signed int *, int, const int);
14011 void vec_dststt (const vector bool int *, int, const int);
14012 void vec_dststt (const vector float *, int, const int);
14013 void vec_dststt (const unsigned char *, int, const int);
14014 void vec_dststt (const signed char *, int, const int);
14015 void vec_dststt (const unsigned short *, int, const int);
14016 void vec_dststt (const short *, int, const int);
14017 void vec_dststt (const unsigned int *, int, const int);
14018 void vec_dststt (const int *, int, const int);
14019 void vec_dststt (const unsigned long *, int, const int);
14020 void vec_dststt (const long *, int, const int);
14021 void vec_dststt (const float *, int, const int);
14023 void vec_dstt (const vector unsigned char *, int, const int);
14024 void vec_dstt (const vector signed char *, int, const int);
14025 void vec_dstt (const vector bool char *, int, const int);
14026 void vec_dstt (const vector unsigned short *, int, const int);
14027 void vec_dstt (const vector signed short *, int, const int);
14028 void vec_dstt (const vector bool short *, int, const int);
14029 void vec_dstt (const vector pixel *, int, const int);
14030 void vec_dstt (const vector unsigned int *, int, const int);
14031 void vec_dstt (const vector signed int *, int, const int);
14032 void vec_dstt (const vector bool int *, int, const int);
14033 void vec_dstt (const vector float *, int, const int);
14034 void vec_dstt (const unsigned char *, int, const int);
14035 void vec_dstt (const signed char *, int, const int);
14036 void vec_dstt (const unsigned short *, int, const int);
14037 void vec_dstt (const short *, int, const int);
14038 void vec_dstt (const unsigned int *, int, const int);
14039 void vec_dstt (const int *, int, const int);
14040 void vec_dstt (const unsigned long *, int, const int);
14041 void vec_dstt (const long *, int, const int);
14042 void vec_dstt (const float *, int, const int);
14044 vector float vec_expte (vector float);
14046 vector float vec_floor (vector float);
14048 vector float vec_ld (int, const vector float *);
14049 vector float vec_ld (int, const float *);
14050 vector bool int vec_ld (int, const vector bool int *);
14051 vector signed int vec_ld (int, const vector signed int *);
14052 vector signed int vec_ld (int, const int *);
14053 vector signed int vec_ld (int, const long *);
14054 vector unsigned int vec_ld (int, const vector unsigned int *);
14055 vector unsigned int vec_ld (int, const unsigned int *);
14056 vector unsigned int vec_ld (int, const unsigned long *);
14057 vector bool short vec_ld (int, const vector bool short *);
14058 vector pixel vec_ld (int, const vector pixel *);
14059 vector signed short vec_ld (int, const vector signed short *);
14060 vector signed short vec_ld (int, const short *);
14061 vector unsigned short vec_ld (int, const vector unsigned short *);
14062 vector unsigned short vec_ld (int, const unsigned short *);
14063 vector bool char vec_ld (int, const vector bool char *);
14064 vector signed char vec_ld (int, const vector signed char *);
14065 vector signed char vec_ld (int, const signed char *);
14066 vector unsigned char vec_ld (int, const vector unsigned char *);
14067 vector unsigned char vec_ld (int, const unsigned char *);
14069 vector signed char vec_lde (int, const signed char *);
14070 vector unsigned char vec_lde (int, const unsigned char *);
14071 vector signed short vec_lde (int, const short *);
14072 vector unsigned short vec_lde (int, const unsigned short *);
14073 vector float vec_lde (int, const float *);
14074 vector signed int vec_lde (int, const int *);
14075 vector unsigned int vec_lde (int, const unsigned int *);
14076 vector signed int vec_lde (int, const long *);
14077 vector unsigned int vec_lde (int, const unsigned long *);
14079 vector float vec_lvewx (int, float *);
14080 vector signed int vec_lvewx (int, int *);
14081 vector unsigned int vec_lvewx (int, unsigned int *);
14082 vector signed int vec_lvewx (int, long *);
14083 vector unsigned int vec_lvewx (int, unsigned long *);
14085 vector signed short vec_lvehx (int, short *);
14086 vector unsigned short vec_lvehx (int, unsigned short *);
14088 vector signed char vec_lvebx (int, char *);
14089 vector unsigned char vec_lvebx (int, unsigned char *);
14091 vector float vec_ldl (int, const vector float *);
14092 vector float vec_ldl (int, const float *);
14093 vector bool int vec_ldl (int, const vector bool int *);
14094 vector signed int vec_ldl (int, const vector signed int *);
14095 vector signed int vec_ldl (int, const int *);
14096 vector signed int vec_ldl (int, const long *);
14097 vector unsigned int vec_ldl (int, const vector unsigned int *);
14098 vector unsigned int vec_ldl (int, const unsigned int *);
14099 vector unsigned int vec_ldl (int, const unsigned long *);
14100 vector bool short vec_ldl (int, const vector bool short *);
14101 vector pixel vec_ldl (int, const vector pixel *);
14102 vector signed short vec_ldl (int, const vector signed short *);
14103 vector signed short vec_ldl (int, const short *);
14104 vector unsigned short vec_ldl (int, const vector unsigned short *);
14105 vector unsigned short vec_ldl (int, const unsigned short *);
14106 vector bool char vec_ldl (int, const vector bool char *);
14107 vector signed char vec_ldl (int, const vector signed char *);
14108 vector signed char vec_ldl (int, const signed char *);
14109 vector unsigned char vec_ldl (int, const vector unsigned char *);
14110 vector unsigned char vec_ldl (int, const unsigned char *);
14112 vector float vec_loge (vector float);
14114 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
14115 vector unsigned char vec_lvsl (int, const volatile signed char *);
14116 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
14117 vector unsigned char vec_lvsl (int, const volatile short *);
14118 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
14119 vector unsigned char vec_lvsl (int, const volatile int *);
14120 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
14121 vector unsigned char vec_lvsl (int, const volatile long *);
14122 vector unsigned char vec_lvsl (int, const volatile float *);
14124 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
14125 vector unsigned char vec_lvsr (int, const volatile signed char *);
14126 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
14127 vector unsigned char vec_lvsr (int, const volatile short *);
14128 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
14129 vector unsigned char vec_lvsr (int, const volatile int *);
14130 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
14131 vector unsigned char vec_lvsr (int, const volatile long *);
14132 vector unsigned char vec_lvsr (int, const volatile float *);
14134 vector float vec_madd (vector float, vector float, vector float);
14136 vector signed short vec_madds (vector signed short,
14137 vector signed short,
14138 vector signed short);
14140 vector unsigned char vec_max (vector bool char, vector unsigned char);
14141 vector unsigned char vec_max (vector unsigned char, vector bool char);
14142 vector unsigned char vec_max (vector unsigned char,
14143 vector unsigned char);
14144 vector signed char vec_max (vector bool char, vector signed char);
14145 vector signed char vec_max (vector signed char, vector bool char);
14146 vector signed char vec_max (vector signed char, vector signed char);
14147 vector unsigned short vec_max (vector bool short,
14148 vector unsigned short);
14149 vector unsigned short vec_max (vector unsigned short,
14150 vector bool short);
14151 vector unsigned short vec_max (vector unsigned short,
14152 vector unsigned short);
14153 vector signed short vec_max (vector bool short, vector signed short);
14154 vector signed short vec_max (vector signed short, vector bool short);
14155 vector signed short vec_max (vector signed short, vector signed short);
14156 vector unsigned int vec_max (vector bool int, vector unsigned int);
14157 vector unsigned int vec_max (vector unsigned int, vector bool int);
14158 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
14159 vector signed int vec_max (vector bool int, vector signed int);
14160 vector signed int vec_max (vector signed int, vector bool int);
14161 vector signed int vec_max (vector signed int, vector signed int);
14162 vector float vec_max (vector float, vector float);
14164 vector float vec_vmaxfp (vector float, vector float);
14166 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14167 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14168 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14170 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14171 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14172 vector unsigned int vec_vmaxuw (vector unsigned int,
14173 vector unsigned int);
14175 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14176 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14177 vector signed short vec_vmaxsh (vector signed short,
14178 vector signed short);
14180 vector unsigned short vec_vmaxuh (vector bool short,
14181 vector unsigned short);
14182 vector unsigned short vec_vmaxuh (vector unsigned short,
14183 vector bool short);
14184 vector unsigned short vec_vmaxuh (vector unsigned short,
14185 vector unsigned short);
14187 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14188 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14189 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14191 vector unsigned char vec_vmaxub (vector bool char,
14192 vector unsigned char);
14193 vector unsigned char vec_vmaxub (vector unsigned char,
14195 vector unsigned char vec_vmaxub (vector unsigned char,
14196 vector unsigned char);
14198 vector bool char vec_mergeh (vector bool char, vector bool char);
14199 vector signed char vec_mergeh (vector signed char, vector signed char);
14200 vector unsigned char vec_mergeh (vector unsigned char,
14201 vector unsigned char);
14202 vector bool short vec_mergeh (vector bool short, vector bool short);
14203 vector pixel vec_mergeh (vector pixel, vector pixel);
14204 vector signed short vec_mergeh (vector signed short,
14205 vector signed short);
14206 vector unsigned short vec_mergeh (vector unsigned short,
14207 vector unsigned short);
14208 vector float vec_mergeh (vector float, vector float);
14209 vector bool int vec_mergeh (vector bool int, vector bool int);
14210 vector signed int vec_mergeh (vector signed int, vector signed int);
14211 vector unsigned int vec_mergeh (vector unsigned int,
14212 vector unsigned int);
14214 vector float vec_vmrghw (vector float, vector float);
14215 vector bool int vec_vmrghw (vector bool int, vector bool int);
14216 vector signed int vec_vmrghw (vector signed int, vector signed int);
14217 vector unsigned int vec_vmrghw (vector unsigned int,
14218 vector unsigned int);
14220 vector bool short vec_vmrghh (vector bool short, vector bool short);
14221 vector signed short vec_vmrghh (vector signed short,
14222 vector signed short);
14223 vector unsigned short vec_vmrghh (vector unsigned short,
14224 vector unsigned short);
14225 vector pixel vec_vmrghh (vector pixel, vector pixel);
14227 vector bool char vec_vmrghb (vector bool char, vector bool char);
14228 vector signed char vec_vmrghb (vector signed char, vector signed char);
14229 vector unsigned char vec_vmrghb (vector unsigned char,
14230 vector unsigned char);
14232 vector bool char vec_mergel (vector bool char, vector bool char);
14233 vector signed char vec_mergel (vector signed char, vector signed char);
14234 vector unsigned char vec_mergel (vector unsigned char,
14235 vector unsigned char);
14236 vector bool short vec_mergel (vector bool short, vector bool short);
14237 vector pixel vec_mergel (vector pixel, vector pixel);
14238 vector signed short vec_mergel (vector signed short,
14239 vector signed short);
14240 vector unsigned short vec_mergel (vector unsigned short,
14241 vector unsigned short);
14242 vector float vec_mergel (vector float, vector float);
14243 vector bool int vec_mergel (vector bool int, vector bool int);
14244 vector signed int vec_mergel (vector signed int, vector signed int);
14245 vector unsigned int vec_mergel (vector unsigned int,
14246 vector unsigned int);
14248 vector float vec_vmrglw (vector float, vector float);
14249 vector signed int vec_vmrglw (vector signed int, vector signed int);
14250 vector unsigned int vec_vmrglw (vector unsigned int,
14251 vector unsigned int);
14252 vector bool int vec_vmrglw (vector bool int, vector bool int);
14254 vector bool short vec_vmrglh (vector bool short, vector bool short);
14255 vector signed short vec_vmrglh (vector signed short,
14256 vector signed short);
14257 vector unsigned short vec_vmrglh (vector unsigned short,
14258 vector unsigned short);
14259 vector pixel vec_vmrglh (vector pixel, vector pixel);
14261 vector bool char vec_vmrglb (vector bool char, vector bool char);
14262 vector signed char vec_vmrglb (vector signed char, vector signed char);
14263 vector unsigned char vec_vmrglb (vector unsigned char,
14264 vector unsigned char);
14266 vector unsigned short vec_mfvscr (void);
14268 vector unsigned char vec_min (vector bool char, vector unsigned char);
14269 vector unsigned char vec_min (vector unsigned char, vector bool char);
14270 vector unsigned char vec_min (vector unsigned char,
14271 vector unsigned char);
14272 vector signed char vec_min (vector bool char, vector signed char);
14273 vector signed char vec_min (vector signed char, vector bool char);
14274 vector signed char vec_min (vector signed char, vector signed char);
14275 vector unsigned short vec_min (vector bool short,
14276 vector unsigned short);
14277 vector unsigned short vec_min (vector unsigned short,
14278 vector bool short);
14279 vector unsigned short vec_min (vector unsigned short,
14280 vector unsigned short);
14281 vector signed short vec_min (vector bool short, vector signed short);
14282 vector signed short vec_min (vector signed short, vector bool short);
14283 vector signed short vec_min (vector signed short, vector signed short);
14284 vector unsigned int vec_min (vector bool int, vector unsigned int);
14285 vector unsigned int vec_min (vector unsigned int, vector bool int);
14286 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14287 vector signed int vec_min (vector bool int, vector signed int);
14288 vector signed int vec_min (vector signed int, vector bool int);
14289 vector signed int vec_min (vector signed int, vector signed int);
14290 vector float vec_min (vector float, vector float);
14292 vector float vec_vminfp (vector float, vector float);
14294 vector signed int vec_vminsw (vector bool int, vector signed int);
14295 vector signed int vec_vminsw (vector signed int, vector bool int);
14296 vector signed int vec_vminsw (vector signed int, vector signed int);
14298 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14299 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14300 vector unsigned int vec_vminuw (vector unsigned int,
14301 vector unsigned int);
14303 vector signed short vec_vminsh (vector bool short, vector signed short);
14304 vector signed short vec_vminsh (vector signed short, vector bool short);
14305 vector signed short vec_vminsh (vector signed short,
14306 vector signed short);
14308 vector unsigned short vec_vminuh (vector bool short,
14309 vector unsigned short);
14310 vector unsigned short vec_vminuh (vector unsigned short,
14311 vector bool short);
14312 vector unsigned short vec_vminuh (vector unsigned short,
14313 vector unsigned short);
14315 vector signed char vec_vminsb (vector bool char, vector signed char);
14316 vector signed char vec_vminsb (vector signed char, vector bool char);
14317 vector signed char vec_vminsb (vector signed char, vector signed char);
14319 vector unsigned char vec_vminub (vector bool char,
14320 vector unsigned char);
14321 vector unsigned char vec_vminub (vector unsigned char,
14323 vector unsigned char vec_vminub (vector unsigned char,
14324 vector unsigned char);
14326 vector signed short vec_mladd (vector signed short,
14327 vector signed short,
14328 vector signed short);
14329 vector signed short vec_mladd (vector signed short,
14330 vector unsigned short,
14331 vector unsigned short);
14332 vector signed short vec_mladd (vector unsigned short,
14333 vector signed short,
14334 vector signed short);
14335 vector unsigned short vec_mladd (vector unsigned short,
14336 vector unsigned short,
14337 vector unsigned short);
14339 vector signed short vec_mradds (vector signed short,
14340 vector signed short,
14341 vector signed short);
14343 vector unsigned int vec_msum (vector unsigned char,
14344 vector unsigned char,
14345 vector unsigned int);
14346 vector signed int vec_msum (vector signed char,
14347 vector unsigned char,
14348 vector signed int);
14349 vector unsigned int vec_msum (vector unsigned short,
14350 vector unsigned short,
14351 vector unsigned int);
14352 vector signed int vec_msum (vector signed short,
14353 vector signed short,
14354 vector signed int);
14356 vector signed int vec_vmsumshm (vector signed short,
14357 vector signed short,
14358 vector signed int);
14360 vector unsigned int vec_vmsumuhm (vector unsigned short,
14361 vector unsigned short,
14362 vector unsigned int);
14364 vector signed int vec_vmsummbm (vector signed char,
14365 vector unsigned char,
14366 vector signed int);
14368 vector unsigned int vec_vmsumubm (vector unsigned char,
14369 vector unsigned char,
14370 vector unsigned int);
14372 vector unsigned int vec_msums (vector unsigned short,
14373 vector unsigned short,
14374 vector unsigned int);
14375 vector signed int vec_msums (vector signed short,
14376 vector signed short,
14377 vector signed int);
14379 vector signed int vec_vmsumshs (vector signed short,
14380 vector signed short,
14381 vector signed int);
14383 vector unsigned int vec_vmsumuhs (vector unsigned short,
14384 vector unsigned short,
14385 vector unsigned int);
14387 void vec_mtvscr (vector signed int);
14388 void vec_mtvscr (vector unsigned int);
14389 void vec_mtvscr (vector bool int);
14390 void vec_mtvscr (vector signed short);
14391 void vec_mtvscr (vector unsigned short);
14392 void vec_mtvscr (vector bool short);
14393 void vec_mtvscr (vector pixel);
14394 void vec_mtvscr (vector signed char);
14395 void vec_mtvscr (vector unsigned char);
14396 void vec_mtvscr (vector bool char);
14398 vector unsigned short vec_mule (vector unsigned char,
14399 vector unsigned char);
14400 vector signed short vec_mule (vector signed char,
14401 vector signed char);
14402 vector unsigned int vec_mule (vector unsigned short,
14403 vector unsigned short);
14404 vector signed int vec_mule (vector signed short, vector signed short);
14406 vector signed int vec_vmulesh (vector signed short,
14407 vector signed short);
14409 vector unsigned int vec_vmuleuh (vector unsigned short,
14410 vector unsigned short);
14412 vector signed short vec_vmulesb (vector signed char,
14413 vector signed char);
14415 vector unsigned short vec_vmuleub (vector unsigned char,
14416 vector unsigned char);
14418 vector unsigned short vec_mulo (vector unsigned char,
14419 vector unsigned char);
14420 vector signed short vec_mulo (vector signed char, vector signed char);
14421 vector unsigned int vec_mulo (vector unsigned short,
14422 vector unsigned short);
14423 vector signed int vec_mulo (vector signed short, vector signed short);
14425 vector signed int vec_vmulosh (vector signed short,
14426 vector signed short);
14428 vector unsigned int vec_vmulouh (vector unsigned short,
14429 vector unsigned short);
14431 vector signed short vec_vmulosb (vector signed char,
14432 vector signed char);
14434 vector unsigned short vec_vmuloub (vector unsigned char,
14435 vector unsigned char);
14437 vector float vec_nmsub (vector float, vector float, vector float);
14439 vector float vec_nor (vector float, vector float);
14440 vector signed int vec_nor (vector signed int, vector signed int);
14441 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14442 vector bool int vec_nor (vector bool int, vector bool int);
14443 vector signed short vec_nor (vector signed short, vector signed short);
14444 vector unsigned short vec_nor (vector unsigned short,
14445 vector unsigned short);
14446 vector bool short vec_nor (vector bool short, vector bool short);
14447 vector signed char vec_nor (vector signed char, vector signed char);
14448 vector unsigned char vec_nor (vector unsigned char,
14449 vector unsigned char);
14450 vector bool char vec_nor (vector bool char, vector bool char);
14452 vector float vec_or (vector float, vector float);
14453 vector float vec_or (vector float, vector bool int);
14454 vector float vec_or (vector bool int, vector float);
14455 vector bool int vec_or (vector bool int, vector bool int);
14456 vector signed int vec_or (vector bool int, vector signed int);
14457 vector signed int vec_or (vector signed int, vector bool int);
14458 vector signed int vec_or (vector signed int, vector signed int);
14459 vector unsigned int vec_or (vector bool int, vector unsigned int);
14460 vector unsigned int vec_or (vector unsigned int, vector bool int);
14461 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14462 vector bool short vec_or (vector bool short, vector bool short);
14463 vector signed short vec_or (vector bool short, vector signed short);
14464 vector signed short vec_or (vector signed short, vector bool short);
14465 vector signed short vec_or (vector signed short, vector signed short);
14466 vector unsigned short vec_or (vector bool short, vector unsigned short);
14467 vector unsigned short vec_or (vector unsigned short, vector bool short);
14468 vector unsigned short vec_or (vector unsigned short,
14469 vector unsigned short);
14470 vector signed char vec_or (vector bool char, vector signed char);
14471 vector bool char vec_or (vector bool char, vector bool char);
14472 vector signed char vec_or (vector signed char, vector bool char);
14473 vector signed char vec_or (vector signed char, vector signed char);
14474 vector unsigned char vec_or (vector bool char, vector unsigned char);
14475 vector unsigned char vec_or (vector unsigned char, vector bool char);
14476 vector unsigned char vec_or (vector unsigned char,
14477 vector unsigned char);
14479 vector signed char vec_pack (vector signed short, vector signed short);
14480 vector unsigned char vec_pack (vector unsigned short,
14481 vector unsigned short);
14482 vector bool char vec_pack (vector bool short, vector bool short);
14483 vector signed short vec_pack (vector signed int, vector signed int);
14484 vector unsigned short vec_pack (vector unsigned int,
14485 vector unsigned int);
14486 vector bool short vec_pack (vector bool int, vector bool int);
14488 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14489 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14490 vector unsigned short vec_vpkuwum (vector unsigned int,
14491 vector unsigned int);
14493 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14494 vector signed char vec_vpkuhum (vector signed short,
14495 vector signed short);
14496 vector unsigned char vec_vpkuhum (vector unsigned short,
14497 vector unsigned short);
14499 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14501 vector unsigned char vec_packs (vector unsigned short,
14502 vector unsigned short);
14503 vector signed char vec_packs (vector signed short, vector signed short);
14504 vector unsigned short vec_packs (vector unsigned int,
14505 vector unsigned int);
14506 vector signed short vec_packs (vector signed int, vector signed int);
14508 vector signed short vec_vpkswss (vector signed int, vector signed int);
14510 vector unsigned short vec_vpkuwus (vector unsigned int,
14511 vector unsigned int);
14513 vector signed char vec_vpkshss (vector signed short,
14514 vector signed short);
14516 vector unsigned char vec_vpkuhus (vector unsigned short,
14517 vector unsigned short);
14519 vector unsigned char vec_packsu (vector unsigned short,
14520 vector unsigned short);
14521 vector unsigned char vec_packsu (vector signed short,
14522 vector signed short);
14523 vector unsigned short vec_packsu (vector unsigned int,
14524 vector unsigned int);
14525 vector unsigned short vec_packsu (vector signed int, vector signed int);
14527 vector unsigned short vec_vpkswus (vector signed int,
14528 vector signed int);
14530 vector unsigned char vec_vpkshus (vector signed short,
14531 vector signed short);
14533 vector float vec_perm (vector float,
14535 vector unsigned char);
14536 vector signed int vec_perm (vector signed int,
14538 vector unsigned char);
14539 vector unsigned int vec_perm (vector unsigned int,
14540 vector unsigned int,
14541 vector unsigned char);
14542 vector bool int vec_perm (vector bool int,
14544 vector unsigned char);
14545 vector signed short vec_perm (vector signed short,
14546 vector signed short,
14547 vector unsigned char);
14548 vector unsigned short vec_perm (vector unsigned short,
14549 vector unsigned short,
14550 vector unsigned char);
14551 vector bool short vec_perm (vector bool short,
14553 vector unsigned char);
14554 vector pixel vec_perm (vector pixel,
14556 vector unsigned char);
14557 vector signed char vec_perm (vector signed char,
14558 vector signed char,
14559 vector unsigned char);
14560 vector unsigned char vec_perm (vector unsigned char,
14561 vector unsigned char,
14562 vector unsigned char);
14563 vector bool char vec_perm (vector bool char,
14565 vector unsigned char);
14567 vector float vec_re (vector float);
14569 vector signed char vec_rl (vector signed char,
14570 vector unsigned char);
14571 vector unsigned char vec_rl (vector unsigned char,
14572 vector unsigned char);
14573 vector signed short vec_rl (vector signed short, vector unsigned short);
14574 vector unsigned short vec_rl (vector unsigned short,
14575 vector unsigned short);
14576 vector signed int vec_rl (vector signed int, vector unsigned int);
14577 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14579 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14580 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14582 vector signed short vec_vrlh (vector signed short,
14583 vector unsigned short);
14584 vector unsigned short vec_vrlh (vector unsigned short,
14585 vector unsigned short);
14587 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14588 vector unsigned char vec_vrlb (vector unsigned char,
14589 vector unsigned char);
14591 vector float vec_round (vector float);
14593 vector float vec_recip (vector float, vector float);
14595 vector float vec_rsqrt (vector float);
14597 vector float vec_rsqrte (vector float);
14599 vector float vec_sel (vector float, vector float, vector bool int);
14600 vector float vec_sel (vector float, vector float, vector unsigned int);
14601 vector signed int vec_sel (vector signed int,
14604 vector signed int vec_sel (vector signed int,
14606 vector unsigned int);
14607 vector unsigned int vec_sel (vector unsigned int,
14608 vector unsigned int,
14610 vector unsigned int vec_sel (vector unsigned int,
14611 vector unsigned int,
14612 vector unsigned int);
14613 vector bool int vec_sel (vector bool int,
14616 vector bool int vec_sel (vector bool int,
14618 vector unsigned int);
14619 vector signed short vec_sel (vector signed short,
14620 vector signed short,
14621 vector bool short);
14622 vector signed short vec_sel (vector signed short,
14623 vector signed short,
14624 vector unsigned short);
14625 vector unsigned short vec_sel (vector unsigned short,
14626 vector unsigned short,
14627 vector bool short);
14628 vector unsigned short vec_sel (vector unsigned short,
14629 vector unsigned short,
14630 vector unsigned short);
14631 vector bool short vec_sel (vector bool short,
14633 vector bool short);
14634 vector bool short vec_sel (vector bool short,
14636 vector unsigned short);
14637 vector signed char vec_sel (vector signed char,
14638 vector signed char,
14640 vector signed char vec_sel (vector signed char,
14641 vector signed char,
14642 vector unsigned char);
14643 vector unsigned char vec_sel (vector unsigned char,
14644 vector unsigned char,
14646 vector unsigned char vec_sel (vector unsigned char,
14647 vector unsigned char,
14648 vector unsigned char);
14649 vector bool char vec_sel (vector bool char,
14652 vector bool char vec_sel (vector bool char,
14654 vector unsigned char);
14656 vector signed char vec_sl (vector signed char,
14657 vector unsigned char);
14658 vector unsigned char vec_sl (vector unsigned char,
14659 vector unsigned char);
14660 vector signed short vec_sl (vector signed short, vector unsigned short);
14661 vector unsigned short vec_sl (vector unsigned short,
14662 vector unsigned short);
14663 vector signed int vec_sl (vector signed int, vector unsigned int);
14664 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14666 vector signed int vec_vslw (vector signed int, vector unsigned int);
14667 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14669 vector signed short vec_vslh (vector signed short,
14670 vector unsigned short);
14671 vector unsigned short vec_vslh (vector unsigned short,
14672 vector unsigned short);
14674 vector signed char vec_vslb (vector signed char, vector unsigned char);
14675 vector unsigned char vec_vslb (vector unsigned char,
14676 vector unsigned char);
14678 vector float vec_sld (vector float, vector float, const int);
14679 vector signed int vec_sld (vector signed int,
14682 vector unsigned int vec_sld (vector unsigned int,
14683 vector unsigned int,
14685 vector bool int vec_sld (vector bool int,
14688 vector signed short vec_sld (vector signed short,
14689 vector signed short,
14691 vector unsigned short vec_sld (vector unsigned short,
14692 vector unsigned short,
14694 vector bool short vec_sld (vector bool short,
14697 vector pixel vec_sld (vector pixel,
14700 vector signed char vec_sld (vector signed char,
14701 vector signed char,
14703 vector unsigned char vec_sld (vector unsigned char,
14704 vector unsigned char,
14706 vector bool char vec_sld (vector bool char,
14710 vector signed int vec_sll (vector signed int,
14711 vector unsigned int);
14712 vector signed int vec_sll (vector signed int,
14713 vector unsigned short);
14714 vector signed int vec_sll (vector signed int,
14715 vector unsigned char);
14716 vector unsigned int vec_sll (vector unsigned int,
14717 vector unsigned int);
14718 vector unsigned int vec_sll (vector unsigned int,
14719 vector unsigned short);
14720 vector unsigned int vec_sll (vector unsigned int,
14721 vector unsigned char);
14722 vector bool int vec_sll (vector bool int,
14723 vector unsigned int);
14724 vector bool int vec_sll (vector bool int,
14725 vector unsigned short);
14726 vector bool int vec_sll (vector bool int,
14727 vector unsigned char);
14728 vector signed short vec_sll (vector signed short,
14729 vector unsigned int);
14730 vector signed short vec_sll (vector signed short,
14731 vector unsigned short);
14732 vector signed short vec_sll (vector signed short,
14733 vector unsigned char);
14734 vector unsigned short vec_sll (vector unsigned short,
14735 vector unsigned int);
14736 vector unsigned short vec_sll (vector unsigned short,
14737 vector unsigned short);
14738 vector unsigned short vec_sll (vector unsigned short,
14739 vector unsigned char);
14740 vector bool short vec_sll (vector bool short, vector unsigned int);
14741 vector bool short vec_sll (vector bool short, vector unsigned short);
14742 vector bool short vec_sll (vector bool short, vector unsigned char);
14743 vector pixel vec_sll (vector pixel, vector unsigned int);
14744 vector pixel vec_sll (vector pixel, vector unsigned short);
14745 vector pixel vec_sll (vector pixel, vector unsigned char);
14746 vector signed char vec_sll (vector signed char, vector unsigned int);
14747 vector signed char vec_sll (vector signed char, vector unsigned short);
14748 vector signed char vec_sll (vector signed char, vector unsigned char);
14749 vector unsigned char vec_sll (vector unsigned char,
14750 vector unsigned int);
14751 vector unsigned char vec_sll (vector unsigned char,
14752 vector unsigned short);
14753 vector unsigned char vec_sll (vector unsigned char,
14754 vector unsigned char);
14755 vector bool char vec_sll (vector bool char, vector unsigned int);
14756 vector bool char vec_sll (vector bool char, vector unsigned short);
14757 vector bool char vec_sll (vector bool char, vector unsigned char);
14759 vector float vec_slo (vector float, vector signed char);
14760 vector float vec_slo (vector float, vector unsigned char);
14761 vector signed int vec_slo (vector signed int, vector signed char);
14762 vector signed int vec_slo (vector signed int, vector unsigned char);
14763 vector unsigned int vec_slo (vector unsigned int, vector signed char);
14764 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
14765 vector signed short vec_slo (vector signed short, vector signed char);
14766 vector signed short vec_slo (vector signed short, vector unsigned char);
14767 vector unsigned short vec_slo (vector unsigned short,
14768 vector signed char);
14769 vector unsigned short vec_slo (vector unsigned short,
14770 vector unsigned char);
14771 vector pixel vec_slo (vector pixel, vector signed char);
14772 vector pixel vec_slo (vector pixel, vector unsigned char);
14773 vector signed char vec_slo (vector signed char, vector signed char);
14774 vector signed char vec_slo (vector signed char, vector unsigned char);
14775 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14776 vector unsigned char vec_slo (vector unsigned char,
14777 vector unsigned char);
14779 vector signed char vec_splat (vector signed char, const int);
14780 vector unsigned char vec_splat (vector unsigned char, const int);
14781 vector bool char vec_splat (vector bool char, const int);
14782 vector signed short vec_splat (vector signed short, const int);
14783 vector unsigned short vec_splat (vector unsigned short, const int);
14784 vector bool short vec_splat (vector bool short, const int);
14785 vector pixel vec_splat (vector pixel, const int);
14786 vector float vec_splat (vector float, const int);
14787 vector signed int vec_splat (vector signed int, const int);
14788 vector unsigned int vec_splat (vector unsigned int, const int);
14789 vector bool int vec_splat (vector bool int, const int);
14790 vector signed long vec_splat (vector signed long, const int);
14791 vector unsigned long vec_splat (vector unsigned long, const int);
14793 vector signed char vec_splats (signed char);
14794 vector unsigned char vec_splats (unsigned char);
14795 vector signed short vec_splats (signed short);
14796 vector unsigned short vec_splats (unsigned short);
14797 vector signed int vec_splats (signed int);
14798 vector unsigned int vec_splats (unsigned int);
14799 vector float vec_splats (float);
14801 vector float vec_vspltw (vector float, const int);
14802 vector signed int vec_vspltw (vector signed int, const int);
14803 vector unsigned int vec_vspltw (vector unsigned int, const int);
14804 vector bool int vec_vspltw (vector bool int, const int);
14806 vector bool short vec_vsplth (vector bool short, const int);
14807 vector signed short vec_vsplth (vector signed short, const int);
14808 vector unsigned short vec_vsplth (vector unsigned short, const int);
14809 vector pixel vec_vsplth (vector pixel, const int);
14811 vector signed char vec_vspltb (vector signed char, const int);
14812 vector unsigned char vec_vspltb (vector unsigned char, const int);
14813 vector bool char vec_vspltb (vector bool char, const int);
14815 vector signed char vec_splat_s8 (const int);
14817 vector signed short vec_splat_s16 (const int);
14819 vector signed int vec_splat_s32 (const int);
14821 vector unsigned char vec_splat_u8 (const int);
14823 vector unsigned short vec_splat_u16 (const int);
14825 vector unsigned int vec_splat_u32 (const int);
14827 vector signed char vec_sr (vector signed char, vector unsigned char);
14828 vector unsigned char vec_sr (vector unsigned char,
14829 vector unsigned char);
14830 vector signed short vec_sr (vector signed short,
14831 vector unsigned short);
14832 vector unsigned short vec_sr (vector unsigned short,
14833 vector unsigned short);
14834 vector signed int vec_sr (vector signed int, vector unsigned int);
14835 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
14837 vector signed int vec_vsrw (vector signed int, vector unsigned int);
14838 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
14840 vector signed short vec_vsrh (vector signed short,
14841 vector unsigned short);
14842 vector unsigned short vec_vsrh (vector unsigned short,
14843 vector unsigned short);
14845 vector signed char vec_vsrb (vector signed char, vector unsigned char);
14846 vector unsigned char vec_vsrb (vector unsigned char,
14847 vector unsigned char);
14849 vector signed char vec_sra (vector signed char, vector unsigned char);
14850 vector unsigned char vec_sra (vector unsigned char,
14851 vector unsigned char);
14852 vector signed short vec_sra (vector signed short,
14853 vector unsigned short);
14854 vector unsigned short vec_sra (vector unsigned short,
14855 vector unsigned short);
14856 vector signed int vec_sra (vector signed int, vector unsigned int);
14857 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
14859 vector signed int vec_vsraw (vector signed int, vector unsigned int);
14860 vector unsigned int vec_vsraw (vector unsigned int,
14861 vector unsigned int);
14863 vector signed short vec_vsrah (vector signed short,
14864 vector unsigned short);
14865 vector unsigned short vec_vsrah (vector unsigned short,
14866 vector unsigned short);
14868 vector signed char vec_vsrab (vector signed char, vector unsigned char);
14869 vector unsigned char vec_vsrab (vector unsigned char,
14870 vector unsigned char);
14872 vector signed int vec_srl (vector signed int, vector unsigned int);
14873 vector signed int vec_srl (vector signed int, vector unsigned short);
14874 vector signed int vec_srl (vector signed int, vector unsigned char);
14875 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
14876 vector unsigned int vec_srl (vector unsigned int,
14877 vector unsigned short);
14878 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
14879 vector bool int vec_srl (vector bool int, vector unsigned int);
14880 vector bool int vec_srl (vector bool int, vector unsigned short);
14881 vector bool int vec_srl (vector bool int, vector unsigned char);
14882 vector signed short vec_srl (vector signed short, vector unsigned int);
14883 vector signed short vec_srl (vector signed short,
14884 vector unsigned short);
14885 vector signed short vec_srl (vector signed short, vector unsigned char);
14886 vector unsigned short vec_srl (vector unsigned short,
14887 vector unsigned int);
14888 vector unsigned short vec_srl (vector unsigned short,
14889 vector unsigned short);
14890 vector unsigned short vec_srl (vector unsigned short,
14891 vector unsigned char);
14892 vector bool short vec_srl (vector bool short, vector unsigned int);
14893 vector bool short vec_srl (vector bool short, vector unsigned short);
14894 vector bool short vec_srl (vector bool short, vector unsigned char);
14895 vector pixel vec_srl (vector pixel, vector unsigned int);
14896 vector pixel vec_srl (vector pixel, vector unsigned short);
14897 vector pixel vec_srl (vector pixel, vector unsigned char);
14898 vector signed char vec_srl (vector signed char, vector unsigned int);
14899 vector signed char vec_srl (vector signed char, vector unsigned short);
14900 vector signed char vec_srl (vector signed char, vector unsigned char);
14901 vector unsigned char vec_srl (vector unsigned char,
14902 vector unsigned int);
14903 vector unsigned char vec_srl (vector unsigned char,
14904 vector unsigned short);
14905 vector unsigned char vec_srl (vector unsigned char,
14906 vector unsigned char);
14907 vector bool char vec_srl (vector bool char, vector unsigned int);
14908 vector bool char vec_srl (vector bool char, vector unsigned short);
14909 vector bool char vec_srl (vector bool char, vector unsigned char);
14911 vector float vec_sro (vector float, vector signed char);
14912 vector float vec_sro (vector float, vector unsigned char);
14913 vector signed int vec_sro (vector signed int, vector signed char);
14914 vector signed int vec_sro (vector signed int, vector unsigned char);
14915 vector unsigned int vec_sro (vector unsigned int, vector signed char);
14916 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
14917 vector signed short vec_sro (vector signed short, vector signed char);
14918 vector signed short vec_sro (vector signed short, vector unsigned char);
14919 vector unsigned short vec_sro (vector unsigned short,
14920 vector signed char);
14921 vector unsigned short vec_sro (vector unsigned short,
14922 vector unsigned char);
14923 vector pixel vec_sro (vector pixel, vector signed char);
14924 vector pixel vec_sro (vector pixel, vector unsigned char);
14925 vector signed char vec_sro (vector signed char, vector signed char);
14926 vector signed char vec_sro (vector signed char, vector unsigned char);
14927 vector unsigned char vec_sro (vector unsigned char, vector signed char);
14928 vector unsigned char vec_sro (vector unsigned char,
14929 vector unsigned char);
14931 void vec_st (vector float, int, vector float *);
14932 void vec_st (vector float, int, float *);
14933 void vec_st (vector signed int, int, vector signed int *);
14934 void vec_st (vector signed int, int, int *);
14935 void vec_st (vector unsigned int, int, vector unsigned int *);
14936 void vec_st (vector unsigned int, int, unsigned int *);
14937 void vec_st (vector bool int, int, vector bool int *);
14938 void vec_st (vector bool int, int, unsigned int *);
14939 void vec_st (vector bool int, int, int *);
14940 void vec_st (vector signed short, int, vector signed short *);
14941 void vec_st (vector signed short, int, short *);
14942 void vec_st (vector unsigned short, int, vector unsigned short *);
14943 void vec_st (vector unsigned short, int, unsigned short *);
14944 void vec_st (vector bool short, int, vector bool short *);
14945 void vec_st (vector bool short, int, unsigned short *);
14946 void vec_st (vector pixel, int, vector pixel *);
14947 void vec_st (vector pixel, int, unsigned short *);
14948 void vec_st (vector pixel, int, short *);
14949 void vec_st (vector bool short, int, short *);
14950 void vec_st (vector signed char, int, vector signed char *);
14951 void vec_st (vector signed char, int, signed char *);
14952 void vec_st (vector unsigned char, int, vector unsigned char *);
14953 void vec_st (vector unsigned char, int, unsigned char *);
14954 void vec_st (vector bool char, int, vector bool char *);
14955 void vec_st (vector bool char, int, unsigned char *);
14956 void vec_st (vector bool char, int, signed char *);
14958 void vec_ste (vector signed char, int, signed char *);
14959 void vec_ste (vector unsigned char, int, unsigned char *);
14960 void vec_ste (vector bool char, int, signed char *);
14961 void vec_ste (vector bool char, int, unsigned char *);
14962 void vec_ste (vector signed short, int, short *);
14963 void vec_ste (vector unsigned short, int, unsigned short *);
14964 void vec_ste (vector bool short, int, short *);
14965 void vec_ste (vector bool short, int, unsigned short *);
14966 void vec_ste (vector pixel, int, short *);
14967 void vec_ste (vector pixel, int, unsigned short *);
14968 void vec_ste (vector float, int, float *);
14969 void vec_ste (vector signed int, int, int *);
14970 void vec_ste (vector unsigned int, int, unsigned int *);
14971 void vec_ste (vector bool int, int, int *);
14972 void vec_ste (vector bool int, int, unsigned int *);
14974 void vec_stvewx (vector float, int, float *);
14975 void vec_stvewx (vector signed int, int, int *);
14976 void vec_stvewx (vector unsigned int, int, unsigned int *);
14977 void vec_stvewx (vector bool int, int, int *);
14978 void vec_stvewx (vector bool int, int, unsigned int *);
14980 void vec_stvehx (vector signed short, int, short *);
14981 void vec_stvehx (vector unsigned short, int, unsigned short *);
14982 void vec_stvehx (vector bool short, int, short *);
14983 void vec_stvehx (vector bool short, int, unsigned short *);
14984 void vec_stvehx (vector pixel, int, short *);
14985 void vec_stvehx (vector pixel, int, unsigned short *);
14987 void vec_stvebx (vector signed char, int, signed char *);
14988 void vec_stvebx (vector unsigned char, int, unsigned char *);
14989 void vec_stvebx (vector bool char, int, signed char *);
14990 void vec_stvebx (vector bool char, int, unsigned char *);
14992 void vec_stl (vector float, int, vector float *);
14993 void vec_stl (vector float, int, float *);
14994 void vec_stl (vector signed int, int, vector signed int *);
14995 void vec_stl (vector signed int, int, int *);
14996 void vec_stl (vector unsigned int, int, vector unsigned int *);
14997 void vec_stl (vector unsigned int, int, unsigned int *);
14998 void vec_stl (vector bool int, int, vector bool int *);
14999 void vec_stl (vector bool int, int, unsigned int *);
15000 void vec_stl (vector bool int, int, int *);
15001 void vec_stl (vector signed short, int, vector signed short *);
15002 void vec_stl (vector signed short, int, short *);
15003 void vec_stl (vector unsigned short, int, vector unsigned short *);
15004 void vec_stl (vector unsigned short, int, unsigned short *);
15005 void vec_stl (vector bool short, int, vector bool short *);
15006 void vec_stl (vector bool short, int, unsigned short *);
15007 void vec_stl (vector bool short, int, short *);
15008 void vec_stl (vector pixel, int, vector pixel *);
15009 void vec_stl (vector pixel, int, unsigned short *);
15010 void vec_stl (vector pixel, int, short *);
15011 void vec_stl (vector signed char, int, vector signed char *);
15012 void vec_stl (vector signed char, int, signed char *);
15013 void vec_stl (vector unsigned char, int, vector unsigned char *);
15014 void vec_stl (vector unsigned char, int, unsigned char *);
15015 void vec_stl (vector bool char, int, vector bool char *);
15016 void vec_stl (vector bool char, int, unsigned char *);
15017 void vec_stl (vector bool char, int, signed char *);
15019 vector signed char vec_sub (vector bool char, vector signed char);
15020 vector signed char vec_sub (vector signed char, vector bool char);
15021 vector signed char vec_sub (vector signed char, vector signed char);
15022 vector unsigned char vec_sub (vector bool char, vector unsigned char);
15023 vector unsigned char vec_sub (vector unsigned char, vector bool char);
15024 vector unsigned char vec_sub (vector unsigned char,
15025 vector unsigned char);
15026 vector signed short vec_sub (vector bool short, vector signed short);
15027 vector signed short vec_sub (vector signed short, vector bool short);
15028 vector signed short vec_sub (vector signed short, vector signed short);
15029 vector unsigned short vec_sub (vector bool short,
15030 vector unsigned short);
15031 vector unsigned short vec_sub (vector unsigned short,
15032 vector bool short);
15033 vector unsigned short vec_sub (vector unsigned short,
15034 vector unsigned short);
15035 vector signed int vec_sub (vector bool int, vector signed int);
15036 vector signed int vec_sub (vector signed int, vector bool int);
15037 vector signed int vec_sub (vector signed int, vector signed int);
15038 vector unsigned int vec_sub (vector bool int, vector unsigned int);
15039 vector unsigned int vec_sub (vector unsigned int, vector bool int);
15040 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
15041 vector float vec_sub (vector float, vector float);
15043 vector float vec_vsubfp (vector float, vector float);
15045 vector signed int vec_vsubuwm (vector bool int, vector signed int);
15046 vector signed int vec_vsubuwm (vector signed int, vector bool int);
15047 vector signed int vec_vsubuwm (vector signed int, vector signed int);
15048 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
15049 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
15050 vector unsigned int vec_vsubuwm (vector unsigned int,
15051 vector unsigned int);
15053 vector signed short vec_vsubuhm (vector bool short,
15054 vector signed short);
15055 vector signed short vec_vsubuhm (vector signed short,
15056 vector bool short);
15057 vector signed short vec_vsubuhm (vector signed short,
15058 vector signed short);
15059 vector unsigned short vec_vsubuhm (vector bool short,
15060 vector unsigned short);
15061 vector unsigned short vec_vsubuhm (vector unsigned short,
15062 vector bool short);
15063 vector unsigned short vec_vsubuhm (vector unsigned short,
15064 vector unsigned short);
15066 vector signed char vec_vsububm (vector bool char, vector signed char);
15067 vector signed char vec_vsububm (vector signed char, vector bool char);
15068 vector signed char vec_vsububm (vector signed char, vector signed char);
15069 vector unsigned char vec_vsububm (vector bool char,
15070 vector unsigned char);
15071 vector unsigned char vec_vsububm (vector unsigned char,
15073 vector unsigned char vec_vsububm (vector unsigned char,
15074 vector unsigned char);
15076 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
15078 vector unsigned char vec_subs (vector bool char, vector unsigned char);
15079 vector unsigned char vec_subs (vector unsigned char, vector bool char);
15080 vector unsigned char vec_subs (vector unsigned char,
15081 vector unsigned char);
15082 vector signed char vec_subs (vector bool char, vector signed char);
15083 vector signed char vec_subs (vector signed char, vector bool char);
15084 vector signed char vec_subs (vector signed char, vector signed char);
15085 vector unsigned short vec_subs (vector bool short,
15086 vector unsigned short);
15087 vector unsigned short vec_subs (vector unsigned short,
15088 vector bool short);
15089 vector unsigned short vec_subs (vector unsigned short,
15090 vector unsigned short);
15091 vector signed short vec_subs (vector bool short, vector signed short);
15092 vector signed short vec_subs (vector signed short, vector bool short);
15093 vector signed short vec_subs (vector signed short, vector signed short);
15094 vector unsigned int vec_subs (vector bool int, vector unsigned int);
15095 vector unsigned int vec_subs (vector unsigned int, vector bool int);
15096 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
15097 vector signed int vec_subs (vector bool int, vector signed int);
15098 vector signed int vec_subs (vector signed int, vector bool int);
15099 vector signed int vec_subs (vector signed int, vector signed int);
15101 vector signed int vec_vsubsws (vector bool int, vector signed int);
15102 vector signed int vec_vsubsws (vector signed int, vector bool int);
15103 vector signed int vec_vsubsws (vector signed int, vector signed int);
15105 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
15106 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
15107 vector unsigned int vec_vsubuws (vector unsigned int,
15108 vector unsigned int);
15110 vector signed short vec_vsubshs (vector bool short,
15111 vector signed short);
15112 vector signed short vec_vsubshs (vector signed short,
15113 vector bool short);
15114 vector signed short vec_vsubshs (vector signed short,
15115 vector signed short);
15117 vector unsigned short vec_vsubuhs (vector bool short,
15118 vector unsigned short);
15119 vector unsigned short vec_vsubuhs (vector unsigned short,
15120 vector bool short);
15121 vector unsigned short vec_vsubuhs (vector unsigned short,
15122 vector unsigned short);
15124 vector signed char vec_vsubsbs (vector bool char, vector signed char);
15125 vector signed char vec_vsubsbs (vector signed char, vector bool char);
15126 vector signed char vec_vsubsbs (vector signed char, vector signed char);
15128 vector unsigned char vec_vsububs (vector bool char,
15129 vector unsigned char);
15130 vector unsigned char vec_vsububs (vector unsigned char,
15132 vector unsigned char vec_vsububs (vector unsigned char,
15133 vector unsigned char);
15135 vector unsigned int vec_sum4s (vector unsigned char,
15136 vector unsigned int);
15137 vector signed int vec_sum4s (vector signed char, vector signed int);
15138 vector signed int vec_sum4s (vector signed short, vector signed int);
15140 vector signed int vec_vsum4shs (vector signed short, vector signed int);
15142 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
15144 vector unsigned int vec_vsum4ubs (vector unsigned char,
15145 vector unsigned int);
15147 vector signed int vec_sum2s (vector signed int, vector signed int);
15149 vector signed int vec_sums (vector signed int, vector signed int);
15151 vector float vec_trunc (vector float);
15153 vector signed short vec_unpackh (vector signed char);
15154 vector bool short vec_unpackh (vector bool char);
15155 vector signed int vec_unpackh (vector signed short);
15156 vector bool int vec_unpackh (vector bool short);
15157 vector unsigned int vec_unpackh (vector pixel);
15159 vector bool int vec_vupkhsh (vector bool short);
15160 vector signed int vec_vupkhsh (vector signed short);
15162 vector unsigned int vec_vupkhpx (vector pixel);
15164 vector bool short vec_vupkhsb (vector bool char);
15165 vector signed short vec_vupkhsb (vector signed char);
15167 vector signed short vec_unpackl (vector signed char);
15168 vector bool short vec_unpackl (vector bool char);
15169 vector unsigned int vec_unpackl (vector pixel);
15170 vector signed int vec_unpackl (vector signed short);
15171 vector bool int vec_unpackl (vector bool short);
15173 vector unsigned int vec_vupklpx (vector pixel);
15175 vector bool int vec_vupklsh (vector bool short);
15176 vector signed int vec_vupklsh (vector signed short);
15178 vector bool short vec_vupklsb (vector bool char);
15179 vector signed short vec_vupklsb (vector signed char);
15181 vector float vec_xor (vector float, vector float);
15182 vector float vec_xor (vector float, vector bool int);
15183 vector float vec_xor (vector bool int, vector float);
15184 vector bool int vec_xor (vector bool int, vector bool int);
15185 vector signed int vec_xor (vector bool int, vector signed int);
15186 vector signed int vec_xor (vector signed int, vector bool int);
15187 vector signed int vec_xor (vector signed int, vector signed int);
15188 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15189 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15190 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15191 vector bool short vec_xor (vector bool short, vector bool short);
15192 vector signed short vec_xor (vector bool short, vector signed short);
15193 vector signed short vec_xor (vector signed short, vector bool short);
15194 vector signed short vec_xor (vector signed short, vector signed short);
15195 vector unsigned short vec_xor (vector bool short,
15196 vector unsigned short);
15197 vector unsigned short vec_xor (vector unsigned short,
15198 vector bool short);
15199 vector unsigned short vec_xor (vector unsigned short,
15200 vector unsigned short);
15201 vector signed char vec_xor (vector bool char, vector signed char);
15202 vector bool char vec_xor (vector bool char, vector bool char);
15203 vector signed char vec_xor (vector signed char, vector bool char);
15204 vector signed char vec_xor (vector signed char, vector signed char);
15205 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15206 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15207 vector unsigned char vec_xor (vector unsigned char,
15208 vector unsigned char);
15210 int vec_all_eq (vector signed char, vector bool char);
15211 int vec_all_eq (vector signed char, vector signed char);
15212 int vec_all_eq (vector unsigned char, vector bool char);
15213 int vec_all_eq (vector unsigned char, vector unsigned char);
15214 int vec_all_eq (vector bool char, vector bool char);
15215 int vec_all_eq (vector bool char, vector unsigned char);
15216 int vec_all_eq (vector bool char, vector signed char);
15217 int vec_all_eq (vector signed short, vector bool short);
15218 int vec_all_eq (vector signed short, vector signed short);
15219 int vec_all_eq (vector unsigned short, vector bool short);
15220 int vec_all_eq (vector unsigned short, vector unsigned short);
15221 int vec_all_eq (vector bool short, vector bool short);
15222 int vec_all_eq (vector bool short, vector unsigned short);
15223 int vec_all_eq (vector bool short, vector signed short);
15224 int vec_all_eq (vector pixel, vector pixel);
15225 int vec_all_eq (vector signed int, vector bool int);
15226 int vec_all_eq (vector signed int, vector signed int);
15227 int vec_all_eq (vector unsigned int, vector bool int);
15228 int vec_all_eq (vector unsigned int, vector unsigned int);
15229 int vec_all_eq (vector bool int, vector bool int);
15230 int vec_all_eq (vector bool int, vector unsigned int);
15231 int vec_all_eq (vector bool int, vector signed int);
15232 int vec_all_eq (vector float, vector float);
15234 int vec_all_ge (vector bool char, vector unsigned char);
15235 int vec_all_ge (vector unsigned char, vector bool char);
15236 int vec_all_ge (vector unsigned char, vector unsigned char);
15237 int vec_all_ge (vector bool char, vector signed char);
15238 int vec_all_ge (vector signed char, vector bool char);
15239 int vec_all_ge (vector signed char, vector signed char);
15240 int vec_all_ge (vector bool short, vector unsigned short);
15241 int vec_all_ge (vector unsigned short, vector bool short);
15242 int vec_all_ge (vector unsigned short, vector unsigned short);
15243 int vec_all_ge (vector signed short, vector signed short);
15244 int vec_all_ge (vector bool short, vector signed short);
15245 int vec_all_ge (vector signed short, vector bool short);
15246 int vec_all_ge (vector bool int, vector unsigned int);
15247 int vec_all_ge (vector unsigned int, vector bool int);
15248 int vec_all_ge (vector unsigned int, vector unsigned int);
15249 int vec_all_ge (vector bool int, vector signed int);
15250 int vec_all_ge (vector signed int, vector bool int);
15251 int vec_all_ge (vector signed int, vector signed int);
15252 int vec_all_ge (vector float, vector float);
15254 int vec_all_gt (vector bool char, vector unsigned char);
15255 int vec_all_gt (vector unsigned char, vector bool char);
15256 int vec_all_gt (vector unsigned char, vector unsigned char);
15257 int vec_all_gt (vector bool char, vector signed char);
15258 int vec_all_gt (vector signed char, vector bool char);
15259 int vec_all_gt (vector signed char, vector signed char);
15260 int vec_all_gt (vector bool short, vector unsigned short);
15261 int vec_all_gt (vector unsigned short, vector bool short);
15262 int vec_all_gt (vector unsigned short, vector unsigned short);
15263 int vec_all_gt (vector bool short, vector signed short);
15264 int vec_all_gt (vector signed short, vector bool short);
15265 int vec_all_gt (vector signed short, vector signed short);
15266 int vec_all_gt (vector bool int, vector unsigned int);
15267 int vec_all_gt (vector unsigned int, vector bool int);
15268 int vec_all_gt (vector unsigned int, vector unsigned int);
15269 int vec_all_gt (vector bool int, vector signed int);
15270 int vec_all_gt (vector signed int, vector bool int);
15271 int vec_all_gt (vector signed int, vector signed int);
15272 int vec_all_gt (vector float, vector float);
15274 int vec_all_in (vector float, vector float);
15276 int vec_all_le (vector bool char, vector unsigned char);
15277 int vec_all_le (vector unsigned char, vector bool char);
15278 int vec_all_le (vector unsigned char, vector unsigned char);
15279 int vec_all_le (vector bool char, vector signed char);
15280 int vec_all_le (vector signed char, vector bool char);
15281 int vec_all_le (vector signed char, vector signed char);
15282 int vec_all_le (vector bool short, vector unsigned short);
15283 int vec_all_le (vector unsigned short, vector bool short);
15284 int vec_all_le (vector unsigned short, vector unsigned short);
15285 int vec_all_le (vector bool short, vector signed short);
15286 int vec_all_le (vector signed short, vector bool short);
15287 int vec_all_le (vector signed short, vector signed short);
15288 int vec_all_le (vector bool int, vector unsigned int);
15289 int vec_all_le (vector unsigned int, vector bool int);
15290 int vec_all_le (vector unsigned int, vector unsigned int);
15291 int vec_all_le (vector bool int, vector signed int);
15292 int vec_all_le (vector signed int, vector bool int);
15293 int vec_all_le (vector signed int, vector signed int);
15294 int vec_all_le (vector float, vector float);
15296 int vec_all_lt (vector bool char, vector unsigned char);
15297 int vec_all_lt (vector unsigned char, vector bool char);
15298 int vec_all_lt (vector unsigned char, vector unsigned char);
15299 int vec_all_lt (vector bool char, vector signed char);
15300 int vec_all_lt (vector signed char, vector bool char);
15301 int vec_all_lt (vector signed char, vector signed char);
15302 int vec_all_lt (vector bool short, vector unsigned short);
15303 int vec_all_lt (vector unsigned short, vector bool short);
15304 int vec_all_lt (vector unsigned short, vector unsigned short);
15305 int vec_all_lt (vector bool short, vector signed short);
15306 int vec_all_lt (vector signed short, vector bool short);
15307 int vec_all_lt (vector signed short, vector signed short);
15308 int vec_all_lt (vector bool int, vector unsigned int);
15309 int vec_all_lt (vector unsigned int, vector bool int);
15310 int vec_all_lt (vector unsigned int, vector unsigned int);
15311 int vec_all_lt (vector bool int, vector signed int);
15312 int vec_all_lt (vector signed int, vector bool int);
15313 int vec_all_lt (vector signed int, vector signed int);
15314 int vec_all_lt (vector float, vector float);
15316 int vec_all_nan (vector float);
15318 int vec_all_ne (vector signed char, vector bool char);
15319 int vec_all_ne (vector signed char, vector signed char);
15320 int vec_all_ne (vector unsigned char, vector bool char);
15321 int vec_all_ne (vector unsigned char, vector unsigned char);
15322 int vec_all_ne (vector bool char, vector bool char);
15323 int vec_all_ne (vector bool char, vector unsigned char);
15324 int vec_all_ne (vector bool char, vector signed char);
15325 int vec_all_ne (vector signed short, vector bool short);
15326 int vec_all_ne (vector signed short, vector signed short);
15327 int vec_all_ne (vector unsigned short, vector bool short);
15328 int vec_all_ne (vector unsigned short, vector unsigned short);
15329 int vec_all_ne (vector bool short, vector bool short);
15330 int vec_all_ne (vector bool short, vector unsigned short);
15331 int vec_all_ne (vector bool short, vector signed short);
15332 int vec_all_ne (vector pixel, vector pixel);
15333 int vec_all_ne (vector signed int, vector bool int);
15334 int vec_all_ne (vector signed int, vector signed int);
15335 int vec_all_ne (vector unsigned int, vector bool int);
15336 int vec_all_ne (vector unsigned int, vector unsigned int);
15337 int vec_all_ne (vector bool int, vector bool int);
15338 int vec_all_ne (vector bool int, vector unsigned int);
15339 int vec_all_ne (vector bool int, vector signed int);
15340 int vec_all_ne (vector float, vector float);
15342 int vec_all_nge (vector float, vector float);
15344 int vec_all_ngt (vector float, vector float);
15346 int vec_all_nle (vector float, vector float);
15348 int vec_all_nlt (vector float, vector float);
15350 int vec_all_numeric (vector float);
15352 int vec_any_eq (vector signed char, vector bool char);
15353 int vec_any_eq (vector signed char, vector signed char);
15354 int vec_any_eq (vector unsigned char, vector bool char);
15355 int vec_any_eq (vector unsigned char, vector unsigned char);
15356 int vec_any_eq (vector bool char, vector bool char);
15357 int vec_any_eq (vector bool char, vector unsigned char);
15358 int vec_any_eq (vector bool char, vector signed char);
15359 int vec_any_eq (vector signed short, vector bool short);
15360 int vec_any_eq (vector signed short, vector signed short);
15361 int vec_any_eq (vector unsigned short, vector bool short);
15362 int vec_any_eq (vector unsigned short, vector unsigned short);
15363 int vec_any_eq (vector bool short, vector bool short);
15364 int vec_any_eq (vector bool short, vector unsigned short);
15365 int vec_any_eq (vector bool short, vector signed short);
15366 int vec_any_eq (vector pixel, vector pixel);
15367 int vec_any_eq (vector signed int, vector bool int);
15368 int vec_any_eq (vector signed int, vector signed int);
15369 int vec_any_eq (vector unsigned int, vector bool int);
15370 int vec_any_eq (vector unsigned int, vector unsigned int);
15371 int vec_any_eq (vector bool int, vector bool int);
15372 int vec_any_eq (vector bool int, vector unsigned int);
15373 int vec_any_eq (vector bool int, vector signed int);
15374 int vec_any_eq (vector float, vector float);
15376 int vec_any_ge (vector signed char, vector bool char);
15377 int vec_any_ge (vector unsigned char, vector bool char);
15378 int vec_any_ge (vector unsigned char, vector unsigned char);
15379 int vec_any_ge (vector signed char, vector signed char);
15380 int vec_any_ge (vector bool char, vector unsigned char);
15381 int vec_any_ge (vector bool char, vector signed char);
15382 int vec_any_ge (vector unsigned short, vector bool short);
15383 int vec_any_ge (vector unsigned short, vector unsigned short);
15384 int vec_any_ge (vector signed short, vector signed short);
15385 int vec_any_ge (vector signed short, vector bool short);
15386 int vec_any_ge (vector bool short, vector unsigned short);
15387 int vec_any_ge (vector bool short, vector signed short);
15388 int vec_any_ge (vector signed int, vector bool int);
15389 int vec_any_ge (vector unsigned int, vector bool int);
15390 int vec_any_ge (vector unsigned int, vector unsigned int);
15391 int vec_any_ge (vector signed int, vector signed int);
15392 int vec_any_ge (vector bool int, vector unsigned int);
15393 int vec_any_ge (vector bool int, vector signed int);
15394 int vec_any_ge (vector float, vector float);
15396 int vec_any_gt (vector bool char, vector unsigned char);
15397 int vec_any_gt (vector unsigned char, vector bool char);
15398 int vec_any_gt (vector unsigned char, vector unsigned char);
15399 int vec_any_gt (vector bool char, vector signed char);
15400 int vec_any_gt (vector signed char, vector bool char);
15401 int vec_any_gt (vector signed char, vector signed char);
15402 int vec_any_gt (vector bool short, vector unsigned short);
15403 int vec_any_gt (vector unsigned short, vector bool short);
15404 int vec_any_gt (vector unsigned short, vector unsigned short);
15405 int vec_any_gt (vector bool short, vector signed short);
15406 int vec_any_gt (vector signed short, vector bool short);
15407 int vec_any_gt (vector signed short, vector signed short);
15408 int vec_any_gt (vector bool int, vector unsigned int);
15409 int vec_any_gt (vector unsigned int, vector bool int);
15410 int vec_any_gt (vector unsigned int, vector unsigned int);
15411 int vec_any_gt (vector bool int, vector signed int);
15412 int vec_any_gt (vector signed int, vector bool int);
15413 int vec_any_gt (vector signed int, vector signed int);
15414 int vec_any_gt (vector float, vector float);
15416 int vec_any_le (vector bool char, vector unsigned char);
15417 int vec_any_le (vector unsigned char, vector bool char);
15418 int vec_any_le (vector unsigned char, vector unsigned char);
15419 int vec_any_le (vector bool char, vector signed char);
15420 int vec_any_le (vector signed char, vector bool char);
15421 int vec_any_le (vector signed char, vector signed char);
15422 int vec_any_le (vector bool short, vector unsigned short);
15423 int vec_any_le (vector unsigned short, vector bool short);
15424 int vec_any_le (vector unsigned short, vector unsigned short);
15425 int vec_any_le (vector bool short, vector signed short);
15426 int vec_any_le (vector signed short, vector bool short);
15427 int vec_any_le (vector signed short, vector signed short);
15428 int vec_any_le (vector bool int, vector unsigned int);
15429 int vec_any_le (vector unsigned int, vector bool int);
15430 int vec_any_le (vector unsigned int, vector unsigned int);
15431 int vec_any_le (vector bool int, vector signed int);
15432 int vec_any_le (vector signed int, vector bool int);
15433 int vec_any_le (vector signed int, vector signed int);
15434 int vec_any_le (vector float, vector float);
15436 int vec_any_lt (vector bool char, vector unsigned char);
15437 int vec_any_lt (vector unsigned char, vector bool char);
15438 int vec_any_lt (vector unsigned char, vector unsigned char);
15439 int vec_any_lt (vector bool char, vector signed char);
15440 int vec_any_lt (vector signed char, vector bool char);
15441 int vec_any_lt (vector signed char, vector signed char);
15442 int vec_any_lt (vector bool short, vector unsigned short);
15443 int vec_any_lt (vector unsigned short, vector bool short);
15444 int vec_any_lt (vector unsigned short, vector unsigned short);
15445 int vec_any_lt (vector bool short, vector signed short);
15446 int vec_any_lt (vector signed short, vector bool short);
15447 int vec_any_lt (vector signed short, vector signed short);
15448 int vec_any_lt (vector bool int, vector unsigned int);
15449 int vec_any_lt (vector unsigned int, vector bool int);
15450 int vec_any_lt (vector unsigned int, vector unsigned int);
15451 int vec_any_lt (vector bool int, vector signed int);
15452 int vec_any_lt (vector signed int, vector bool int);
15453 int vec_any_lt (vector signed int, vector signed int);
15454 int vec_any_lt (vector float, vector float);
15456 int vec_any_nan (vector float);
15458 int vec_any_ne (vector signed char, vector bool char);
15459 int vec_any_ne (vector signed char, vector signed char);
15460 int vec_any_ne (vector unsigned char, vector bool char);
15461 int vec_any_ne (vector unsigned char, vector unsigned char);
15462 int vec_any_ne (vector bool char, vector bool char);
15463 int vec_any_ne (vector bool char, vector unsigned char);
15464 int vec_any_ne (vector bool char, vector signed char);
15465 int vec_any_ne (vector signed short, vector bool short);
15466 int vec_any_ne (vector signed short, vector signed short);
15467 int vec_any_ne (vector unsigned short, vector bool short);
15468 int vec_any_ne (vector unsigned short, vector unsigned short);
15469 int vec_any_ne (vector bool short, vector bool short);
15470 int vec_any_ne (vector bool short, vector unsigned short);
15471 int vec_any_ne (vector bool short, vector signed short);
15472 int vec_any_ne (vector pixel, vector pixel);
15473 int vec_any_ne (vector signed int, vector bool int);
15474 int vec_any_ne (vector signed int, vector signed int);
15475 int vec_any_ne (vector unsigned int, vector bool int);
15476 int vec_any_ne (vector unsigned int, vector unsigned int);
15477 int vec_any_ne (vector bool int, vector bool int);
15478 int vec_any_ne (vector bool int, vector unsigned int);
15479 int vec_any_ne (vector bool int, vector signed int);
15480 int vec_any_ne (vector float, vector float);
15482 int vec_any_nge (vector float, vector float);
15484 int vec_any_ngt (vector float, vector float);
15486 int vec_any_nle (vector float, vector float);
15488 int vec_any_nlt (vector float, vector float);
15490 int vec_any_numeric (vector float);
15492 int vec_any_out (vector float, vector float);
15495 If the vector/scalar (VSX) instruction set is available, the following
15496 additional functions are available:
15499 vector double vec_abs (vector double);
15500 vector double vec_add (vector double, vector double);
15501 vector double vec_and (vector double, vector double);
15502 vector double vec_and (vector double, vector bool long);
15503 vector double vec_and (vector bool long, vector double);
15504 vector long vec_and (vector long, vector long);
15505 vector long vec_and (vector long, vector bool long);
15506 vector long vec_and (vector bool long, vector long);
15507 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15508 vector unsigned long vec_and (vector unsigned long, vector bool long);
15509 vector unsigned long vec_and (vector bool long, vector unsigned long);
15510 vector double vec_andc (vector double, vector double);
15511 vector double vec_andc (vector double, vector bool long);
15512 vector double vec_andc (vector bool long, vector double);
15513 vector long vec_andc (vector long, vector long);
15514 vector long vec_andc (vector long, vector bool long);
15515 vector long vec_andc (vector bool long, vector long);
15516 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15517 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15518 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15519 vector double vec_ceil (vector double);
15520 vector bool long vec_cmpeq (vector double, vector double);
15521 vector bool long vec_cmpge (vector double, vector double);
15522 vector bool long vec_cmpgt (vector double, vector double);
15523 vector bool long vec_cmple (vector double, vector double);
15524 vector bool long vec_cmplt (vector double, vector double);
15525 vector double vec_cpsgn (vector double, vector double);
15526 vector float vec_div (vector float, vector float);
15527 vector double vec_div (vector double, vector double);
15528 vector long vec_div (vector long, vector long);
15529 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15530 vector double vec_floor (vector double);
15531 vector double vec_ld (int, const vector double *);
15532 vector double vec_ld (int, const double *);
15533 vector double vec_ldl (int, const vector double *);
15534 vector double vec_ldl (int, const double *);
15535 vector unsigned char vec_lvsl (int, const volatile double *);
15536 vector unsigned char vec_lvsr (int, const volatile double *);
15537 vector double vec_madd (vector double, vector double, vector double);
15538 vector double vec_max (vector double, vector double);
15539 vector signed long vec_mergeh (vector signed long, vector signed long);
15540 vector signed long vec_mergeh (vector signed long, vector bool long);
15541 vector signed long vec_mergeh (vector bool long, vector signed long);
15542 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15543 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15544 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15545 vector signed long vec_mergel (vector signed long, vector signed long);
15546 vector signed long vec_mergel (vector signed long, vector bool long);
15547 vector signed long vec_mergel (vector bool long, vector signed long);
15548 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15549 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15550 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15551 vector double vec_min (vector double, vector double);
15552 vector float vec_msub (vector float, vector float, vector float);
15553 vector double vec_msub (vector double, vector double, vector double);
15554 vector float vec_mul (vector float, vector float);
15555 vector double vec_mul (vector double, vector double);
15556 vector long vec_mul (vector long, vector long);
15557 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15558 vector float vec_nearbyint (vector float);
15559 vector double vec_nearbyint (vector double);
15560 vector float vec_nmadd (vector float, vector float, vector float);
15561 vector double vec_nmadd (vector double, vector double, vector double);
15562 vector double vec_nmsub (vector double, vector double, vector double);
15563 vector double vec_nor (vector double, vector double);
15564 vector long vec_nor (vector long, vector long);
15565 vector long vec_nor (vector long, vector bool long);
15566 vector long vec_nor (vector bool long, vector long);
15567 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15568 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15569 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15570 vector double vec_or (vector double, vector double);
15571 vector double vec_or (vector double, vector bool long);
15572 vector double vec_or (vector bool long, vector double);
15573 vector long vec_or (vector long, vector long);
15574 vector long vec_or (vector long, vector bool long);
15575 vector long vec_or (vector bool long, vector long);
15576 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15577 vector unsigned long vec_or (vector unsigned long, vector bool long);
15578 vector unsigned long vec_or (vector bool long, vector unsigned long);
15579 vector double vec_perm (vector double, vector double, vector unsigned char);
15580 vector long vec_perm (vector long, vector long, vector unsigned char);
15581 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15582 vector unsigned char);
15583 vector double vec_rint (vector double);
15584 vector double vec_recip (vector double, vector double);
15585 vector double vec_rsqrt (vector double);
15586 vector double vec_rsqrte (vector double);
15587 vector double vec_sel (vector double, vector double, vector bool long);
15588 vector double vec_sel (vector double, vector double, vector unsigned long);
15589 vector long vec_sel (vector long, vector long, vector long);
15590 vector long vec_sel (vector long, vector long, vector unsigned long);
15591 vector long vec_sel (vector long, vector long, vector bool long);
15592 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15594 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15595 vector unsigned long);
15596 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15598 vector double vec_splats (double);
15599 vector signed long vec_splats (signed long);
15600 vector unsigned long vec_splats (unsigned long);
15601 vector float vec_sqrt (vector float);
15602 vector double vec_sqrt (vector double);
15603 void vec_st (vector double, int, vector double *);
15604 void vec_st (vector double, int, double *);
15605 vector double vec_sub (vector double, vector double);
15606 vector double vec_trunc (vector double);
15607 vector double vec_xor (vector double, vector double);
15608 vector double vec_xor (vector double, vector bool long);
15609 vector double vec_xor (vector bool long, vector double);
15610 vector long vec_xor (vector long, vector long);
15611 vector long vec_xor (vector long, vector bool long);
15612 vector long vec_xor (vector bool long, vector long);
15613 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15614 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15615 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15616 int vec_all_eq (vector double, vector double);
15617 int vec_all_ge (vector double, vector double);
15618 int vec_all_gt (vector double, vector double);
15619 int vec_all_le (vector double, vector double);
15620 int vec_all_lt (vector double, vector double);
15621 int vec_all_nan (vector double);
15622 int vec_all_ne (vector double, vector double);
15623 int vec_all_nge (vector double, vector double);
15624 int vec_all_ngt (vector double, vector double);
15625 int vec_all_nle (vector double, vector double);
15626 int vec_all_nlt (vector double, vector double);
15627 int vec_all_numeric (vector double);
15628 int vec_any_eq (vector double, vector double);
15629 int vec_any_ge (vector double, vector double);
15630 int vec_any_gt (vector double, vector double);
15631 int vec_any_le (vector double, vector double);
15632 int vec_any_lt (vector double, vector double);
15633 int vec_any_nan (vector double);
15634 int vec_any_ne (vector double, vector double);
15635 int vec_any_nge (vector double, vector double);
15636 int vec_any_ngt (vector double, vector double);
15637 int vec_any_nle (vector double, vector double);
15638 int vec_any_nlt (vector double, vector double);
15639 int vec_any_numeric (vector double);
15641 vector double vec_vsx_ld (int, const vector double *);
15642 vector double vec_vsx_ld (int, const double *);
15643 vector float vec_vsx_ld (int, const vector float *);
15644 vector float vec_vsx_ld (int, const float *);
15645 vector bool int vec_vsx_ld (int, const vector bool int *);
15646 vector signed int vec_vsx_ld (int, const vector signed int *);
15647 vector signed int vec_vsx_ld (int, const int *);
15648 vector signed int vec_vsx_ld (int, const long *);
15649 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15650 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15651 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15652 vector bool short vec_vsx_ld (int, const vector bool short *);
15653 vector pixel vec_vsx_ld (int, const vector pixel *);
15654 vector signed short vec_vsx_ld (int, const vector signed short *);
15655 vector signed short vec_vsx_ld (int, const short *);
15656 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15657 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15658 vector bool char vec_vsx_ld (int, const vector bool char *);
15659 vector signed char vec_vsx_ld (int, const vector signed char *);
15660 vector signed char vec_vsx_ld (int, const signed char *);
15661 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15662 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15664 void vec_vsx_st (vector double, int, vector double *);
15665 void vec_vsx_st (vector double, int, double *);
15666 void vec_vsx_st (vector float, int, vector float *);
15667 void vec_vsx_st (vector float, int, float *);
15668 void vec_vsx_st (vector signed int, int, vector signed int *);
15669 void vec_vsx_st (vector signed int, int, int *);
15670 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15671 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15672 void vec_vsx_st (vector bool int, int, vector bool int *);
15673 void vec_vsx_st (vector bool int, int, unsigned int *);
15674 void vec_vsx_st (vector bool int, int, int *);
15675 void vec_vsx_st (vector signed short, int, vector signed short *);
15676 void vec_vsx_st (vector signed short, int, short *);
15677 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15678 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15679 void vec_vsx_st (vector bool short, int, vector bool short *);
15680 void vec_vsx_st (vector bool short, int, unsigned short *);
15681 void vec_vsx_st (vector pixel, int, vector pixel *);
15682 void vec_vsx_st (vector pixel, int, unsigned short *);
15683 void vec_vsx_st (vector pixel, int, short *);
15684 void vec_vsx_st (vector bool short, int, short *);
15685 void vec_vsx_st (vector signed char, int, vector signed char *);
15686 void vec_vsx_st (vector signed char, int, signed char *);
15687 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15688 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15689 void vec_vsx_st (vector bool char, int, vector bool char *);
15690 void vec_vsx_st (vector bool char, int, unsigned char *);
15691 void vec_vsx_st (vector bool char, int, signed char *);
15693 vector double vec_xxpermdi (vector double, vector double, int);
15694 vector float vec_xxpermdi (vector float, vector float, int);
15695 vector long long vec_xxpermdi (vector long long, vector long long, int);
15696 vector unsigned long long vec_xxpermdi (vector unsigned long long,
15697 vector unsigned long long, int);
15698 vector int vec_xxpermdi (vector int, vector int, int);
15699 vector unsigned int vec_xxpermdi (vector unsigned int,
15700 vector unsigned int, int);
15701 vector short vec_xxpermdi (vector short, vector short, int);
15702 vector unsigned short vec_xxpermdi (vector unsigned short,
15703 vector unsigned short, int);
15704 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
15705 vector unsigned char vec_xxpermdi (vector unsigned char,
15706 vector unsigned char, int);
15708 vector double vec_xxsldi (vector double, vector double, int);
15709 vector float vec_xxsldi (vector float, vector float, int);
15710 vector long long vec_xxsldi (vector long long, vector long long, int);
15711 vector unsigned long long vec_xxsldi (vector unsigned long long,
15712 vector unsigned long long, int);
15713 vector int vec_xxsldi (vector int, vector int, int);
15714 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
15715 vector short vec_xxsldi (vector short, vector short, int);
15716 vector unsigned short vec_xxsldi (vector unsigned short,
15717 vector unsigned short, int);
15718 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
15719 vector unsigned char vec_xxsldi (vector unsigned char,
15720 vector unsigned char, int);
15723 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
15724 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
15725 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
15726 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
15727 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
15729 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15730 instruction set is available, the following additional functions are
15731 available for both 32-bit and 64-bit targets. For 64-bit targets, you
15732 can use @var{vector long} instead of @var{vector long long},
15733 @var{vector bool long} instead of @var{vector bool long long}, and
15734 @var{vector unsigned long} instead of @var{vector unsigned long long}.
15737 vector long long vec_abs (vector long long);
15739 vector long long vec_add (vector long long, vector long long);
15740 vector unsigned long long vec_add (vector unsigned long long,
15741 vector unsigned long long);
15743 int vec_all_eq (vector long long, vector long long);
15744 int vec_all_eq (vector unsigned long long, vector unsigned long long);
15745 int vec_all_ge (vector long long, vector long long);
15746 int vec_all_ge (vector unsigned long long, vector unsigned long long);
15747 int vec_all_gt (vector long long, vector long long);
15748 int vec_all_gt (vector unsigned long long, vector unsigned long long);
15749 int vec_all_le (vector long long, vector long long);
15750 int vec_all_le (vector unsigned long long, vector unsigned long long);
15751 int vec_all_lt (vector long long, vector long long);
15752 int vec_all_lt (vector unsigned long long, vector unsigned long long);
15753 int vec_all_ne (vector long long, vector long long);
15754 int vec_all_ne (vector unsigned long long, vector unsigned long long);
15756 int vec_any_eq (vector long long, vector long long);
15757 int vec_any_eq (vector unsigned long long, vector unsigned long long);
15758 int vec_any_ge (vector long long, vector long long);
15759 int vec_any_ge (vector unsigned long long, vector unsigned long long);
15760 int vec_any_gt (vector long long, vector long long);
15761 int vec_any_gt (vector unsigned long long, vector unsigned long long);
15762 int vec_any_le (vector long long, vector long long);
15763 int vec_any_le (vector unsigned long long, vector unsigned long long);
15764 int vec_any_lt (vector long long, vector long long);
15765 int vec_any_lt (vector unsigned long long, vector unsigned long long);
15766 int vec_any_ne (vector long long, vector long long);
15767 int vec_any_ne (vector unsigned long long, vector unsigned long long);
15769 vector long long vec_eqv (vector long long, vector long long);
15770 vector long long vec_eqv (vector bool long long, vector long long);
15771 vector long long vec_eqv (vector long long, vector bool long long);
15772 vector unsigned long long vec_eqv (vector unsigned long long,
15773 vector unsigned long long);
15774 vector unsigned long long vec_eqv (vector bool long long,
15775 vector unsigned long long);
15776 vector unsigned long long vec_eqv (vector unsigned long long,
15777 vector bool long long);
15778 vector int vec_eqv (vector int, vector int);
15779 vector int vec_eqv (vector bool int, vector int);
15780 vector int vec_eqv (vector int, vector bool int);
15781 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
15782 vector unsigned int vec_eqv (vector bool unsigned int,
15783 vector unsigned int);
15784 vector unsigned int vec_eqv (vector unsigned int,
15785 vector bool unsigned int);
15786 vector short vec_eqv (vector short, vector short);
15787 vector short vec_eqv (vector bool short, vector short);
15788 vector short vec_eqv (vector short, vector bool short);
15789 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
15790 vector unsigned short vec_eqv (vector bool unsigned short,
15791 vector unsigned short);
15792 vector unsigned short vec_eqv (vector unsigned short,
15793 vector bool unsigned short);
15794 vector signed char vec_eqv (vector signed char, vector signed char);
15795 vector signed char vec_eqv (vector bool signed char, vector signed char);
15796 vector signed char vec_eqv (vector signed char, vector bool signed char);
15797 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
15798 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
15799 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
15801 vector long long vec_max (vector long long, vector long long);
15802 vector unsigned long long vec_max (vector unsigned long long,
15803 vector unsigned long long);
15805 vector signed int vec_mergee (vector signed int, vector signed int);
15806 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
15807 vector bool int vec_mergee (vector bool int, vector bool int);
15809 vector signed int vec_mergeo (vector signed int, vector signed int);
15810 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
15811 vector bool int vec_mergeo (vector bool int, vector bool int);
15813 vector long long vec_min (vector long long, vector long long);
15814 vector unsigned long long vec_min (vector unsigned long long,
15815 vector unsigned long long);
15817 vector long long vec_nand (vector long long, vector long long);
15818 vector long long vec_nand (vector bool long long, vector long long);
15819 vector long long vec_nand (vector long long, vector bool long long);
15820 vector unsigned long long vec_nand (vector unsigned long long,
15821 vector unsigned long long);
15822 vector unsigned long long vec_nand (vector bool long long,
15823 vector unsigned long long);
15824 vector unsigned long long vec_nand (vector unsigned long long,
15825 vector bool long long);
15826 vector int vec_nand (vector int, vector int);
15827 vector int vec_nand (vector bool int, vector int);
15828 vector int vec_nand (vector int, vector bool int);
15829 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
15830 vector unsigned int vec_nand (vector bool unsigned int,
15831 vector unsigned int);
15832 vector unsigned int vec_nand (vector unsigned int,
15833 vector bool unsigned int);
15834 vector short vec_nand (vector short, vector short);
15835 vector short vec_nand (vector bool short, vector short);
15836 vector short vec_nand (vector short, vector bool short);
15837 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
15838 vector unsigned short vec_nand (vector bool unsigned short,
15839 vector unsigned short);
15840 vector unsigned short vec_nand (vector unsigned short,
15841 vector bool unsigned short);
15842 vector signed char vec_nand (vector signed char, vector signed char);
15843 vector signed char vec_nand (vector bool signed char, vector signed char);
15844 vector signed char vec_nand (vector signed char, vector bool signed char);
15845 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
15846 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
15847 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
15849 vector long long vec_orc (vector long long, vector long long);
15850 vector long long vec_orc (vector bool long long, vector long long);
15851 vector long long vec_orc (vector long long, vector bool long long);
15852 vector unsigned long long vec_orc (vector unsigned long long,
15853 vector unsigned long long);
15854 vector unsigned long long vec_orc (vector bool long long,
15855 vector unsigned long long);
15856 vector unsigned long long vec_orc (vector unsigned long long,
15857 vector bool long long);
15858 vector int vec_orc (vector int, vector int);
15859 vector int vec_orc (vector bool int, vector int);
15860 vector int vec_orc (vector int, vector bool int);
15861 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
15862 vector unsigned int vec_orc (vector bool unsigned int,
15863 vector unsigned int);
15864 vector unsigned int vec_orc (vector unsigned int,
15865 vector bool unsigned int);
15866 vector short vec_orc (vector short, vector short);
15867 vector short vec_orc (vector bool short, vector short);
15868 vector short vec_orc (vector short, vector bool short);
15869 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
15870 vector unsigned short vec_orc (vector bool unsigned short,
15871 vector unsigned short);
15872 vector unsigned short vec_orc (vector unsigned short,
15873 vector bool unsigned short);
15874 vector signed char vec_orc (vector signed char, vector signed char);
15875 vector signed char vec_orc (vector bool signed char, vector signed char);
15876 vector signed char vec_orc (vector signed char, vector bool signed char);
15877 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
15878 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
15879 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
15881 vector int vec_pack (vector long long, vector long long);
15882 vector unsigned int vec_pack (vector unsigned long long,
15883 vector unsigned long long);
15884 vector bool int vec_pack (vector bool long long, vector bool long long);
15886 vector int vec_packs (vector long long, vector long long);
15887 vector unsigned int vec_packs (vector unsigned long long,
15888 vector unsigned long long);
15890 vector unsigned int vec_packsu (vector long long, vector long long);
15891 vector unsigned int vec_packsu (vector unsigned long long,
15892 vector unsigned long long);
15894 vector long long vec_rl (vector long long,
15895 vector unsigned long long);
15896 vector long long vec_rl (vector unsigned long long,
15897 vector unsigned long long);
15899 vector long long vec_sl (vector long long, vector unsigned long long);
15900 vector long long vec_sl (vector unsigned long long,
15901 vector unsigned long long);
15903 vector long long vec_sr (vector long long, vector unsigned long long);
15904 vector unsigned long long char vec_sr (vector unsigned long long,
15905 vector unsigned long long);
15907 vector long long vec_sra (vector long long, vector unsigned long long);
15908 vector unsigned long long vec_sra (vector unsigned long long,
15909 vector unsigned long long);
15911 vector long long vec_sub (vector long long, vector long long);
15912 vector unsigned long long vec_sub (vector unsigned long long,
15913 vector unsigned long long);
15915 vector long long vec_unpackh (vector int);
15916 vector unsigned long long vec_unpackh (vector unsigned int);
15918 vector long long vec_unpackl (vector int);
15919 vector unsigned long long vec_unpackl (vector unsigned int);
15921 vector long long vec_vaddudm (vector long long, vector long long);
15922 vector long long vec_vaddudm (vector bool long long, vector long long);
15923 vector long long vec_vaddudm (vector long long, vector bool long long);
15924 vector unsigned long long vec_vaddudm (vector unsigned long long,
15925 vector unsigned long long);
15926 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
15927 vector unsigned long long);
15928 vector unsigned long long vec_vaddudm (vector unsigned long long,
15929 vector bool unsigned long long);
15931 vector long long vec_vbpermq (vector signed char, vector signed char);
15932 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
15934 vector long long vec_cntlz (vector long long);
15935 vector unsigned long long vec_cntlz (vector unsigned long long);
15936 vector int vec_cntlz (vector int);
15937 vector unsigned int vec_cntlz (vector int);
15938 vector short vec_cntlz (vector short);
15939 vector unsigned short vec_cntlz (vector unsigned short);
15940 vector signed char vec_cntlz (vector signed char);
15941 vector unsigned char vec_cntlz (vector unsigned char);
15943 vector long long vec_vclz (vector long long);
15944 vector unsigned long long vec_vclz (vector unsigned long long);
15945 vector int vec_vclz (vector int);
15946 vector unsigned int vec_vclz (vector int);
15947 vector short vec_vclz (vector short);
15948 vector unsigned short vec_vclz (vector unsigned short);
15949 vector signed char vec_vclz (vector signed char);
15950 vector unsigned char vec_vclz (vector unsigned char);
15952 vector signed char vec_vclzb (vector signed char);
15953 vector unsigned char vec_vclzb (vector unsigned char);
15955 vector long long vec_vclzd (vector long long);
15956 vector unsigned long long vec_vclzd (vector unsigned long long);
15958 vector short vec_vclzh (vector short);
15959 vector unsigned short vec_vclzh (vector unsigned short);
15961 vector int vec_vclzw (vector int);
15962 vector unsigned int vec_vclzw (vector int);
15964 vector signed char vec_vgbbd (vector signed char);
15965 vector unsigned char vec_vgbbd (vector unsigned char);
15967 vector long long vec_vmaxsd (vector long long, vector long long);
15969 vector unsigned long long vec_vmaxud (vector unsigned long long,
15970 unsigned vector long long);
15972 vector long long vec_vminsd (vector long long, vector long long);
15974 vector unsigned long long vec_vminud (vector long long,
15977 vector int vec_vpksdss (vector long long, vector long long);
15978 vector unsigned int vec_vpksdss (vector long long, vector long long);
15980 vector unsigned int vec_vpkudus (vector unsigned long long,
15981 vector unsigned long long);
15983 vector int vec_vpkudum (vector long long, vector long long);
15984 vector unsigned int vec_vpkudum (vector unsigned long long,
15985 vector unsigned long long);
15986 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
15988 vector long long vec_vpopcnt (vector long long);
15989 vector unsigned long long vec_vpopcnt (vector unsigned long long);
15990 vector int vec_vpopcnt (vector int);
15991 vector unsigned int vec_vpopcnt (vector int);
15992 vector short vec_vpopcnt (vector short);
15993 vector unsigned short vec_vpopcnt (vector unsigned short);
15994 vector signed char vec_vpopcnt (vector signed char);
15995 vector unsigned char vec_vpopcnt (vector unsigned char);
15997 vector signed char vec_vpopcntb (vector signed char);
15998 vector unsigned char vec_vpopcntb (vector unsigned char);
16000 vector long long vec_vpopcntd (vector long long);
16001 vector unsigned long long vec_vpopcntd (vector unsigned long long);
16003 vector short vec_vpopcnth (vector short);
16004 vector unsigned short vec_vpopcnth (vector unsigned short);
16006 vector int vec_vpopcntw (vector int);
16007 vector unsigned int vec_vpopcntw (vector int);
16009 vector long long vec_vrld (vector long long, vector unsigned long long);
16010 vector unsigned long long vec_vrld (vector unsigned long long,
16011 vector unsigned long long);
16013 vector long long vec_vsld (vector long long, vector unsigned long long);
16014 vector long long vec_vsld (vector unsigned long long,
16015 vector unsigned long long);
16017 vector long long vec_vsrad (vector long long, vector unsigned long long);
16018 vector unsigned long long vec_vsrad (vector unsigned long long,
16019 vector unsigned long long);
16021 vector long long vec_vsrd (vector long long, vector unsigned long long);
16022 vector unsigned long long char vec_vsrd (vector unsigned long long,
16023 vector unsigned long long);
16025 vector long long vec_vsubudm (vector long long, vector long long);
16026 vector long long vec_vsubudm (vector bool long long, vector long long);
16027 vector long long vec_vsubudm (vector long long, vector bool long long);
16028 vector unsigned long long vec_vsubudm (vector unsigned long long,
16029 vector unsigned long long);
16030 vector unsigned long long vec_vsubudm (vector bool long long,
16031 vector unsigned long long);
16032 vector unsigned long long vec_vsubudm (vector unsigned long long,
16033 vector bool long long);
16035 vector long long vec_vupkhsw (vector int);
16036 vector unsigned long long vec_vupkhsw (vector unsigned int);
16038 vector long long vec_vupklsw (vector int);
16039 vector unsigned long long vec_vupklsw (vector int);
16042 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16043 instruction set is available, the following additional functions are
16044 available for 64-bit targets. New vector types
16045 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
16046 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
16049 The normal vector extract, and set operations work on
16050 @var{vector __int128_t} and @var{vector __uint128_t} types,
16051 but the index value must be 0.
16054 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
16055 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
16057 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
16058 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
16060 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
16061 vector __int128_t);
16062 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
16063 vector __uint128_t);
16065 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
16066 vector __int128_t);
16067 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
16068 vector __uint128_t);
16070 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
16071 vector __int128_t);
16072 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
16073 vector __uint128_t);
16075 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
16076 vector __int128_t);
16077 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
16078 vector __uint128_t);
16080 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
16081 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
16083 __int128_t vec_vsubuqm (__int128_t, __int128_t);
16084 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
16086 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
16087 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
16088 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
16089 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
16090 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
16091 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
16092 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
16093 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
16094 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
16095 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
16098 If the cryptographic instructions are enabled (@option{-mcrypto} or
16099 @option{-mcpu=power8}), the following builtins are enabled.
16102 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
16104 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
16105 vector unsigned long long);
16107 vector unsigned long long __builtin_crypto_vcipherlast
16108 (vector unsigned long long,
16109 vector unsigned long long);
16111 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
16112 vector unsigned long long);
16114 vector unsigned long long __builtin_crypto_vncipherlast
16115 (vector unsigned long long,
16116 vector unsigned long long);
16118 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
16119 vector unsigned char,
16120 vector unsigned char);
16122 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
16123 vector unsigned short,
16124 vector unsigned short);
16126 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
16127 vector unsigned int,
16128 vector unsigned int);
16130 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
16131 vector unsigned long long,
16132 vector unsigned long long);
16134 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
16135 vector unsigned char);
16137 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
16138 vector unsigned short);
16140 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
16141 vector unsigned int);
16143 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
16144 vector unsigned long long);
16146 vector unsigned long long __builtin_crypto_vshasigmad
16147 (vector unsigned long long, int, int);
16149 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
16153 The second argument to the @var{__builtin_crypto_vshasigmad} and
16154 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
16155 integer that is 0 or 1. The third argument to these builtin functions
16156 must be a constant integer in the range of 0 to 15.
16158 @node PowerPC Hardware Transactional Memory Built-in Functions
16159 @subsection PowerPC Hardware Transactional Memory Built-in Functions
16160 GCC provides two interfaces for accessing the Hardware Transactional
16161 Memory (HTM) instructions available on some of the PowerPC family
16162 of processors (eg, POWER8). The two interfaces come in a low level
16163 interface, consisting of built-in functions specific to PowerPC and a
16164 higher level interface consisting of inline functions that are common
16165 between PowerPC and S/390.
16167 @subsubsection PowerPC HTM Low Level Built-in Functions
16169 The following low level built-in functions are available with
16170 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16171 They all generate the machine instruction that is part of the name.
16173 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16174 the full 4-bit condition register value set by their associated hardware
16175 instruction. The header file @code{htmintrin.h} defines some macros that can
16176 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16177 returns a simple true or false value depending on whether a transaction was
16178 successfully started or not. The arguments of the builtins match exactly the
16179 type and order of the associated hardware instruction's operands, except for
16180 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16181 Refer to the ISA manual for a description of each instruction's operands.
16184 unsigned int __builtin_tbegin (unsigned int)
16185 unsigned int __builtin_tend (unsigned int)
16187 unsigned int __builtin_tabort (unsigned int)
16188 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16189 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16190 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16191 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16193 unsigned int __builtin_tcheck (void)
16194 unsigned int __builtin_treclaim (unsigned int)
16195 unsigned int __builtin_trechkpt (void)
16196 unsigned int __builtin_tsr (unsigned int)
16199 In addition to the above HTM built-ins, we have added built-ins for
16200 some common extended mnemonics of the HTM instructions:
16203 unsigned int __builtin_tendall (void)
16204 unsigned int __builtin_tresume (void)
16205 unsigned int __builtin_tsuspend (void)
16208 Note that the semantics of the above HTM builtins are required to mimic
16209 the locking semantics used for critical sections. Builtins that are used
16210 to create a new transaction or restart a suspended transaction must have
16211 lock acquisition like semantics while those builtins that end or suspend a
16212 transaction must have lock release like semantics. Specifically, this must
16213 mimic lock semantics as specified by C++11, for example: Lock acquisition is
16214 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
16215 that returns 0, and lock release is as-if an execution of
16216 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
16217 implicit implementation-defined lock used for all transactions. The HTM
16218 instructions associated with with the builtins inherently provide the
16219 correct acquisition and release hardware barriers required. However,
16220 the compiler must also be prohibited from moving loads and stores across
16221 the builtins in a way that would violate their semantics. This has been
16222 accomplished by adding memory barriers to the associated HTM instructions
16223 (which is a conservative approach to provide acquire and release semantics).
16224 Earlier versions of the compiler did not treat the HTM instructions as
16225 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
16226 be used to determine whether the current compiler treats HTM instructions
16227 as memory barriers or not. This allows the user to explicitly add memory
16228 barriers to their code when using an older version of the compiler.
16230 The following set of built-in functions are available to gain access
16231 to the HTM specific special purpose registers.
16234 unsigned long __builtin_get_texasr (void)
16235 unsigned long __builtin_get_texasru (void)
16236 unsigned long __builtin_get_tfhar (void)
16237 unsigned long __builtin_get_tfiar (void)
16239 void __builtin_set_texasr (unsigned long);
16240 void __builtin_set_texasru (unsigned long);
16241 void __builtin_set_tfhar (unsigned long);
16242 void __builtin_set_tfiar (unsigned long);
16245 Example usage of these low level built-in functions may look like:
16248 #include <htmintrin.h>
16250 int num_retries = 10;
16254 if (__builtin_tbegin (0))
16256 /* Transaction State Initiated. */
16257 if (is_locked (lock))
16258 __builtin_tabort (0);
16259 ... transaction code...
16260 __builtin_tend (0);
16265 /* Transaction State Failed. Use locks if the transaction
16266 failure is "persistent" or we've tried too many times. */
16267 if (num_retries-- <= 0
16268 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16270 acquire_lock (lock);
16271 ... non transactional fallback path...
16272 release_lock (lock);
16279 One final built-in function has been added that returns the value of
16280 the 2-bit Transaction State field of the Machine Status Register (MSR)
16281 as stored in @code{CR0}.
16284 unsigned long __builtin_ttest (void)
16287 This built-in can be used to determine the current transaction state
16288 using the following code example:
16291 #include <htmintrin.h>
16293 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16295 if (tx_state == _HTM_TRANSACTIONAL)
16297 /* Code to use in transactional state. */
16299 else if (tx_state == _HTM_NONTRANSACTIONAL)
16301 /* Code to use in non-transactional state. */
16303 else if (tx_state == _HTM_SUSPENDED)
16305 /* Code to use in transaction suspended state. */
16309 @subsubsection PowerPC HTM High Level Inline Functions
16311 The following high level HTM interface is made available by including
16312 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16313 where CPU is `power8' or later. This interface is common between PowerPC
16314 and S/390, allowing users to write one HTM source implementation that
16315 can be compiled and executed on either system.
16318 long __TM_simple_begin (void)
16319 long __TM_begin (void* const TM_buff)
16320 long __TM_end (void)
16321 void __TM_abort (void)
16322 void __TM_named_abort (unsigned char const code)
16323 void __TM_resume (void)
16324 void __TM_suspend (void)
16326 long __TM_is_user_abort (void* const TM_buff)
16327 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16328 long __TM_is_illegal (void* const TM_buff)
16329 long __TM_is_footprint_exceeded (void* const TM_buff)
16330 long __TM_nesting_depth (void* const TM_buff)
16331 long __TM_is_nested_too_deep(void* const TM_buff)
16332 long __TM_is_conflict(void* const TM_buff)
16333 long __TM_is_failure_persistent(void* const TM_buff)
16334 long __TM_failure_address(void* const TM_buff)
16335 long long __TM_failure_code(void* const TM_buff)
16338 Using these common set of HTM inline functions, we can create
16339 a more portable version of the HTM example in the previous
16340 section that will work on either PowerPC or S/390:
16343 #include <htmxlintrin.h>
16345 int num_retries = 10;
16346 TM_buff_type TM_buff;
16350 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16352 /* Transaction State Initiated. */
16353 if (is_locked (lock))
16355 ... transaction code...
16361 /* Transaction State Failed. Use locks if the transaction
16362 failure is "persistent" or we've tried too many times. */
16363 if (num_retries-- <= 0
16364 || __TM_is_failure_persistent (TM_buff))
16366 acquire_lock (lock);
16367 ... non transactional fallback path...
16368 release_lock (lock);
16375 @node RX Built-in Functions
16376 @subsection RX Built-in Functions
16377 GCC supports some of the RX instructions which cannot be expressed in
16378 the C programming language via the use of built-in functions. The
16379 following functions are supported:
16381 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16382 Generates the @code{brk} machine instruction.
16385 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16386 Generates the @code{clrpsw} machine instruction to clear the specified
16387 bit in the processor status word.
16390 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16391 Generates the @code{int} machine instruction to generate an interrupt
16392 with the specified value.
16395 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16396 Generates the @code{machi} machine instruction to add the result of
16397 multiplying the top 16 bits of the two arguments into the
16401 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16402 Generates the @code{maclo} machine instruction to add the result of
16403 multiplying the bottom 16 bits of the two arguments into the
16407 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16408 Generates the @code{mulhi} machine instruction to place the result of
16409 multiplying the top 16 bits of the two arguments into the
16413 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16414 Generates the @code{mullo} machine instruction to place the result of
16415 multiplying the bottom 16 bits of the two arguments into the
16419 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16420 Generates the @code{mvfachi} machine instruction to read the top
16421 32 bits of the accumulator.
16424 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16425 Generates the @code{mvfacmi} machine instruction to read the middle
16426 32 bits of the accumulator.
16429 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16430 Generates the @code{mvfc} machine instruction which reads the control
16431 register specified in its argument and returns its value.
16434 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16435 Generates the @code{mvtachi} machine instruction to set the top
16436 32 bits of the accumulator.
16439 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16440 Generates the @code{mvtaclo} machine instruction to set the bottom
16441 32 bits of the accumulator.
16444 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16445 Generates the @code{mvtc} machine instruction which sets control
16446 register number @code{reg} to @code{val}.
16449 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16450 Generates the @code{mvtipl} machine instruction set the interrupt
16454 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16455 Generates the @code{racw} machine instruction to round the accumulator
16456 according to the specified mode.
16459 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16460 Generates the @code{revw} machine instruction which swaps the bytes in
16461 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16462 and also bits 16--23 occupy bits 24--31 and vice versa.
16465 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16466 Generates the @code{rmpa} machine instruction which initiates a
16467 repeated multiply and accumulate sequence.
16470 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16471 Generates the @code{round} machine instruction which returns the
16472 floating-point argument rounded according to the current rounding mode
16473 set in the floating-point status word register.
16476 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16477 Generates the @code{sat} machine instruction which returns the
16478 saturated value of the argument.
16481 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16482 Generates the @code{setpsw} machine instruction to set the specified
16483 bit in the processor status word.
16486 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16487 Generates the @code{wait} machine instruction.
16490 @node S/390 System z Built-in Functions
16491 @subsection S/390 System z Built-in Functions
16492 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16493 Generates the @code{tbegin} machine instruction starting a
16494 non-constraint hardware transaction. If the parameter is non-NULL the
16495 memory area is used to store the transaction diagnostic buffer and
16496 will be passed as first operand to @code{tbegin}. This buffer can be
16497 defined using the @code{struct __htm_tdb} C struct defined in
16498 @code{htmintrin.h} and must reside on a double-word boundary. The
16499 second tbegin operand is set to @code{0xff0c}. This enables
16500 save/restore of all GPRs and disables aborts for FPR and AR
16501 manipulations inside the transaction body. The condition code set by
16502 the tbegin instruction is returned as integer value. The tbegin
16503 instruction by definition overwrites the content of all FPRs. The
16504 compiler will generate code which saves and restores the FPRs. For
16505 soft-float code it is recommended to used the @code{*_nofloat}
16506 variant. In order to prevent a TDB from being written it is required
16507 to pass an constant zero value as parameter. Passing the zero value
16508 through a variable is not sufficient. Although modifications of
16509 access registers inside the transaction will not trigger an
16510 transaction abort it is not supported to actually modify them. Access
16511 registers do not get saved when entering a transaction. They will have
16512 undefined state when reaching the abort code.
16515 Macros for the possible return codes of tbegin are defined in the
16516 @code{htmintrin.h} header file:
16519 @item _HTM_TBEGIN_STARTED
16520 @code{tbegin} has been executed as part of normal processing. The
16521 transaction body is supposed to be executed.
16522 @item _HTM_TBEGIN_INDETERMINATE
16523 The transaction was aborted due to an indeterminate condition which
16524 might be persistent.
16525 @item _HTM_TBEGIN_TRANSIENT
16526 The transaction aborted due to a transient failure. The transaction
16527 should be re-executed in that case.
16528 @item _HTM_TBEGIN_PERSISTENT
16529 The transaction aborted due to a persistent failure. Re-execution
16530 under same circumstances will not be productive.
16533 @defmac _HTM_FIRST_USER_ABORT_CODE
16534 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16535 specifies the first abort code which can be used for
16536 @code{__builtin_tabort}. Values below this threshold are reserved for
16540 @deftp {Data type} {struct __htm_tdb}
16541 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16542 the structure of the transaction diagnostic block as specified in the
16543 Principles of Operation manual chapter 5-91.
16546 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16547 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16548 Using this variant in code making use of FPRs will leave the FPRs in
16549 undefined state when entering the transaction abort handler code.
16552 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16553 In addition to @code{__builtin_tbegin} a loop for transient failures
16554 is generated. If tbegin returns a condition code of 2 the transaction
16555 will be retried as often as specified in the second argument. The
16556 perform processor assist instruction is used to tell the CPU about the
16557 number of fails so far.
16560 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16561 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16562 restores. Using this variant in code making use of FPRs will leave
16563 the FPRs in undefined state when entering the transaction abort
16567 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16568 Generates the @code{tbeginc} machine instruction starting a constraint
16569 hardware transaction. The second operand is set to @code{0xff08}.
16572 @deftypefn {Built-in Function} int __builtin_tend (void)
16573 Generates the @code{tend} machine instruction finishing a transaction
16574 and making the changes visible to other threads. The condition code
16575 generated by tend is returned as integer value.
16578 @deftypefn {Built-in Function} void __builtin_tabort (int)
16579 Generates the @code{tabort} machine instruction with the specified
16580 abort code. Abort codes from 0 through 255 are reserved and will
16581 result in an error message.
16584 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16585 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16586 integer parameter is loaded into rX and a value of zero is loaded into
16587 rY. The integer parameter specifies the number of times the
16588 transaction repeatedly aborted.
16591 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16592 Generates the @code{etnd} machine instruction. The current nesting
16593 depth is returned as integer value. For a nesting depth of 0 the code
16594 is not executed as part of an transaction.
16597 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16599 Generates the @code{ntstg} machine instruction. The second argument
16600 is written to the first arguments location. The store operation will
16601 not be rolled-back in case of an transaction abort.
16604 @node SH Built-in Functions
16605 @subsection SH Built-in Functions
16606 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16607 families of processors:
16609 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16610 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16611 used by system code that manages threads and execution contexts. The compiler
16612 normally does not generate code that modifies the contents of @samp{GBR} and
16613 thus the value is preserved across function calls. Changing the @samp{GBR}
16614 value in user code must be done with caution, since the compiler might use
16615 @samp{GBR} in order to access thread local variables.
16619 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16620 Returns the value that is currently set in the @samp{GBR} register.
16621 Memory loads and stores that use the thread pointer as a base address are
16622 turned into @samp{GBR} based displacement loads and stores, if possible.
16630 int get_tcb_value (void)
16632 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16633 return ((my_tcb*)__builtin_thread_pointer ())->c;
16639 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16640 Returns the value that is currently set in the @samp{FPSCR} register.
16643 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16644 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16645 preserving the current values of the FR, SZ and PR bits.
16648 @node SPARC VIS Built-in Functions
16649 @subsection SPARC VIS Built-in Functions
16651 GCC supports SIMD operations on the SPARC using both the generic vector
16652 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16653 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16654 switch, the VIS extension is exposed as the following built-in functions:
16657 typedef int v1si __attribute__ ((vector_size (4)));
16658 typedef int v2si __attribute__ ((vector_size (8)));
16659 typedef short v4hi __attribute__ ((vector_size (8)));
16660 typedef short v2hi __attribute__ ((vector_size (4)));
16661 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16662 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16664 void __builtin_vis_write_gsr (int64_t);
16665 int64_t __builtin_vis_read_gsr (void);
16667 void * __builtin_vis_alignaddr (void *, long);
16668 void * __builtin_vis_alignaddrl (void *, long);
16669 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16670 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16671 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16672 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16674 v4hi __builtin_vis_fexpand (v4qi);
16676 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16677 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16678 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16679 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16680 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16681 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16682 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16684 v4qi __builtin_vis_fpack16 (v4hi);
16685 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16686 v2hi __builtin_vis_fpackfix (v2si);
16687 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16689 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
16691 long __builtin_vis_edge8 (void *, void *);
16692 long __builtin_vis_edge8l (void *, void *);
16693 long __builtin_vis_edge16 (void *, void *);
16694 long __builtin_vis_edge16l (void *, void *);
16695 long __builtin_vis_edge32 (void *, void *);
16696 long __builtin_vis_edge32l (void *, void *);
16698 long __builtin_vis_fcmple16 (v4hi, v4hi);
16699 long __builtin_vis_fcmple32 (v2si, v2si);
16700 long __builtin_vis_fcmpne16 (v4hi, v4hi);
16701 long __builtin_vis_fcmpne32 (v2si, v2si);
16702 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
16703 long __builtin_vis_fcmpgt32 (v2si, v2si);
16704 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
16705 long __builtin_vis_fcmpeq32 (v2si, v2si);
16707 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
16708 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
16709 v2si __builtin_vis_fpadd32 (v2si, v2si);
16710 v1si __builtin_vis_fpadd32s (v1si, v1si);
16711 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
16712 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
16713 v2si __builtin_vis_fpsub32 (v2si, v2si);
16714 v1si __builtin_vis_fpsub32s (v1si, v1si);
16716 long __builtin_vis_array8 (long, long);
16717 long __builtin_vis_array16 (long, long);
16718 long __builtin_vis_array32 (long, long);
16721 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
16722 functions also become available:
16725 long __builtin_vis_bmask (long, long);
16726 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
16727 v2si __builtin_vis_bshufflev2si (v2si, v2si);
16728 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
16729 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
16731 long __builtin_vis_edge8n (void *, void *);
16732 long __builtin_vis_edge8ln (void *, void *);
16733 long __builtin_vis_edge16n (void *, void *);
16734 long __builtin_vis_edge16ln (void *, void *);
16735 long __builtin_vis_edge32n (void *, void *);
16736 long __builtin_vis_edge32ln (void *, void *);
16739 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
16740 functions also become available:
16743 void __builtin_vis_cmask8 (long);
16744 void __builtin_vis_cmask16 (long);
16745 void __builtin_vis_cmask32 (long);
16747 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
16749 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
16750 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
16751 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
16752 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
16753 v2si __builtin_vis_fsll16 (v2si, v2si);
16754 v2si __builtin_vis_fslas16 (v2si, v2si);
16755 v2si __builtin_vis_fsrl16 (v2si, v2si);
16756 v2si __builtin_vis_fsra16 (v2si, v2si);
16758 long __builtin_vis_pdistn (v8qi, v8qi);
16760 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
16762 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
16763 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
16765 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
16766 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
16767 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
16768 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
16769 v2si __builtin_vis_fpadds32 (v2si, v2si);
16770 v1si __builtin_vis_fpadds32s (v1si, v1si);
16771 v2si __builtin_vis_fpsubs32 (v2si, v2si);
16772 v1si __builtin_vis_fpsubs32s (v1si, v1si);
16774 long __builtin_vis_fucmple8 (v8qi, v8qi);
16775 long __builtin_vis_fucmpne8 (v8qi, v8qi);
16776 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
16777 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
16779 float __builtin_vis_fhadds (float, float);
16780 double __builtin_vis_fhaddd (double, double);
16781 float __builtin_vis_fhsubs (float, float);
16782 double __builtin_vis_fhsubd (double, double);
16783 float __builtin_vis_fnhadds (float, float);
16784 double __builtin_vis_fnhaddd (double, double);
16786 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
16787 int64_t __builtin_vis_xmulx (int64_t, int64_t);
16788 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
16791 @node SPU Built-in Functions
16792 @subsection SPU Built-in Functions
16794 GCC provides extensions for the SPU processor as described in the
16795 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
16796 found at @uref{http://cell.scei.co.jp/} or
16797 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
16798 implementation differs in several ways.
16803 The optional extension of specifying vector constants in parentheses is
16807 A vector initializer requires no cast if the vector constant is of the
16808 same type as the variable it is initializing.
16811 If @code{signed} or @code{unsigned} is omitted, the signedness of the
16812 vector type is the default signedness of the base type. The default
16813 varies depending on the operating system, so a portable program should
16814 always specify the signedness.
16817 By default, the keyword @code{__vector} is added. The macro
16818 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
16822 GCC allows using a @code{typedef} name as the type specifier for a
16826 For C, overloaded functions are implemented with macros so the following
16830 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
16834 Since @code{spu_add} is a macro, the vector constant in the example
16835 is treated as four separate arguments. Wrap the entire argument in
16836 parentheses for this to work.
16839 The extended version of @code{__builtin_expect} is not supported.
16843 @emph{Note:} Only the interface described in the aforementioned
16844 specification is supported. Internally, GCC uses built-in functions to
16845 implement the required functionality, but these are not supported and
16846 are subject to change without notice.
16848 @node TI C6X Built-in Functions
16849 @subsection TI C6X Built-in Functions
16851 GCC provides intrinsics to access certain instructions of the TI C6X
16852 processors. These intrinsics, listed below, are available after
16853 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
16854 to C6X instructions.
16858 int _sadd (int, int)
16859 int _ssub (int, int)
16860 int _sadd2 (int, int)
16861 int _ssub2 (int, int)
16862 long long _mpy2 (int, int)
16863 long long _smpy2 (int, int)
16864 int _add4 (int, int)
16865 int _sub4 (int, int)
16866 int _saddu4 (int, int)
16868 int _smpy (int, int)
16869 int _smpyh (int, int)
16870 int _smpyhl (int, int)
16871 int _smpylh (int, int)
16873 int _sshl (int, int)
16874 int _subc (int, int)
16876 int _avg2 (int, int)
16877 int _avgu4 (int, int)
16879 int _clrr (int, int)
16880 int _extr (int, int)
16881 int _extru (int, int)
16887 @node TILE-Gx Built-in Functions
16888 @subsection TILE-Gx Built-in Functions
16890 GCC provides intrinsics to access every instruction of the TILE-Gx
16891 processor. The intrinsics are of the form:
16895 unsigned long long __insn_@var{op} (...)
16899 Where @var{op} is the name of the instruction. Refer to the ISA manual
16900 for the complete list of instructions.
16902 GCC also provides intrinsics to directly access the network registers.
16903 The intrinsics are:
16907 unsigned long long __tile_idn0_receive (void)
16908 unsigned long long __tile_idn1_receive (void)
16909 unsigned long long __tile_udn0_receive (void)
16910 unsigned long long __tile_udn1_receive (void)
16911 unsigned long long __tile_udn2_receive (void)
16912 unsigned long long __tile_udn3_receive (void)
16913 void __tile_idn_send (unsigned long long)
16914 void __tile_udn_send (unsigned long long)
16918 The intrinsic @code{void __tile_network_barrier (void)} is used to
16919 guarantee that no network operations before it are reordered with
16922 @node TILEPro Built-in Functions
16923 @subsection TILEPro Built-in Functions
16925 GCC provides intrinsics to access every instruction of the TILEPro
16926 processor. The intrinsics are of the form:
16930 unsigned __insn_@var{op} (...)
16935 where @var{op} is the name of the instruction. Refer to the ISA manual
16936 for the complete list of instructions.
16938 GCC also provides intrinsics to directly access the network registers.
16939 The intrinsics are:
16943 unsigned __tile_idn0_receive (void)
16944 unsigned __tile_idn1_receive (void)
16945 unsigned __tile_sn_receive (void)
16946 unsigned __tile_udn0_receive (void)
16947 unsigned __tile_udn1_receive (void)
16948 unsigned __tile_udn2_receive (void)
16949 unsigned __tile_udn3_receive (void)
16950 void __tile_idn_send (unsigned)
16951 void __tile_sn_send (unsigned)
16952 void __tile_udn_send (unsigned)
16956 The intrinsic @code{void __tile_network_barrier (void)} is used to
16957 guarantee that no network operations before it are reordered with
16960 @node x86 Built-in Functions
16961 @subsection x86 Built-in Functions
16963 These built-in functions are available for the x86-32 and x86-64 family
16964 of computers, depending on the command-line switches used.
16966 If you specify command-line switches such as @option{-msse},
16967 the compiler could use the extended instruction sets even if the built-ins
16968 are not used explicitly in the program. For this reason, applications
16969 that perform run-time CPU detection must compile separate files for each
16970 supported architecture, using the appropriate flags. In particular,
16971 the file containing the CPU detection code should be compiled without
16974 The following machine modes are available for use with MMX built-in functions
16975 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
16976 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
16977 vector of eight 8-bit integers. Some of the built-in functions operate on
16978 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
16980 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
16981 of two 32-bit floating-point values.
16983 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
16984 floating-point values. Some instructions use a vector of four 32-bit
16985 integers, these use @code{V4SI}. Finally, some instructions operate on an
16986 entire vector register, interpreting it as a 128-bit integer, these use mode
16989 In 64-bit mode, the x86-64 family of processors uses additional built-in
16990 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
16991 floating point and @code{TC} 128-bit complex floating-point values.
16993 The following floating-point built-in functions are available in 64-bit
16994 mode. All of them implement the function that is part of the name.
16997 __float128 __builtin_fabsq (__float128)
16998 __float128 __builtin_copysignq (__float128, __float128)
17001 The following built-in function is always available.
17004 @item void __builtin_ia32_pause (void)
17005 Generates the @code{pause} machine instruction with a compiler memory
17009 The following floating-point built-in functions are made available in the
17013 @item __float128 __builtin_infq (void)
17014 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
17015 @findex __builtin_infq
17017 @item __float128 __builtin_huge_valq (void)
17018 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
17019 @findex __builtin_huge_valq
17022 The following built-in functions are always available and can be used to
17023 check the target platform type.
17025 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
17026 This function runs the CPU detection code to check the type of CPU and the
17027 features supported. This built-in function needs to be invoked along with the built-in functions
17028 to check CPU type and features, @code{__builtin_cpu_is} and
17029 @code{__builtin_cpu_supports}, only when used in a function that is
17030 executed before any constructors are called. The CPU detection code is
17031 automatically executed in a very high priority constructor.
17033 For example, this function has to be used in @code{ifunc} resolvers that
17034 check for CPU type using the built-in functions @code{__builtin_cpu_is}
17035 and @code{__builtin_cpu_supports}, or in constructors on targets that
17036 don't support constructor priority.
17039 static void (*resolve_memcpy (void)) (void)
17041 // ifunc resolvers fire before constructors, explicitly call the init
17043 __builtin_cpu_init ();
17044 if (__builtin_cpu_supports ("ssse3"))
17045 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
17047 return default_memcpy;
17050 void *memcpy (void *, const void *, size_t)
17051 __attribute__ ((ifunc ("resolve_memcpy")));
17056 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
17057 This function returns a positive integer if the run-time CPU
17058 is of type @var{cpuname}
17059 and returns @code{0} otherwise. The following CPU names can be detected:
17075 Intel Core i7 Nehalem CPU.
17078 Intel Core i7 Westmere CPU.
17081 Intel Core i7 Sandy Bridge CPU.
17087 AMD Family 10h CPU.
17090 AMD Family 10h Barcelona CPU.
17093 AMD Family 10h Shanghai CPU.
17096 AMD Family 10h Istanbul CPU.
17099 AMD Family 14h CPU.
17102 AMD Family 15h CPU.
17105 AMD Family 15h Bulldozer version 1.
17108 AMD Family 15h Bulldozer version 2.
17111 AMD Family 15h Bulldozer version 3.
17114 AMD Family 15h Bulldozer version 4.
17117 AMD Family 16h CPU.
17120 AMD Family 17h CPU.
17123 Here is an example:
17125 if (__builtin_cpu_is ("corei7"))
17127 do_corei7 (); // Core i7 specific implementation.
17131 do_generic (); // Generic implementation.
17136 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
17137 This function returns a positive integer if the run-time CPU
17138 supports @var{feature}
17139 and returns @code{0} otherwise. The following features can be detected:
17147 POPCNT instruction.
17155 SSSE3 instructions.
17157 SSE4.1 instructions.
17159 SSE4.2 instructions.
17165 AVX512F instructions.
17168 Here is an example:
17170 if (__builtin_cpu_supports ("popcnt"))
17172 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
17176 count = generic_countbits (n); //generic implementation.
17182 The following built-in functions are made available by @option{-mmmx}.
17183 All of them generate the machine instruction that is part of the name.
17186 v8qi __builtin_ia32_paddb (v8qi, v8qi)
17187 v4hi __builtin_ia32_paddw (v4hi, v4hi)
17188 v2si __builtin_ia32_paddd (v2si, v2si)
17189 v8qi __builtin_ia32_psubb (v8qi, v8qi)
17190 v4hi __builtin_ia32_psubw (v4hi, v4hi)
17191 v2si __builtin_ia32_psubd (v2si, v2si)
17192 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17193 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17194 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17195 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17196 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17197 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17198 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17199 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17200 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17201 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17202 di __builtin_ia32_pand (di, di)
17203 di __builtin_ia32_pandn (di,di)
17204 di __builtin_ia32_por (di, di)
17205 di __builtin_ia32_pxor (di, di)
17206 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17207 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17208 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17209 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17210 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17211 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17212 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17213 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17214 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17215 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17216 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17217 v2si __builtin_ia32_punpckldq (v2si, v2si)
17218 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17219 v4hi __builtin_ia32_packssdw (v2si, v2si)
17220 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17222 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17223 v2si __builtin_ia32_pslld (v2si, v2si)
17224 v1di __builtin_ia32_psllq (v1di, v1di)
17225 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17226 v2si __builtin_ia32_psrld (v2si, v2si)
17227 v1di __builtin_ia32_psrlq (v1di, v1di)
17228 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17229 v2si __builtin_ia32_psrad (v2si, v2si)
17230 v4hi __builtin_ia32_psllwi (v4hi, int)
17231 v2si __builtin_ia32_pslldi (v2si, int)
17232 v1di __builtin_ia32_psllqi (v1di, int)
17233 v4hi __builtin_ia32_psrlwi (v4hi, int)
17234 v2si __builtin_ia32_psrldi (v2si, int)
17235 v1di __builtin_ia32_psrlqi (v1di, int)
17236 v4hi __builtin_ia32_psrawi (v4hi, int)
17237 v2si __builtin_ia32_psradi (v2si, int)
17241 The following built-in functions are made available either with
17242 @option{-msse}, or with a combination of @option{-m3dnow} and
17243 @option{-march=athlon}. All of them generate the machine
17244 instruction that is part of the name.
17247 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17248 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17249 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17250 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17251 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17252 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17253 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17254 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17255 int __builtin_ia32_pmovmskb (v8qi)
17256 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17257 void __builtin_ia32_movntq (di *, di)
17258 void __builtin_ia32_sfence (void)
17261 The following built-in functions are available when @option{-msse} is used.
17262 All of them generate the machine instruction that is part of the name.
17265 int __builtin_ia32_comieq (v4sf, v4sf)
17266 int __builtin_ia32_comineq (v4sf, v4sf)
17267 int __builtin_ia32_comilt (v4sf, v4sf)
17268 int __builtin_ia32_comile (v4sf, v4sf)
17269 int __builtin_ia32_comigt (v4sf, v4sf)
17270 int __builtin_ia32_comige (v4sf, v4sf)
17271 int __builtin_ia32_ucomieq (v4sf, v4sf)
17272 int __builtin_ia32_ucomineq (v4sf, v4sf)
17273 int __builtin_ia32_ucomilt (v4sf, v4sf)
17274 int __builtin_ia32_ucomile (v4sf, v4sf)
17275 int __builtin_ia32_ucomigt (v4sf, v4sf)
17276 int __builtin_ia32_ucomige (v4sf, v4sf)
17277 v4sf __builtin_ia32_addps (v4sf, v4sf)
17278 v4sf __builtin_ia32_subps (v4sf, v4sf)
17279 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17280 v4sf __builtin_ia32_divps (v4sf, v4sf)
17281 v4sf __builtin_ia32_addss (v4sf, v4sf)
17282 v4sf __builtin_ia32_subss (v4sf, v4sf)
17283 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17284 v4sf __builtin_ia32_divss (v4sf, v4sf)
17285 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17286 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17287 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17288 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17289 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17290 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17291 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17292 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17293 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17294 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17295 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17296 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17297 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17298 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17299 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17300 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17301 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17302 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17303 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17304 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17305 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17306 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17307 v4sf __builtin_ia32_minps (v4sf, v4sf)
17308 v4sf __builtin_ia32_minss (v4sf, v4sf)
17309 v4sf __builtin_ia32_andps (v4sf, v4sf)
17310 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17311 v4sf __builtin_ia32_orps (v4sf, v4sf)
17312 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17313 v4sf __builtin_ia32_movss (v4sf, v4sf)
17314 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17315 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17316 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17317 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17318 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17319 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17320 v2si __builtin_ia32_cvtps2pi (v4sf)
17321 int __builtin_ia32_cvtss2si (v4sf)
17322 v2si __builtin_ia32_cvttps2pi (v4sf)
17323 int __builtin_ia32_cvttss2si (v4sf)
17324 v4sf __builtin_ia32_rcpps (v4sf)
17325 v4sf __builtin_ia32_rsqrtps (v4sf)
17326 v4sf __builtin_ia32_sqrtps (v4sf)
17327 v4sf __builtin_ia32_rcpss (v4sf)
17328 v4sf __builtin_ia32_rsqrtss (v4sf)
17329 v4sf __builtin_ia32_sqrtss (v4sf)
17330 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17331 void __builtin_ia32_movntps (float *, v4sf)
17332 int __builtin_ia32_movmskps (v4sf)
17335 The following built-in functions are available when @option{-msse} is used.
17338 @item v4sf __builtin_ia32_loadups (float *)
17339 Generates the @code{movups} machine instruction as a load from memory.
17340 @item void __builtin_ia32_storeups (float *, v4sf)
17341 Generates the @code{movups} machine instruction as a store to memory.
17342 @item v4sf __builtin_ia32_loadss (float *)
17343 Generates the @code{movss} machine instruction as a load from memory.
17344 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17345 Generates the @code{movhps} machine instruction as a load from memory.
17346 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17347 Generates the @code{movlps} machine instruction as a load from memory
17348 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17349 Generates the @code{movhps} machine instruction as a store to memory.
17350 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17351 Generates the @code{movlps} machine instruction as a store to memory.
17354 The following built-in functions are available when @option{-msse2} is used.
17355 All of them generate the machine instruction that is part of the name.
17358 int __builtin_ia32_comisdeq (v2df, v2df)
17359 int __builtin_ia32_comisdlt (v2df, v2df)
17360 int __builtin_ia32_comisdle (v2df, v2df)
17361 int __builtin_ia32_comisdgt (v2df, v2df)
17362 int __builtin_ia32_comisdge (v2df, v2df)
17363 int __builtin_ia32_comisdneq (v2df, v2df)
17364 int __builtin_ia32_ucomisdeq (v2df, v2df)
17365 int __builtin_ia32_ucomisdlt (v2df, v2df)
17366 int __builtin_ia32_ucomisdle (v2df, v2df)
17367 int __builtin_ia32_ucomisdgt (v2df, v2df)
17368 int __builtin_ia32_ucomisdge (v2df, v2df)
17369 int __builtin_ia32_ucomisdneq (v2df, v2df)
17370 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17371 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17372 v2df __builtin_ia32_cmplepd (v2df, v2df)
17373 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17374 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17375 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17376 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17377 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17378 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17379 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17380 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17381 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17382 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17383 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17384 v2df __builtin_ia32_cmplesd (v2df, v2df)
17385 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17386 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17387 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17388 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17389 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17390 v2di __builtin_ia32_paddq (v2di, v2di)
17391 v2di __builtin_ia32_psubq (v2di, v2di)
17392 v2df __builtin_ia32_addpd (v2df, v2df)
17393 v2df __builtin_ia32_subpd (v2df, v2df)
17394 v2df __builtin_ia32_mulpd (v2df, v2df)
17395 v2df __builtin_ia32_divpd (v2df, v2df)
17396 v2df __builtin_ia32_addsd (v2df, v2df)
17397 v2df __builtin_ia32_subsd (v2df, v2df)
17398 v2df __builtin_ia32_mulsd (v2df, v2df)
17399 v2df __builtin_ia32_divsd (v2df, v2df)
17400 v2df __builtin_ia32_minpd (v2df, v2df)
17401 v2df __builtin_ia32_maxpd (v2df, v2df)
17402 v2df __builtin_ia32_minsd (v2df, v2df)
17403 v2df __builtin_ia32_maxsd (v2df, v2df)
17404 v2df __builtin_ia32_andpd (v2df, v2df)
17405 v2df __builtin_ia32_andnpd (v2df, v2df)
17406 v2df __builtin_ia32_orpd (v2df, v2df)
17407 v2df __builtin_ia32_xorpd (v2df, v2df)
17408 v2df __builtin_ia32_movsd (v2df, v2df)
17409 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17410 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17411 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17412 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17413 v4si __builtin_ia32_paddd128 (v4si, v4si)
17414 v2di __builtin_ia32_paddq128 (v2di, v2di)
17415 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17416 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17417 v4si __builtin_ia32_psubd128 (v4si, v4si)
17418 v2di __builtin_ia32_psubq128 (v2di, v2di)
17419 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17420 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17421 v2di __builtin_ia32_pand128 (v2di, v2di)
17422 v2di __builtin_ia32_pandn128 (v2di, v2di)
17423 v2di __builtin_ia32_por128 (v2di, v2di)
17424 v2di __builtin_ia32_pxor128 (v2di, v2di)
17425 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17426 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17427 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17428 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17429 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17430 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17431 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17432 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17433 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17434 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17435 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17436 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17437 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17438 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17439 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17440 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17441 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17442 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17443 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17444 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17445 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17446 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17447 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17448 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17449 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17450 v2df __builtin_ia32_loadupd (double *)
17451 void __builtin_ia32_storeupd (double *, v2df)
17452 v2df __builtin_ia32_loadhpd (v2df, double const *)
17453 v2df __builtin_ia32_loadlpd (v2df, double const *)
17454 int __builtin_ia32_movmskpd (v2df)
17455 int __builtin_ia32_pmovmskb128 (v16qi)
17456 void __builtin_ia32_movnti (int *, int)
17457 void __builtin_ia32_movnti64 (long long int *, long long int)
17458 void __builtin_ia32_movntpd (double *, v2df)
17459 void __builtin_ia32_movntdq (v2df *, v2df)
17460 v4si __builtin_ia32_pshufd (v4si, int)
17461 v8hi __builtin_ia32_pshuflw (v8hi, int)
17462 v8hi __builtin_ia32_pshufhw (v8hi, int)
17463 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17464 v2df __builtin_ia32_sqrtpd (v2df)
17465 v2df __builtin_ia32_sqrtsd (v2df)
17466 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17467 v2df __builtin_ia32_cvtdq2pd (v4si)
17468 v4sf __builtin_ia32_cvtdq2ps (v4si)
17469 v4si __builtin_ia32_cvtpd2dq (v2df)
17470 v2si __builtin_ia32_cvtpd2pi (v2df)
17471 v4sf __builtin_ia32_cvtpd2ps (v2df)
17472 v4si __builtin_ia32_cvttpd2dq (v2df)
17473 v2si __builtin_ia32_cvttpd2pi (v2df)
17474 v2df __builtin_ia32_cvtpi2pd (v2si)
17475 int __builtin_ia32_cvtsd2si (v2df)
17476 int __builtin_ia32_cvttsd2si (v2df)
17477 long long __builtin_ia32_cvtsd2si64 (v2df)
17478 long long __builtin_ia32_cvttsd2si64 (v2df)
17479 v4si __builtin_ia32_cvtps2dq (v4sf)
17480 v2df __builtin_ia32_cvtps2pd (v4sf)
17481 v4si __builtin_ia32_cvttps2dq (v4sf)
17482 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17483 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17484 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17485 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17486 void __builtin_ia32_clflush (const void *)
17487 void __builtin_ia32_lfence (void)
17488 void __builtin_ia32_mfence (void)
17489 v16qi __builtin_ia32_loaddqu (const char *)
17490 void __builtin_ia32_storedqu (char *, v16qi)
17491 v1di __builtin_ia32_pmuludq (v2si, v2si)
17492 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17493 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17494 v4si __builtin_ia32_pslld128 (v4si, v4si)
17495 v2di __builtin_ia32_psllq128 (v2di, v2di)
17496 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17497 v4si __builtin_ia32_psrld128 (v4si, v4si)
17498 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17499 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17500 v4si __builtin_ia32_psrad128 (v4si, v4si)
17501 v2di __builtin_ia32_pslldqi128 (v2di, int)
17502 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17503 v4si __builtin_ia32_pslldi128 (v4si, int)
17504 v2di __builtin_ia32_psllqi128 (v2di, int)
17505 v2di __builtin_ia32_psrldqi128 (v2di, int)
17506 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17507 v4si __builtin_ia32_psrldi128 (v4si, int)
17508 v2di __builtin_ia32_psrlqi128 (v2di, int)
17509 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17510 v4si __builtin_ia32_psradi128 (v4si, int)
17511 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17512 v2di __builtin_ia32_movq128 (v2di)
17515 The following built-in functions are available when @option{-msse3} is used.
17516 All of them generate the machine instruction that is part of the name.
17519 v2df __builtin_ia32_addsubpd (v2df, v2df)
17520 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17521 v2df __builtin_ia32_haddpd (v2df, v2df)
17522 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17523 v2df __builtin_ia32_hsubpd (v2df, v2df)
17524 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17525 v16qi __builtin_ia32_lddqu (char const *)
17526 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17527 v4sf __builtin_ia32_movshdup (v4sf)
17528 v4sf __builtin_ia32_movsldup (v4sf)
17529 void __builtin_ia32_mwait (unsigned int, unsigned int)
17532 The following built-in functions are available when @option{-mssse3} is used.
17533 All of them generate the machine instruction that is part of the name.
17536 v2si __builtin_ia32_phaddd (v2si, v2si)
17537 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17538 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17539 v2si __builtin_ia32_phsubd (v2si, v2si)
17540 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17541 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17542 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17543 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17544 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17545 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17546 v2si __builtin_ia32_psignd (v2si, v2si)
17547 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17548 v1di __builtin_ia32_palignr (v1di, v1di, int)
17549 v8qi __builtin_ia32_pabsb (v8qi)
17550 v2si __builtin_ia32_pabsd (v2si)
17551 v4hi __builtin_ia32_pabsw (v4hi)
17554 The following built-in functions are available when @option{-mssse3} is used.
17555 All of them generate the machine instruction that is part of the name.
17558 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17559 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17560 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17561 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17562 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17563 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17564 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17565 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17566 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17567 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17568 v4si __builtin_ia32_psignd128 (v4si, v4si)
17569 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17570 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17571 v16qi __builtin_ia32_pabsb128 (v16qi)
17572 v4si __builtin_ia32_pabsd128 (v4si)
17573 v8hi __builtin_ia32_pabsw128 (v8hi)
17576 The following built-in functions are available when @option{-msse4.1} is
17577 used. All of them generate the machine instruction that is part of the
17581 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17582 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17583 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17584 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17585 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17586 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17587 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17588 v2di __builtin_ia32_movntdqa (v2di *);
17589 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17590 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17591 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17592 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17593 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17594 v8hi __builtin_ia32_phminposuw128 (v8hi)
17595 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17596 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17597 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17598 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17599 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17600 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17601 v4si __builtin_ia32_pminud128 (v4si, v4si)
17602 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17603 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17604 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17605 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17606 v2di __builtin_ia32_pmovsxdq128 (v4si)
17607 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17608 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17609 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17610 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17611 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17612 v2di __builtin_ia32_pmovzxdq128 (v4si)
17613 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17614 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17615 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17616 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17617 int __builtin_ia32_ptestc128 (v2di, v2di)
17618 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17619 int __builtin_ia32_ptestz128 (v2di, v2di)
17620 v2df __builtin_ia32_roundpd (v2df, const int)
17621 v4sf __builtin_ia32_roundps (v4sf, const int)
17622 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17623 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17626 The following built-in functions are available when @option{-msse4.1} is
17630 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17631 Generates the @code{insertps} machine instruction.
17632 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17633 Generates the @code{pextrb} machine instruction.
17634 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17635 Generates the @code{pinsrb} machine instruction.
17636 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17637 Generates the @code{pinsrd} machine instruction.
17638 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17639 Generates the @code{pinsrq} machine instruction in 64bit mode.
17642 The following built-in functions are changed to generate new SSE4.1
17643 instructions when @option{-msse4.1} is used.
17646 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17647 Generates the @code{extractps} machine instruction.
17648 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17649 Generates the @code{pextrd} machine instruction.
17650 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17651 Generates the @code{pextrq} machine instruction in 64bit mode.
17654 The following built-in functions are available when @option{-msse4.2} is
17655 used. All of them generate the machine instruction that is part of the
17659 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17660 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17661 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17662 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17663 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17664 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17665 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17666 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17667 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17668 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17669 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17670 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17671 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17672 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17673 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17676 The following built-in functions are available when @option{-msse4.2} is
17680 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17681 Generates the @code{crc32b} machine instruction.
17682 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17683 Generates the @code{crc32w} machine instruction.
17684 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17685 Generates the @code{crc32l} machine instruction.
17686 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17687 Generates the @code{crc32q} machine instruction.
17690 The following built-in functions are changed to generate new SSE4.2
17691 instructions when @option{-msse4.2} is used.
17694 @item int __builtin_popcount (unsigned int)
17695 Generates the @code{popcntl} machine instruction.
17696 @item int __builtin_popcountl (unsigned long)
17697 Generates the @code{popcntl} or @code{popcntq} machine instruction,
17698 depending on the size of @code{unsigned long}.
17699 @item int __builtin_popcountll (unsigned long long)
17700 Generates the @code{popcntq} machine instruction.
17703 The following built-in functions are available when @option{-mavx} is
17704 used. All of them generate the machine instruction that is part of the
17708 v4df __builtin_ia32_addpd256 (v4df,v4df)
17709 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
17710 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
17711 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
17712 v4df __builtin_ia32_andnpd256 (v4df,v4df)
17713 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
17714 v4df __builtin_ia32_andpd256 (v4df,v4df)
17715 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
17716 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
17717 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
17718 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
17719 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
17720 v2df __builtin_ia32_cmppd (v2df,v2df,int)
17721 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
17722 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
17723 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
17724 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
17725 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
17726 v4df __builtin_ia32_cvtdq2pd256 (v4si)
17727 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
17728 v4si __builtin_ia32_cvtpd2dq256 (v4df)
17729 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
17730 v8si __builtin_ia32_cvtps2dq256 (v8sf)
17731 v4df __builtin_ia32_cvtps2pd256 (v4sf)
17732 v4si __builtin_ia32_cvttpd2dq256 (v4df)
17733 v8si __builtin_ia32_cvttps2dq256 (v8sf)
17734 v4df __builtin_ia32_divpd256 (v4df,v4df)
17735 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
17736 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
17737 v4df __builtin_ia32_haddpd256 (v4df,v4df)
17738 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
17739 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
17740 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
17741 v32qi __builtin_ia32_lddqu256 (pcchar)
17742 v32qi __builtin_ia32_loaddqu256 (pcchar)
17743 v4df __builtin_ia32_loadupd256 (pcdouble)
17744 v8sf __builtin_ia32_loadups256 (pcfloat)
17745 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
17746 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
17747 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
17748 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
17749 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
17750 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
17751 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
17752 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
17753 v4df __builtin_ia32_maxpd256 (v4df,v4df)
17754 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
17755 v4df __builtin_ia32_minpd256 (v4df,v4df)
17756 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
17757 v4df __builtin_ia32_movddup256 (v4df)
17758 int __builtin_ia32_movmskpd256 (v4df)
17759 int __builtin_ia32_movmskps256 (v8sf)
17760 v8sf __builtin_ia32_movshdup256 (v8sf)
17761 v8sf __builtin_ia32_movsldup256 (v8sf)
17762 v4df __builtin_ia32_mulpd256 (v4df,v4df)
17763 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
17764 v4df __builtin_ia32_orpd256 (v4df,v4df)
17765 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
17766 v2df __builtin_ia32_pd_pd256 (v4df)
17767 v4df __builtin_ia32_pd256_pd (v2df)
17768 v4sf __builtin_ia32_ps_ps256 (v8sf)
17769 v8sf __builtin_ia32_ps256_ps (v4sf)
17770 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
17771 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
17772 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
17773 v8sf __builtin_ia32_rcpps256 (v8sf)
17774 v4df __builtin_ia32_roundpd256 (v4df,int)
17775 v8sf __builtin_ia32_roundps256 (v8sf,int)
17776 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
17777 v8sf __builtin_ia32_rsqrtps256 (v8sf)
17778 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
17779 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
17780 v4si __builtin_ia32_si_si256 (v8si)
17781 v8si __builtin_ia32_si256_si (v4si)
17782 v4df __builtin_ia32_sqrtpd256 (v4df)
17783 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
17784 v8sf __builtin_ia32_sqrtps256 (v8sf)
17785 void __builtin_ia32_storedqu256 (pchar,v32qi)
17786 void __builtin_ia32_storeupd256 (pdouble,v4df)
17787 void __builtin_ia32_storeups256 (pfloat,v8sf)
17788 v4df __builtin_ia32_subpd256 (v4df,v4df)
17789 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
17790 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
17791 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
17792 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
17793 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
17794 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
17795 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
17796 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
17797 v4sf __builtin_ia32_vbroadcastss (pcfloat)
17798 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
17799 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
17800 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
17801 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
17802 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
17803 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
17804 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
17805 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
17806 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
17807 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
17808 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
17809 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
17810 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
17811 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
17812 v2df __builtin_ia32_vpermilpd (v2df,int)
17813 v4df __builtin_ia32_vpermilpd256 (v4df,int)
17814 v4sf __builtin_ia32_vpermilps (v4sf,int)
17815 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
17816 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
17817 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
17818 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
17819 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
17820 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
17821 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
17822 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
17823 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
17824 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
17825 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
17826 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
17827 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
17828 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
17829 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
17830 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
17831 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
17832 void __builtin_ia32_vzeroall (void)
17833 void __builtin_ia32_vzeroupper (void)
17834 v4df __builtin_ia32_xorpd256 (v4df,v4df)
17835 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
17838 The following built-in functions are available when @option{-mavx2} is
17839 used. All of them generate the machine instruction that is part of the
17843 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
17844 v32qi __builtin_ia32_pabsb256 (v32qi)
17845 v16hi __builtin_ia32_pabsw256 (v16hi)
17846 v8si __builtin_ia32_pabsd256 (v8si)
17847 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
17848 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
17849 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
17850 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
17851 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
17852 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
17853 v8si __builtin_ia32_paddd256 (v8si,v8si)
17854 v4di __builtin_ia32_paddq256 (v4di,v4di)
17855 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
17856 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
17857 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
17858 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
17859 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
17860 v4di __builtin_ia32_andsi256 (v4di,v4di)
17861 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
17862 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
17863 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
17864 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
17865 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
17866 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
17867 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
17868 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
17869 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
17870 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
17871 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
17872 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
17873 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
17874 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
17875 v8si __builtin_ia32_phaddd256 (v8si,v8si)
17876 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
17877 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
17878 v8si __builtin_ia32_phsubd256 (v8si,v8si)
17879 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
17880 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
17881 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
17882 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
17883 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
17884 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
17885 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
17886 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
17887 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
17888 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
17889 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
17890 v8si __builtin_ia32_pminsd256 (v8si,v8si)
17891 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
17892 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
17893 v8si __builtin_ia32_pminud256 (v8si,v8si)
17894 int __builtin_ia32_pmovmskb256 (v32qi)
17895 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
17896 v8si __builtin_ia32_pmovsxbd256 (v16qi)
17897 v4di __builtin_ia32_pmovsxbq256 (v16qi)
17898 v8si __builtin_ia32_pmovsxwd256 (v8hi)
17899 v4di __builtin_ia32_pmovsxwq256 (v8hi)
17900 v4di __builtin_ia32_pmovsxdq256 (v4si)
17901 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
17902 v8si __builtin_ia32_pmovzxbd256 (v16qi)
17903 v4di __builtin_ia32_pmovzxbq256 (v16qi)
17904 v8si __builtin_ia32_pmovzxwd256 (v8hi)
17905 v4di __builtin_ia32_pmovzxwq256 (v8hi)
17906 v4di __builtin_ia32_pmovzxdq256 (v4si)
17907 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
17908 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
17909 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
17910 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
17911 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
17912 v8si __builtin_ia32_pmulld256 (v8si,v8si)
17913 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
17914 v4di __builtin_ia32_por256 (v4di,v4di)
17915 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
17916 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
17917 v8si __builtin_ia32_pshufd256 (v8si,int)
17918 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
17919 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
17920 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
17921 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
17922 v8si __builtin_ia32_psignd256 (v8si,v8si)
17923 v4di __builtin_ia32_pslldqi256 (v4di,int)
17924 v16hi __builtin_ia32_psllwi256 (16hi,int)
17925 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
17926 v8si __builtin_ia32_pslldi256 (v8si,int)
17927 v8si __builtin_ia32_pslld256(v8si,v4si)
17928 v4di __builtin_ia32_psllqi256 (v4di,int)
17929 v4di __builtin_ia32_psllq256(v4di,v2di)
17930 v16hi __builtin_ia32_psrawi256 (v16hi,int)
17931 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
17932 v8si __builtin_ia32_psradi256 (v8si,int)
17933 v8si __builtin_ia32_psrad256 (v8si,v4si)
17934 v4di __builtin_ia32_psrldqi256 (v4di, int)
17935 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
17936 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
17937 v8si __builtin_ia32_psrldi256 (v8si,int)
17938 v8si __builtin_ia32_psrld256 (v8si,v4si)
17939 v4di __builtin_ia32_psrlqi256 (v4di,int)
17940 v4di __builtin_ia32_psrlq256(v4di,v2di)
17941 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
17942 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
17943 v8si __builtin_ia32_psubd256 (v8si,v8si)
17944 v4di __builtin_ia32_psubq256 (v4di,v4di)
17945 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
17946 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
17947 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
17948 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
17949 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
17950 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
17951 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
17952 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
17953 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
17954 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
17955 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
17956 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
17957 v4di __builtin_ia32_pxor256 (v4di,v4di)
17958 v4di __builtin_ia32_movntdqa256 (pv4di)
17959 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
17960 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
17961 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
17962 v4di __builtin_ia32_vbroadcastsi256 (v2di)
17963 v4si __builtin_ia32_pblendd128 (v4si,v4si)
17964 v8si __builtin_ia32_pblendd256 (v8si,v8si)
17965 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
17966 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
17967 v8si __builtin_ia32_pbroadcastd256 (v4si)
17968 v4di __builtin_ia32_pbroadcastq256 (v2di)
17969 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
17970 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
17971 v4si __builtin_ia32_pbroadcastd128 (v4si)
17972 v2di __builtin_ia32_pbroadcastq128 (v2di)
17973 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
17974 v4df __builtin_ia32_permdf256 (v4df,int)
17975 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
17976 v4di __builtin_ia32_permdi256 (v4di,int)
17977 v4di __builtin_ia32_permti256 (v4di,v4di,int)
17978 v4di __builtin_ia32_extract128i256 (v4di,int)
17979 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
17980 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
17981 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
17982 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
17983 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
17984 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
17985 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
17986 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
17987 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
17988 v8si __builtin_ia32_psllv8si (v8si,v8si)
17989 v4si __builtin_ia32_psllv4si (v4si,v4si)
17990 v4di __builtin_ia32_psllv4di (v4di,v4di)
17991 v2di __builtin_ia32_psllv2di (v2di,v2di)
17992 v8si __builtin_ia32_psrav8si (v8si,v8si)
17993 v4si __builtin_ia32_psrav4si (v4si,v4si)
17994 v8si __builtin_ia32_psrlv8si (v8si,v8si)
17995 v4si __builtin_ia32_psrlv4si (v4si,v4si)
17996 v4di __builtin_ia32_psrlv4di (v4di,v4di)
17997 v2di __builtin_ia32_psrlv2di (v2di,v2di)
17998 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
17999 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
18000 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
18001 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
18002 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
18003 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
18004 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
18005 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
18006 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
18007 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
18008 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
18009 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
18010 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
18011 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
18012 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
18013 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
18016 The following built-in functions are available when @option{-maes} is
18017 used. All of them generate the machine instruction that is part of the
18021 v2di __builtin_ia32_aesenc128 (v2di, v2di)
18022 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
18023 v2di __builtin_ia32_aesdec128 (v2di, v2di)
18024 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
18025 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
18026 v2di __builtin_ia32_aesimc128 (v2di)
18029 The following built-in function is available when @option{-mpclmul} is
18033 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
18034 Generates the @code{pclmulqdq} machine instruction.
18037 The following built-in function is available when @option{-mfsgsbase} is
18038 used. All of them generate the machine instruction that is part of the
18042 unsigned int __builtin_ia32_rdfsbase32 (void)
18043 unsigned long long __builtin_ia32_rdfsbase64 (void)
18044 unsigned int __builtin_ia32_rdgsbase32 (void)
18045 unsigned long long __builtin_ia32_rdgsbase64 (void)
18046 void _writefsbase_u32 (unsigned int)
18047 void _writefsbase_u64 (unsigned long long)
18048 void _writegsbase_u32 (unsigned int)
18049 void _writegsbase_u64 (unsigned long long)
18052 The following built-in function is available when @option{-mrdrnd} is
18053 used. All of them generate the machine instruction that is part of the
18057 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
18058 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
18059 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
18062 The following built-in functions are available when @option{-msse4a} is used.
18063 All of them generate the machine instruction that is part of the name.
18066 void __builtin_ia32_movntsd (double *, v2df)
18067 void __builtin_ia32_movntss (float *, v4sf)
18068 v2di __builtin_ia32_extrq (v2di, v16qi)
18069 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
18070 v2di __builtin_ia32_insertq (v2di, v2di)
18071 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
18074 The following built-in functions are available when @option{-mxop} is used.
18076 v2df __builtin_ia32_vfrczpd (v2df)
18077 v4sf __builtin_ia32_vfrczps (v4sf)
18078 v2df __builtin_ia32_vfrczsd (v2df)
18079 v4sf __builtin_ia32_vfrczss (v4sf)
18080 v4df __builtin_ia32_vfrczpd256 (v4df)
18081 v8sf __builtin_ia32_vfrczps256 (v8sf)
18082 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
18083 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
18084 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
18085 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
18086 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
18087 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
18088 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
18089 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
18090 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
18091 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
18092 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
18093 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
18094 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
18095 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
18096 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18097 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
18098 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
18099 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
18100 v4si __builtin_ia32_vpcomequd (v4si, v4si)
18101 v2di __builtin_ia32_vpcomequq (v2di, v2di)
18102 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
18103 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18104 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
18105 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
18106 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
18107 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
18108 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
18109 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
18110 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
18111 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
18112 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
18113 v4si __builtin_ia32_vpcomged (v4si, v4si)
18114 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
18115 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
18116 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
18117 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
18118 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
18119 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
18120 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
18121 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
18122 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
18123 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
18124 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
18125 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
18126 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
18127 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
18128 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
18129 v4si __builtin_ia32_vpcomled (v4si, v4si)
18130 v2di __builtin_ia32_vpcomleq (v2di, v2di)
18131 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
18132 v4si __builtin_ia32_vpcomleud (v4si, v4si)
18133 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
18134 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
18135 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
18136 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
18137 v4si __builtin_ia32_vpcomltd (v4si, v4si)
18138 v2di __builtin_ia32_vpcomltq (v2di, v2di)
18139 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
18140 v4si __builtin_ia32_vpcomltud (v4si, v4si)
18141 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
18142 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
18143 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
18144 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
18145 v4si __builtin_ia32_vpcomned (v4si, v4si)
18146 v2di __builtin_ia32_vpcomneq (v2di, v2di)
18147 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
18148 v4si __builtin_ia32_vpcomneud (v4si, v4si)
18149 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
18150 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
18151 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
18152 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
18153 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
18154 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
18155 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
18156 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
18157 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
18158 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
18159 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
18160 v4si __builtin_ia32_vphaddbd (v16qi)
18161 v2di __builtin_ia32_vphaddbq (v16qi)
18162 v8hi __builtin_ia32_vphaddbw (v16qi)
18163 v2di __builtin_ia32_vphadddq (v4si)
18164 v4si __builtin_ia32_vphaddubd (v16qi)
18165 v2di __builtin_ia32_vphaddubq (v16qi)
18166 v8hi __builtin_ia32_vphaddubw (v16qi)
18167 v2di __builtin_ia32_vphaddudq (v4si)
18168 v4si __builtin_ia32_vphadduwd (v8hi)
18169 v2di __builtin_ia32_vphadduwq (v8hi)
18170 v4si __builtin_ia32_vphaddwd (v8hi)
18171 v2di __builtin_ia32_vphaddwq (v8hi)
18172 v8hi __builtin_ia32_vphsubbw (v16qi)
18173 v2di __builtin_ia32_vphsubdq (v4si)
18174 v4si __builtin_ia32_vphsubwd (v8hi)
18175 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
18176 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
18177 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
18178 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
18179 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
18180 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
18181 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
18182 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
18183 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
18184 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
18185 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
18186 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
18187 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
18188 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
18189 v4si __builtin_ia32_vprotd (v4si, v4si)
18190 v2di __builtin_ia32_vprotq (v2di, v2di)
18191 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18192 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18193 v4si __builtin_ia32_vpshad (v4si, v4si)
18194 v2di __builtin_ia32_vpshaq (v2di, v2di)
18195 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18196 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18197 v4si __builtin_ia32_vpshld (v4si, v4si)
18198 v2di __builtin_ia32_vpshlq (v2di, v2di)
18199 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18202 The following built-in functions are available when @option{-mfma4} is used.
18203 All of them generate the machine instruction that is part of the name.
18206 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18207 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18208 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18209 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18210 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18211 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18212 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18213 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18214 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18215 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18216 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18217 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18218 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18219 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18220 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18221 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18222 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18223 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18224 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18225 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18226 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18227 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18228 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18229 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18230 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18231 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18232 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18233 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18234 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18235 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18236 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18237 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18241 The following built-in functions are available when @option{-mlwp} is used.
18244 void __builtin_ia32_llwpcb16 (void *);
18245 void __builtin_ia32_llwpcb32 (void *);
18246 void __builtin_ia32_llwpcb64 (void *);
18247 void * __builtin_ia32_llwpcb16 (void);
18248 void * __builtin_ia32_llwpcb32 (void);
18249 void * __builtin_ia32_llwpcb64 (void);
18250 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18251 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18252 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18253 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18254 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18255 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18258 The following built-in functions are available when @option{-mbmi} is used.
18259 All of them generate the machine instruction that is part of the name.
18261 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18262 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18265 The following built-in functions are available when @option{-mbmi2} is used.
18266 All of them generate the machine instruction that is part of the name.
18268 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18269 unsigned int _pdep_u32 (unsigned int, unsigned int)
18270 unsigned int _pext_u32 (unsigned int, unsigned int)
18271 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18272 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18273 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18276 The following built-in functions are available when @option{-mlzcnt} is used.
18277 All of them generate the machine instruction that is part of the name.
18279 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18280 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18281 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18284 The following built-in functions are available when @option{-mfxsr} is used.
18285 All of them generate the machine instruction that is part of the name.
18287 void __builtin_ia32_fxsave (void *)
18288 void __builtin_ia32_fxrstor (void *)
18289 void __builtin_ia32_fxsave64 (void *)
18290 void __builtin_ia32_fxrstor64 (void *)
18293 The following built-in functions are available when @option{-mxsave} is used.
18294 All of them generate the machine instruction that is part of the name.
18296 void __builtin_ia32_xsave (void *, long long)
18297 void __builtin_ia32_xrstor (void *, long long)
18298 void __builtin_ia32_xsave64 (void *, long long)
18299 void __builtin_ia32_xrstor64 (void *, long long)
18302 The following built-in functions are available when @option{-mxsaveopt} is used.
18303 All of them generate the machine instruction that is part of the name.
18305 void __builtin_ia32_xsaveopt (void *, long long)
18306 void __builtin_ia32_xsaveopt64 (void *, long long)
18309 The following built-in functions are available when @option{-mtbm} is used.
18310 Both of them generate the immediate form of the bextr machine instruction.
18312 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18313 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18317 The following built-in functions are available when @option{-m3dnow} is used.
18318 All of them generate the machine instruction that is part of the name.
18321 void __builtin_ia32_femms (void)
18322 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18323 v2si __builtin_ia32_pf2id (v2sf)
18324 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18325 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18326 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18327 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18328 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18329 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18330 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18331 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18332 v2sf __builtin_ia32_pfrcp (v2sf)
18333 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18334 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18335 v2sf __builtin_ia32_pfrsqrt (v2sf)
18336 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18337 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18338 v2sf __builtin_ia32_pi2fd (v2si)
18339 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18342 The following built-in functions are available when both @option{-m3dnow}
18343 and @option{-march=athlon} are used. All of them generate the machine
18344 instruction that is part of the name.
18347 v2si __builtin_ia32_pf2iw (v2sf)
18348 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18349 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18350 v2sf __builtin_ia32_pi2fw (v2si)
18351 v2sf __builtin_ia32_pswapdsf (v2sf)
18352 v2si __builtin_ia32_pswapdsi (v2si)
18355 The following built-in functions are available when @option{-mrtm} is used
18356 They are used for restricted transactional memory. These are the internal
18357 low level functions. Normally the functions in
18358 @ref{x86 transactional memory intrinsics} should be used instead.
18361 int __builtin_ia32_xbegin ()
18362 void __builtin_ia32_xend ()
18363 void __builtin_ia32_xabort (status)
18364 int __builtin_ia32_xtest ()
18367 The following built-in functions are available when @option{-mmwaitx} is used.
18368 All of them generate the machine instruction that is part of the name.
18370 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18371 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18374 @node x86 transactional memory intrinsics
18375 @subsection x86 Transactional Memory Intrinsics
18377 These hardware transactional memory intrinsics for x86 allow you to use
18378 memory transactions with RTM (Restricted Transactional Memory).
18379 This support is enabled with the @option{-mrtm} option.
18380 For using HLE (Hardware Lock Elision) see
18381 @ref{x86 specific memory model extensions for transactional memory} instead.
18383 A memory transaction commits all changes to memory in an atomic way,
18384 as visible to other threads. If the transaction fails it is rolled back
18385 and all side effects discarded.
18387 Generally there is no guarantee that a memory transaction ever succeeds
18388 and suitable fallback code always needs to be supplied.
18390 @deftypefn {RTM Function} {unsigned} _xbegin ()
18391 Start a RTM (Restricted Transactional Memory) transaction.
18392 Returns @code{_XBEGIN_STARTED} when the transaction
18393 started successfully (note this is not 0, so the constant has to be
18394 explicitly tested).
18396 If the transaction aborts, all side-effects
18397 are undone and an abort code encoded as a bit mask is returned.
18398 The following macros are defined:
18401 @item _XABORT_EXPLICIT
18402 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18403 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18404 @item _XABORT_RETRY
18405 Transaction retry is possible.
18406 @item _XABORT_CONFLICT
18407 Transaction abort due to a memory conflict with another thread.
18408 @item _XABORT_CAPACITY
18409 Transaction abort due to the transaction using too much memory.
18410 @item _XABORT_DEBUG
18411 Transaction abort due to a debug trap.
18412 @item _XABORT_NESTED
18413 Transaction abort in an inner nested transaction.
18416 There is no guarantee
18417 any transaction ever succeeds, so there always needs to be a valid
18421 @deftypefn {RTM Function} {void} _xend ()
18422 Commit the current transaction. When no transaction is active this faults.
18423 All memory side-effects of the transaction become visible
18424 to other threads in an atomic manner.
18427 @deftypefn {RTM Function} {int} _xtest ()
18428 Return a nonzero value if a transaction is currently active, otherwise 0.
18431 @deftypefn {RTM Function} {void} _xabort (status)
18432 Abort the current transaction. When no transaction is active this is a no-op.
18433 The @var{status} is an 8-bit constant; its value is encoded in the return
18434 value from @code{_xbegin}.
18437 Here is an example showing handling for @code{_XABORT_RETRY}
18438 and a fallback path for other failures:
18441 #include <immintrin.h>
18443 int n_tries, max_tries;
18444 unsigned status = _XABORT_EXPLICIT;
18447 for (n_tries = 0; n_tries < max_tries; n_tries++)
18449 status = _xbegin ();
18450 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18453 if (status == _XBEGIN_STARTED)
18455 ... transaction code...
18460 ... non-transactional fallback path...
18465 Note that, in most cases, the transactional and non-transactional code
18466 must synchronize together to ensure consistency.
18468 @node Target Format Checks
18469 @section Format Checks Specific to Particular Target Machines
18471 For some target machines, GCC supports additional options to the
18473 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18476 * Solaris Format Checks::
18477 * Darwin Format Checks::
18480 @node Solaris Format Checks
18481 @subsection Solaris Format Checks
18483 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18484 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18485 conversions, and the two-argument @code{%b} conversion for displaying
18486 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18488 @node Darwin Format Checks
18489 @subsection Darwin Format Checks
18491 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18492 attribute context. Declarations made with such attribution are parsed for correct syntax
18493 and format argument types. However, parsing of the format string itself is currently undefined
18494 and is not carried out by this version of the compiler.
18496 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18497 also be used as format arguments. Note that the relevant headers are only likely to be
18498 available on Darwin (OSX) installations. On such installations, the XCode and system
18499 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18500 associated functions.
18503 @section Pragmas Accepted by GCC
18505 @cindex @code{#pragma}
18507 GCC supports several types of pragmas, primarily in order to compile
18508 code originally written for other compilers. Note that in general
18509 we do not recommend the use of pragmas; @xref{Function Attributes},
18510 for further explanation.
18513 * AArch64 Pragmas::
18517 * RS/6000 and PowerPC Pragmas::
18519 * Solaris Pragmas::
18520 * Symbol-Renaming Pragmas::
18521 * Structure-Layout Pragmas::
18523 * Diagnostic Pragmas::
18524 * Visibility Pragmas::
18525 * Push/Pop Macro Pragmas::
18526 * Function Specific Option Pragmas::
18527 * Loop-Specific Pragmas::
18530 @node AArch64 Pragmas
18531 @subsection AArch64 Pragmas
18533 The pragmas defined by the AArch64 target correspond to the AArch64
18534 target function attributes. They can be specified as below:
18536 #pragma GCC target("string")
18539 where @code{@var{string}} can be any string accepted as an AArch64 target
18540 attribute. @xref{AArch64 Function Attributes}, for more details
18541 on the permissible values of @code{string}.
18544 @subsection ARM Pragmas
18546 The ARM target defines pragmas for controlling the default addition of
18547 @code{long_call} and @code{short_call} attributes to functions.
18548 @xref{Function Attributes}, for information about the effects of these
18553 @cindex pragma, long_calls
18554 Set all subsequent functions to have the @code{long_call} attribute.
18556 @item no_long_calls
18557 @cindex pragma, no_long_calls
18558 Set all subsequent functions to have the @code{short_call} attribute.
18560 @item long_calls_off
18561 @cindex pragma, long_calls_off
18562 Do not affect the @code{long_call} or @code{short_call} attributes of
18563 subsequent functions.
18567 @subsection M32C Pragmas
18570 @item GCC memregs @var{number}
18571 @cindex pragma, memregs
18572 Overrides the command-line option @code{-memregs=} for the current
18573 file. Use with care! This pragma must be before any function in the
18574 file, and mixing different memregs values in different objects may
18575 make them incompatible. This pragma is useful when a
18576 performance-critical function uses a memreg for temporary values,
18577 as it may allow you to reduce the number of memregs used.
18579 @item ADDRESS @var{name} @var{address}
18580 @cindex pragma, address
18581 For any declared symbols matching @var{name}, this does three things
18582 to that symbol: it forces the symbol to be located at the given
18583 address (a number), it forces the symbol to be volatile, and it
18584 changes the symbol's scope to be static. This pragma exists for
18585 compatibility with other compilers, but note that the common
18586 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18590 #pragma ADDRESS port3 0x103
18597 @subsection MeP Pragmas
18601 @item custom io_volatile (on|off)
18602 @cindex pragma, custom io_volatile
18603 Overrides the command-line option @code{-mio-volatile} for the current
18604 file. Note that for compatibility with future GCC releases, this
18605 option should only be used once before any @code{io} variables in each
18608 @item GCC coprocessor available @var{registers}
18609 @cindex pragma, coprocessor available
18610 Specifies which coprocessor registers are available to the register
18611 allocator. @var{registers} may be a single register, register range
18612 separated by ellipses, or comma-separated list of those. Example:
18615 #pragma GCC coprocessor available $c0...$c10, $c28
18618 @item GCC coprocessor call_saved @var{registers}
18619 @cindex pragma, coprocessor call_saved
18620 Specifies which coprocessor registers are to be saved and restored by
18621 any function using them. @var{registers} may be a single register,
18622 register range separated by ellipses, or comma-separated list of
18626 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18629 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18630 @cindex pragma, coprocessor subclass
18631 Creates and defines a register class. These register classes can be
18632 used by inline @code{asm} constructs. @var{registers} may be a single
18633 register, register range separated by ellipses, or comma-separated
18634 list of those. Example:
18637 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18639 asm ("cpfoo %0" : "=B" (x));
18642 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18643 @cindex pragma, disinterrupt
18644 For the named functions, the compiler adds code to disable interrupts
18645 for the duration of those functions. If any functions so named
18646 are not encountered in the source, a warning is emitted that the pragma is
18647 not used. Examples:
18650 #pragma disinterrupt foo
18651 #pragma disinterrupt bar, grill
18652 int foo () @{ @dots{} @}
18655 @item GCC call @var{name} , @var{name} @dots{}
18656 @cindex pragma, call
18657 For the named functions, the compiler always uses a register-indirect
18658 call model when calling the named functions. Examples:
18667 @node RS/6000 and PowerPC Pragmas
18668 @subsection RS/6000 and PowerPC Pragmas
18670 The RS/6000 and PowerPC targets define one pragma for controlling
18671 whether or not the @code{longcall} attribute is added to function
18672 declarations by default. This pragma overrides the @option{-mlongcall}
18673 option, but not the @code{longcall} and @code{shortcall} attributes.
18674 @xref{RS/6000 and PowerPC Options}, for more information about when long
18675 calls are and are not necessary.
18679 @cindex pragma, longcall
18680 Apply the @code{longcall} attribute to all subsequent function
18684 Do not apply the @code{longcall} attribute to subsequent function
18688 @c Describe h8300 pragmas here.
18689 @c Describe sh pragmas here.
18690 @c Describe v850 pragmas here.
18692 @node Darwin Pragmas
18693 @subsection Darwin Pragmas
18695 The following pragmas are available for all architectures running the
18696 Darwin operating system. These are useful for compatibility with other
18700 @item mark @var{tokens}@dots{}
18701 @cindex pragma, mark
18702 This pragma is accepted, but has no effect.
18704 @item options align=@var{alignment}
18705 @cindex pragma, options align
18706 This pragma sets the alignment of fields in structures. The values of
18707 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
18708 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
18709 properly; to restore the previous setting, use @code{reset} for the
18712 @item segment @var{tokens}@dots{}
18713 @cindex pragma, segment
18714 This pragma is accepted, but has no effect.
18716 @item unused (@var{var} [, @var{var}]@dots{})
18717 @cindex pragma, unused
18718 This pragma declares variables to be possibly unused. GCC does not
18719 produce warnings for the listed variables. The effect is similar to
18720 that of the @code{unused} attribute, except that this pragma may appear
18721 anywhere within the variables' scopes.
18724 @node Solaris Pragmas
18725 @subsection Solaris Pragmas
18727 The Solaris target supports @code{#pragma redefine_extname}
18728 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
18729 @code{#pragma} directives for compatibility with the system compiler.
18732 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
18733 @cindex pragma, align
18735 Increase the minimum alignment of each @var{variable} to @var{alignment}.
18736 This is the same as GCC's @code{aligned} attribute @pxref{Variable
18737 Attributes}). Macro expansion occurs on the arguments to this pragma
18738 when compiling C and Objective-C@. It does not currently occur when
18739 compiling C++, but this is a bug which may be fixed in a future
18742 @item fini (@var{function} [, @var{function}]...)
18743 @cindex pragma, fini
18745 This pragma causes each listed @var{function} to be called after
18746 main, or during shared module unloading, by adding a call to the
18747 @code{.fini} section.
18749 @item init (@var{function} [, @var{function}]...)
18750 @cindex pragma, init
18752 This pragma causes each listed @var{function} to be called during
18753 initialization (before @code{main}) or during shared module loading, by
18754 adding a call to the @code{.init} section.
18758 @node Symbol-Renaming Pragmas
18759 @subsection Symbol-Renaming Pragmas
18761 GCC supports a @code{#pragma} directive that changes the name used in
18762 assembly for a given declaration. While this pragma is supported on all
18763 platforms, it is intended primarily to provide compatibility with the
18764 Solaris system headers. This effect can also be achieved using the asm
18765 labels extension (@pxref{Asm Labels}).
18768 @item redefine_extname @var{oldname} @var{newname}
18769 @cindex pragma, redefine_extname
18771 This pragma gives the C function @var{oldname} the assembly symbol
18772 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
18773 is defined if this pragma is available (currently on all platforms).
18776 This pragma and the asm labels extension interact in a complicated
18777 manner. Here are some corner cases you may want to be aware of:
18780 @item This pragma silently applies only to declarations with external
18781 linkage. Asm labels do not have this restriction.
18783 @item In C++, this pragma silently applies only to declarations with
18784 ``C'' linkage. Again, asm labels do not have this restriction.
18786 @item If either of the ways of changing the assembly name of a
18787 declaration are applied to a declaration whose assembly name has
18788 already been determined (either by a previous use of one of these
18789 features, or because the compiler needed the assembly name in order to
18790 generate code), and the new name is different, a warning issues and
18791 the name does not change.
18793 @item The @var{oldname} used by @code{#pragma redefine_extname} is
18794 always the C-language name.
18797 @node Structure-Layout Pragmas
18798 @subsection Structure-Layout Pragmas
18800 For compatibility with Microsoft Windows compilers, GCC supports a
18801 set of @code{#pragma} directives that change the maximum alignment of
18802 members of structures (other than zero-width bit-fields), unions, and
18803 classes subsequently defined. The @var{n} value below always is required
18804 to be a small power of two and specifies the new alignment in bytes.
18807 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
18808 @item @code{#pragma pack()} sets the alignment to the one that was in
18809 effect when compilation started (see also command-line option
18810 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
18811 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
18812 setting on an internal stack and then optionally sets the new alignment.
18813 @item @code{#pragma pack(pop)} restores the alignment setting to the one
18814 saved at the top of the internal stack (and removes that stack entry).
18815 Note that @code{#pragma pack([@var{n}])} does not influence this internal
18816 stack; thus it is possible to have @code{#pragma pack(push)} followed by
18817 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
18818 @code{#pragma pack(pop)}.
18821 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
18822 directive which lays out structures and unions subsequently defined as the
18823 documented @code{__attribute__ ((ms_struct))}.
18826 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
18827 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
18828 @item @code{#pragma ms_struct reset} goes back to the default layout.
18831 Most targets also support the @code{#pragma scalar_storage_order} directive
18832 which lays out structures and unions subsequently defined as the documented
18833 @code{__attribute__ ((scalar_storage_order))}.
18836 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
18837 of the scalar fields to big-endian.
18838 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
18839 of the scalar fields to little-endian.
18840 @item @code{#pragma scalar_storage_order default} goes back to the endianness
18841 that was in effect when compilation started (see also command-line option
18842 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
18846 @subsection Weak Pragmas
18848 For compatibility with SVR4, GCC supports a set of @code{#pragma}
18849 directives for declaring symbols to be weak, and defining weak
18853 @item #pragma weak @var{symbol}
18854 @cindex pragma, weak
18855 This pragma declares @var{symbol} to be weak, as if the declaration
18856 had the attribute of the same name. The pragma may appear before
18857 or after the declaration of @var{symbol}. It is not an error for
18858 @var{symbol} to never be defined at all.
18860 @item #pragma weak @var{symbol1} = @var{symbol2}
18861 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
18862 It is an error if @var{symbol2} is not defined in the current
18866 @node Diagnostic Pragmas
18867 @subsection Diagnostic Pragmas
18869 GCC allows the user to selectively enable or disable certain types of
18870 diagnostics, and change the kind of the diagnostic. For example, a
18871 project's policy might require that all sources compile with
18872 @option{-Werror} but certain files might have exceptions allowing
18873 specific types of warnings. Or, a project might selectively enable
18874 diagnostics and treat them as errors depending on which preprocessor
18875 macros are defined.
18878 @item #pragma GCC diagnostic @var{kind} @var{option}
18879 @cindex pragma, diagnostic
18881 Modifies the disposition of a diagnostic. Note that not all
18882 diagnostics are modifiable; at the moment only warnings (normally
18883 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
18884 Use @option{-fdiagnostics-show-option} to determine which diagnostics
18885 are controllable and which option controls them.
18887 @var{kind} is @samp{error} to treat this diagnostic as an error,
18888 @samp{warning} to treat it like a warning (even if @option{-Werror} is
18889 in effect), or @samp{ignored} if the diagnostic is to be ignored.
18890 @var{option} is a double quoted string that matches the command-line
18894 #pragma GCC diagnostic warning "-Wformat"
18895 #pragma GCC diagnostic error "-Wformat"
18896 #pragma GCC diagnostic ignored "-Wformat"
18899 Note that these pragmas override any command-line options. GCC keeps
18900 track of the location of each pragma, and issues diagnostics according
18901 to the state as of that point in the source file. Thus, pragmas occurring
18902 after a line do not affect diagnostics caused by that line.
18904 @item #pragma GCC diagnostic push
18905 @itemx #pragma GCC diagnostic pop
18907 Causes GCC to remember the state of the diagnostics as of each
18908 @code{push}, and restore to that point at each @code{pop}. If a
18909 @code{pop} has no matching @code{push}, the command-line options are
18913 #pragma GCC diagnostic error "-Wuninitialized"
18914 foo(a); /* error is given for this one */
18915 #pragma GCC diagnostic push
18916 #pragma GCC diagnostic ignored "-Wuninitialized"
18917 foo(b); /* no diagnostic for this one */
18918 #pragma GCC diagnostic pop
18919 foo(c); /* error is given for this one */
18920 #pragma GCC diagnostic pop
18921 foo(d); /* depends on command-line options */
18926 GCC also offers a simple mechanism for printing messages during
18930 @item #pragma message @var{string}
18931 @cindex pragma, diagnostic
18933 Prints @var{string} as a compiler message on compilation. The message
18934 is informational only, and is neither a compilation warning nor an error.
18937 #pragma message "Compiling " __FILE__ "..."
18940 @var{string} may be parenthesized, and is printed with location
18941 information. For example,
18944 #define DO_PRAGMA(x) _Pragma (#x)
18945 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
18947 TODO(Remember to fix this)
18951 prints @samp{/tmp/file.c:4: note: #pragma message:
18952 TODO - Remember to fix this}.
18956 @node Visibility Pragmas
18957 @subsection Visibility Pragmas
18960 @item #pragma GCC visibility push(@var{visibility})
18961 @itemx #pragma GCC visibility pop
18962 @cindex pragma, visibility
18964 This pragma allows the user to set the visibility for multiple
18965 declarations without having to give each a visibility attribute
18966 (@pxref{Function Attributes}).
18968 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
18969 declarations. Class members and template specializations are not
18970 affected; if you want to override the visibility for a particular
18971 member or instantiation, you must use an attribute.
18976 @node Push/Pop Macro Pragmas
18977 @subsection Push/Pop Macro Pragmas
18979 For compatibility with Microsoft Windows compilers, GCC supports
18980 @samp{#pragma push_macro(@var{"macro_name"})}
18981 and @samp{#pragma pop_macro(@var{"macro_name"})}.
18984 @item #pragma push_macro(@var{"macro_name"})
18985 @cindex pragma, push_macro
18986 This pragma saves the value of the macro named as @var{macro_name} to
18987 the top of the stack for this macro.
18989 @item #pragma pop_macro(@var{"macro_name"})
18990 @cindex pragma, pop_macro
18991 This pragma sets the value of the macro named as @var{macro_name} to
18992 the value on top of the stack for this macro. If the stack for
18993 @var{macro_name} is empty, the value of the macro remains unchanged.
19000 #pragma push_macro("X")
19003 #pragma pop_macro("X")
19008 In this example, the definition of X as 1 is saved by @code{#pragma
19009 push_macro} and restored by @code{#pragma pop_macro}.
19011 @node Function Specific Option Pragmas
19012 @subsection Function Specific Option Pragmas
19015 @item #pragma GCC target (@var{"string"}...)
19016 @cindex pragma GCC target
19018 This pragma allows you to set target specific options for functions
19019 defined later in the source file. One or more strings can be
19020 specified. Each function that is defined after this point is as
19021 if @code{attribute((target("STRING")))} was specified for that
19022 function. The parenthesis around the options is optional.
19023 @xref{Function Attributes}, for more information about the
19024 @code{target} attribute and the attribute syntax.
19026 The @code{#pragma GCC target} pragma is presently implemented for
19027 x86, PowerPC, and Nios II targets only.
19031 @item #pragma GCC optimize (@var{"string"}...)
19032 @cindex pragma GCC optimize
19034 This pragma allows you to set global optimization 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((optimize("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{optimize} attribute and the attribute syntax.
19044 @item #pragma GCC push_options
19045 @itemx #pragma GCC pop_options
19046 @cindex pragma GCC push_options
19047 @cindex pragma GCC pop_options
19049 These pragmas maintain a stack of the current target and optimization
19050 options. It is intended for include files where you temporarily want
19051 to switch to using a different @samp{#pragma GCC target} or
19052 @samp{#pragma GCC optimize} and then to pop back to the previous
19057 @item #pragma GCC reset_options
19058 @cindex pragma GCC reset_options
19060 This pragma clears the current @code{#pragma GCC target} and
19061 @code{#pragma GCC optimize} to use the default switches as specified
19062 on the command line.
19065 @node Loop-Specific Pragmas
19066 @subsection Loop-Specific Pragmas
19069 @item #pragma GCC ivdep
19070 @cindex pragma GCC ivdep
19073 With this pragma, the programmer asserts that there are no loop-carried
19074 dependencies which would prevent consecutive iterations of
19075 the following loop from executing concurrently with SIMD
19076 (single instruction multiple data) instructions.
19078 For example, the compiler can only unconditionally vectorize the following
19079 loop with the pragma:
19082 void foo (int n, int *a, int *b, int *c)
19086 for (i = 0; i < n; ++i)
19087 a[i] = b[i] + c[i];
19092 In this example, using the @code{restrict} qualifier had the same
19093 effect. In the following example, that would not be possible. Assume
19094 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
19095 that it can unconditionally vectorize the following loop:
19098 void ignore_vec_dep (int *a, int k, int c, int m)
19101 for (int i = 0; i < m; i++)
19102 a[i] = a[i + k] * c;
19107 @node Unnamed Fields
19108 @section Unnamed Structure and Union Fields
19109 @cindex @code{struct}
19110 @cindex @code{union}
19112 As permitted by ISO C11 and for compatibility with other compilers,
19113 GCC allows you to define
19114 a structure or union that contains, as fields, structures and unions
19115 without names. For example:
19129 In this example, you are able to access members of the unnamed
19130 union with code like @samp{foo.b}. Note that only unnamed structs and
19131 unions are allowed, you may not have, for example, an unnamed
19134 You must never create such structures that cause ambiguous field definitions.
19135 For example, in this structure:
19147 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
19148 The compiler gives errors for such constructs.
19150 @opindex fms-extensions
19151 Unless @option{-fms-extensions} is used, the unnamed field must be a
19152 structure or union definition without a tag (for example, @samp{struct
19153 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
19154 also be a definition with a tag such as @samp{struct foo @{ int a;
19155 @};}, a reference to a previously defined structure or union such as
19156 @samp{struct foo;}, or a reference to a @code{typedef} name for a
19157 previously defined structure or union type.
19159 @opindex fplan9-extensions
19160 The option @option{-fplan9-extensions} enables
19161 @option{-fms-extensions} as well as two other extensions. First, a
19162 pointer to a structure is automatically converted to a pointer to an
19163 anonymous field for assignments and function calls. For example:
19166 struct s1 @{ int a; @};
19167 struct s2 @{ struct s1; @};
19168 extern void f1 (struct s1 *);
19169 void f2 (struct s2 *p) @{ f1 (p); @}
19173 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
19174 converted into a pointer to the anonymous field.
19176 Second, when the type of an anonymous field is a @code{typedef} for a
19177 @code{struct} or @code{union}, code may refer to the field using the
19178 name of the @code{typedef}.
19181 typedef struct @{ int a; @} s1;
19182 struct s2 @{ s1; @};
19183 s1 f1 (struct s2 *p) @{ return p->s1; @}
19186 These usages are only permitted when they are not ambiguous.
19189 @section Thread-Local Storage
19190 @cindex Thread-Local Storage
19191 @cindex @acronym{TLS}
19192 @cindex @code{__thread}
19194 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
19195 are allocated such that there is one instance of the variable per extant
19196 thread. The runtime model GCC uses to implement this originates
19197 in the IA-64 processor-specific ABI, but has since been migrated
19198 to other processors as well. It requires significant support from
19199 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
19200 system libraries (@file{libc.so} and @file{libpthread.so}), so it
19201 is not available everywhere.
19203 At the user level, the extension is visible with a new storage
19204 class keyword: @code{__thread}. For example:
19208 extern __thread struct state s;
19209 static __thread char *p;
19212 The @code{__thread} specifier may be used alone, with the @code{extern}
19213 or @code{static} specifiers, but with no other storage class specifier.
19214 When used with @code{extern} or @code{static}, @code{__thread} must appear
19215 immediately after the other storage class specifier.
19217 The @code{__thread} specifier may be applied to any global, file-scoped
19218 static, function-scoped static, or static data member of a class. It may
19219 not be applied to block-scoped automatic or non-static data member.
19221 When the address-of operator is applied to a thread-local variable, it is
19222 evaluated at run time and returns the address of the current thread's
19223 instance of that variable. An address so obtained may be used by any
19224 thread. When a thread terminates, any pointers to thread-local variables
19225 in that thread become invalid.
19227 No static initialization may refer to the address of a thread-local variable.
19229 In C++, if an initializer is present for a thread-local variable, it must
19230 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19233 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19234 ELF Handling For Thread-Local Storage} for a detailed explanation of
19235 the four thread-local storage addressing models, and how the runtime
19236 is expected to function.
19239 * C99 Thread-Local Edits::
19240 * C++98 Thread-Local Edits::
19243 @node C99 Thread-Local Edits
19244 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19246 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19247 that document the exact semantics of the language extension.
19251 @cite{5.1.2 Execution environments}
19253 Add new text after paragraph 1
19256 Within either execution environment, a @dfn{thread} is a flow of
19257 control within a program. It is implementation defined whether
19258 or not there may be more than one thread associated with a program.
19259 It is implementation defined how threads beyond the first are
19260 created, the name and type of the function called at thread
19261 startup, and how threads may be terminated. However, objects
19262 with thread storage duration shall be initialized before thread
19267 @cite{6.2.4 Storage durations of objects}
19269 Add new text before paragraph 3
19272 An object whose identifier is declared with the storage-class
19273 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19274 Its lifetime is the entire execution of the thread, and its
19275 stored value is initialized only once, prior to thread startup.
19279 @cite{6.4.1 Keywords}
19281 Add @code{__thread}.
19284 @cite{6.7.1 Storage-class specifiers}
19286 Add @code{__thread} to the list of storage class specifiers in
19289 Change paragraph 2 to
19292 With the exception of @code{__thread}, at most one storage-class
19293 specifier may be given [@dots{}]. The @code{__thread} specifier may
19294 be used alone, or immediately following @code{extern} or
19298 Add new text after paragraph 6
19301 The declaration of an identifier for a variable that has
19302 block scope that specifies @code{__thread} shall also
19303 specify either @code{extern} or @code{static}.
19305 The @code{__thread} specifier shall be used only with
19310 @node C++98 Thread-Local Edits
19311 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19313 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19314 that document the exact semantics of the language extension.
19318 @b{[intro.execution]}
19320 New text after paragraph 4
19323 A @dfn{thread} is a flow of control within the abstract machine.
19324 It is implementation defined whether or not there may be more than
19328 New text after paragraph 7
19331 It is unspecified whether additional action must be taken to
19332 ensure when and whether side effects are visible to other threads.
19338 Add @code{__thread}.
19341 @b{[basic.start.main]}
19343 Add after paragraph 5
19346 The thread that begins execution at the @code{main} function is called
19347 the @dfn{main thread}. It is implementation defined how functions
19348 beginning threads other than the main thread are designated or typed.
19349 A function so designated, as well as the @code{main} function, is called
19350 a @dfn{thread startup function}. It is implementation defined what
19351 happens if a thread startup function returns. It is implementation
19352 defined what happens to other threads when any thread calls @code{exit}.
19356 @b{[basic.start.init]}
19358 Add after paragraph 4
19361 The storage for an object of thread storage duration shall be
19362 statically initialized before the first statement of the thread startup
19363 function. An object of thread storage duration shall not require
19364 dynamic initialization.
19368 @b{[basic.start.term]}
19370 Add after paragraph 3
19373 The type of an object with thread storage duration shall not have a
19374 non-trivial destructor, nor shall it be an array type whose elements
19375 (directly or indirectly) have non-trivial destructors.
19381 Add ``thread storage duration'' to the list in paragraph 1.
19386 Thread, static, and automatic storage durations are associated with
19387 objects introduced by declarations [@dots{}].
19390 Add @code{__thread} to the list of specifiers in paragraph 3.
19393 @b{[basic.stc.thread]}
19395 New section before @b{[basic.stc.static]}
19398 The keyword @code{__thread} applied to a non-local object gives the
19399 object thread storage duration.
19401 A local variable or class data member declared both @code{static}
19402 and @code{__thread} gives the variable or member thread storage
19407 @b{[basic.stc.static]}
19412 All objects that have neither thread storage duration, dynamic
19413 storage duration nor are local [@dots{}].
19419 Add @code{__thread} to the list in paragraph 1.
19424 With the exception of @code{__thread}, at most one
19425 @var{storage-class-specifier} shall appear in a given
19426 @var{decl-specifier-seq}. The @code{__thread} specifier may
19427 be used alone, or immediately following the @code{extern} or
19428 @code{static} specifiers. [@dots{}]
19431 Add after paragraph 5
19434 The @code{__thread} specifier can be applied only to the names of objects
19435 and to anonymous unions.
19441 Add after paragraph 6
19444 Non-@code{static} members shall not be @code{__thread}.
19448 @node Binary constants
19449 @section Binary Constants using the @samp{0b} Prefix
19450 @cindex Binary constants using the @samp{0b} prefix
19452 Integer constants can be written as binary constants, consisting of a
19453 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19454 @samp{0B}. This is particularly useful in environments that operate a
19455 lot on the bit level (like microcontrollers).
19457 The following statements are identical:
19466 The type of these constants follows the same rules as for octal or
19467 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19470 @node C++ Extensions
19471 @chapter Extensions to the C++ Language
19472 @cindex extensions, C++ language
19473 @cindex C++ language extensions
19475 The GNU compiler provides these extensions to the C++ language (and you
19476 can also use most of the C language extensions in your C++ programs). If you
19477 want to write code that checks whether these features are available, you can
19478 test for the GNU compiler the same way as for C programs: check for a
19479 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19480 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19481 Predefined Macros,cpp,The GNU C Preprocessor}).
19484 * C++ Volatiles:: What constitutes an access to a volatile object.
19485 * Restricted Pointers:: C99 restricted pointers and references.
19486 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19487 * C++ Interface:: You can use a single C++ header file for both
19488 declarations and definitions.
19489 * Template Instantiation:: Methods for ensuring that exactly one copy of
19490 each needed template instantiation is emitted.
19491 * Bound member functions:: You can extract a function pointer to the
19492 method denoted by a @samp{->*} or @samp{.*} expression.
19493 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19494 * Function Multiversioning:: Declaring multiple function versions.
19495 * Namespace Association:: Strong using-directives for namespace association.
19496 * Type Traits:: Compiler support for type traits.
19497 * C++ Concepts:: Improved support for generic programming.
19498 * Java Exceptions:: Tweaking exception handling to work with Java.
19499 * Deprecated Features:: Things will disappear from G++.
19500 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19503 @node C++ Volatiles
19504 @section When is a Volatile C++ Object Accessed?
19505 @cindex accessing volatiles
19506 @cindex volatile read
19507 @cindex volatile write
19508 @cindex volatile access
19510 The C++ standard differs from the C standard in its treatment of
19511 volatile objects. It fails to specify what constitutes a volatile
19512 access, except to say that C++ should behave in a similar manner to C
19513 with respect to volatiles, where possible. However, the different
19514 lvalueness of expressions between C and C++ complicate the behavior.
19515 G++ behaves the same as GCC for volatile access, @xref{C
19516 Extensions,,Volatiles}, for a description of GCC's behavior.
19518 The C and C++ language specifications differ when an object is
19519 accessed in a void context:
19522 volatile int *src = @var{somevalue};
19526 The C++ standard specifies that such expressions do not undergo lvalue
19527 to rvalue conversion, and that the type of the dereferenced object may
19528 be incomplete. The C++ standard does not specify explicitly that it
19529 is lvalue to rvalue conversion that is responsible for causing an
19530 access. There is reason to believe that it is, because otherwise
19531 certain simple expressions become undefined. However, because it
19532 would surprise most programmers, G++ treats dereferencing a pointer to
19533 volatile object of complete type as GCC would do for an equivalent
19534 type in C@. When the object has incomplete type, G++ issues a
19535 warning; if you wish to force an error, you must force a conversion to
19536 rvalue with, for instance, a static cast.
19538 When using a reference to volatile, G++ does not treat equivalent
19539 expressions as accesses to volatiles, but instead issues a warning that
19540 no volatile is accessed. The rationale for this is that otherwise it
19541 becomes difficult to determine where volatile access occur, and not
19542 possible to ignore the return value from functions returning volatile
19543 references. Again, if you wish to force a read, cast the reference to
19546 G++ implements the same behavior as GCC does when assigning to a
19547 volatile object---there is no reread of the assigned-to object, the
19548 assigned rvalue is reused. Note that in C++ assignment expressions
19549 are lvalues, and if used as an lvalue, the volatile object is
19550 referred to. For instance, @var{vref} refers to @var{vobj}, as
19551 expected, in the following example:
19555 volatile int &vref = vobj = @var{something};
19558 @node Restricted Pointers
19559 @section Restricting Pointer Aliasing
19560 @cindex restricted pointers
19561 @cindex restricted references
19562 @cindex restricted this pointer
19564 As with the C front end, G++ understands the C99 feature of restricted pointers,
19565 specified with the @code{__restrict__}, or @code{__restrict} type
19566 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19567 language flag, @code{restrict} is not a keyword in C++.
19569 In addition to allowing restricted pointers, you can specify restricted
19570 references, which indicate that the reference is not aliased in the local
19574 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19581 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19582 @var{rref} refers to a (different) unaliased integer.
19584 You may also specify whether a member function's @var{this} pointer is
19585 unaliased by using @code{__restrict__} as a member function qualifier.
19588 void T::fn () __restrict__
19595 Within the body of @code{T::fn}, @var{this} has the effective
19596 definition @code{T *__restrict__ const this}. Notice that the
19597 interpretation of a @code{__restrict__} member function qualifier is
19598 different to that of @code{const} or @code{volatile} qualifier, in that it
19599 is applied to the pointer rather than the object. This is consistent with
19600 other compilers that implement restricted pointers.
19602 As with all outermost parameter qualifiers, @code{__restrict__} is
19603 ignored in function definition matching. This means you only need to
19604 specify @code{__restrict__} in a function definition, rather than
19605 in a function prototype as well.
19607 @node Vague Linkage
19608 @section Vague Linkage
19609 @cindex vague linkage
19611 There are several constructs in C++ that require space in the object
19612 file but are not clearly tied to a single translation unit. We say that
19613 these constructs have ``vague linkage''. Typically such constructs are
19614 emitted wherever they are needed, though sometimes we can be more
19618 @item Inline Functions
19619 Inline functions are typically defined in a header file which can be
19620 included in many different compilations. Hopefully they can usually be
19621 inlined, but sometimes an out-of-line copy is necessary, if the address
19622 of the function is taken or if inlining fails. In general, we emit an
19623 out-of-line copy in all translation units where one is needed. As an
19624 exception, we only emit inline virtual functions with the vtable, since
19625 it always requires a copy.
19627 Local static variables and string constants used in an inline function
19628 are also considered to have vague linkage, since they must be shared
19629 between all inlined and out-of-line instances of the function.
19633 C++ virtual functions are implemented in most compilers using a lookup
19634 table, known as a vtable. The vtable contains pointers to the virtual
19635 functions provided by a class, and each object of the class contains a
19636 pointer to its vtable (or vtables, in some multiple-inheritance
19637 situations). If the class declares any non-inline, non-pure virtual
19638 functions, the first one is chosen as the ``key method'' for the class,
19639 and the vtable is only emitted in the translation unit where the key
19642 @emph{Note:} If the chosen key method is later defined as inline, the
19643 vtable is still emitted in every translation unit that defines it.
19644 Make sure that any inline virtuals are declared inline in the class
19645 body, even if they are not defined there.
19647 @item @code{type_info} objects
19648 @cindex @code{type_info}
19650 C++ requires information about types to be written out in order to
19651 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19652 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19653 object is written out along with the vtable so that @samp{dynamic_cast}
19654 can determine the dynamic type of a class object at run time. For all
19655 other types, we write out the @samp{type_info} object when it is used: when
19656 applying @samp{typeid} to an expression, throwing an object, or
19657 referring to a type in a catch clause or exception specification.
19659 @item Template Instantiations
19660 Most everything in this section also applies to template instantiations,
19661 but there are other options as well.
19662 @xref{Template Instantiation,,Where's the Template?}.
19666 When used with GNU ld version 2.8 or later on an ELF system such as
19667 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
19668 these constructs will be discarded at link time. This is known as
19671 On targets that don't support COMDAT, but do support weak symbols, GCC
19672 uses them. This way one copy overrides all the others, but
19673 the unused copies still take up space in the executable.
19675 For targets that do not support either COMDAT or weak symbols,
19676 most entities with vague linkage are emitted as local symbols to
19677 avoid duplicate definition errors from the linker. This does not happen
19678 for local statics in inlines, however, as having multiple copies
19679 almost certainly breaks things.
19681 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
19682 another way to control placement of these constructs.
19684 @node C++ Interface
19685 @section C++ Interface and Implementation Pragmas
19687 @cindex interface and implementation headers, C++
19688 @cindex C++ interface and implementation headers
19689 @cindex pragmas, interface and implementation
19691 @code{#pragma interface} and @code{#pragma implementation} provide the
19692 user with a way of explicitly directing the compiler to emit entities
19693 with vague linkage (and debugging information) in a particular
19696 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
19697 by COMDAT support and the ``key method'' heuristic
19698 mentioned in @ref{Vague Linkage}. Using them can actually cause your
19699 program to grow due to unnecessary out-of-line copies of inline
19703 @item #pragma interface
19704 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
19705 @kindex #pragma interface
19706 Use this directive in @emph{header files} that define object classes, to save
19707 space in most of the object files that use those classes. Normally,
19708 local copies of certain information (backup copies of inline member
19709 functions, debugging information, and the internal tables that implement
19710 virtual functions) must be kept in each object file that includes class
19711 definitions. You can use this pragma to avoid such duplication. When a
19712 header file containing @samp{#pragma interface} is included in a
19713 compilation, this auxiliary information is not generated (unless
19714 the main input source file itself uses @samp{#pragma implementation}).
19715 Instead, the object files contain references to be resolved at link
19718 The second form of this directive is useful for the case where you have
19719 multiple headers with the same name in different directories. If you
19720 use this form, you must specify the same string to @samp{#pragma
19723 @item #pragma implementation
19724 @itemx #pragma implementation "@var{objects}.h"
19725 @kindex #pragma implementation
19726 Use this pragma in a @emph{main input file}, when you want full output from
19727 included header files to be generated (and made globally visible). The
19728 included header file, in turn, should use @samp{#pragma interface}.
19729 Backup copies of inline member functions, debugging information, and the
19730 internal tables used to implement virtual functions are all generated in
19731 implementation files.
19733 @cindex implied @code{#pragma implementation}
19734 @cindex @code{#pragma implementation}, implied
19735 @cindex naming convention, implementation headers
19736 If you use @samp{#pragma implementation} with no argument, it applies to
19737 an include file with the same basename@footnote{A file's @dfn{basename}
19738 is the name stripped of all leading path information and of trailing
19739 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
19740 file. For example, in @file{allclass.cc}, giving just
19741 @samp{#pragma implementation}
19742 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
19744 Use the string argument if you want a single implementation file to
19745 include code from multiple header files. (You must also use
19746 @samp{#include} to include the header file; @samp{#pragma
19747 implementation} only specifies how to use the file---it doesn't actually
19750 There is no way to split up the contents of a single header file into
19751 multiple implementation files.
19754 @cindex inlining and C++ pragmas
19755 @cindex C++ pragmas, effect on inlining
19756 @cindex pragmas in C++, effect on inlining
19757 @samp{#pragma implementation} and @samp{#pragma interface} also have an
19758 effect on function inlining.
19760 If you define a class in a header file marked with @samp{#pragma
19761 interface}, the effect on an inline function defined in that class is
19762 similar to an explicit @code{extern} declaration---the compiler emits
19763 no code at all to define an independent version of the function. Its
19764 definition is used only for inlining with its callers.
19766 @opindex fno-implement-inlines
19767 Conversely, when you include the same header file in a main source file
19768 that declares it as @samp{#pragma implementation}, the compiler emits
19769 code for the function itself; this defines a version of the function
19770 that can be found via pointers (or by callers compiled without
19771 inlining). If all calls to the function can be inlined, you can avoid
19772 emitting the function by compiling with @option{-fno-implement-inlines}.
19773 If any calls are not inlined, you will get linker errors.
19775 @node Template Instantiation
19776 @section Where's the Template?
19777 @cindex template instantiation
19779 C++ templates were the first language feature to require more
19780 intelligence from the environment than was traditionally found on a UNIX
19781 system. Somehow the compiler and linker have to make sure that each
19782 template instance occurs exactly once in the executable if it is needed,
19783 and not at all otherwise. There are two basic approaches to this
19784 problem, which are referred to as the Borland model and the Cfront model.
19787 @item Borland model
19788 Borland C++ solved the template instantiation problem by adding the code
19789 equivalent of common blocks to their linker; the compiler emits template
19790 instances in each translation unit that uses them, and the linker
19791 collapses them together. The advantage of this model is that the linker
19792 only has to consider the object files themselves; there is no external
19793 complexity to worry about. The disadvantage is that compilation time
19794 is increased because the template code is being compiled repeatedly.
19795 Code written for this model tends to include definitions of all
19796 templates in the header file, since they must be seen to be
19800 The AT&T C++ translator, Cfront, solved the template instantiation
19801 problem by creating the notion of a template repository, an
19802 automatically maintained place where template instances are stored. A
19803 more modern version of the repository works as follows: As individual
19804 object files are built, the compiler places any template definitions and
19805 instantiations encountered in the repository. At link time, the link
19806 wrapper adds in the objects in the repository and compiles any needed
19807 instances that were not previously emitted. The advantages of this
19808 model are more optimal compilation speed and the ability to use the
19809 system linker; to implement the Borland model a compiler vendor also
19810 needs to replace the linker. The disadvantages are vastly increased
19811 complexity, and thus potential for error; for some code this can be
19812 just as transparent, but in practice it can been very difficult to build
19813 multiple programs in one directory and one program in multiple
19814 directories. Code written for this model tends to separate definitions
19815 of non-inline member templates into a separate file, which should be
19816 compiled separately.
19819 G++ implements the Borland model on targets where the linker supports it,
19820 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
19821 Otherwise G++ implements neither automatic model.
19823 You have the following options for dealing with template instantiations:
19827 Do nothing. Code written for the Borland model works fine, but
19828 each translation unit contains instances of each of the templates it
19829 uses. The duplicate instances will be discarded by the linker, but in
19830 a large program, this can lead to an unacceptable amount of code
19831 duplication in object files or shared libraries.
19833 Duplicate instances of a template can be avoided by defining an explicit
19834 instantiation in one object file, and preventing the compiler from doing
19835 implicit instantiations in any other object files by using an explicit
19836 instantiation declaration, using the @code{extern template} syntax:
19839 extern template int max (int, int);
19842 This syntax is defined in the C++ 2011 standard, but has been supported by
19843 G++ and other compilers since well before 2011.
19845 Explicit instantiations can be used for the largest or most frequently
19846 duplicated instances, without having to know exactly which other instances
19847 are used in the rest of the program. You can scatter the explicit
19848 instantiations throughout your program, perhaps putting them in the
19849 translation units where the instances are used or the translation units
19850 that define the templates themselves; you can put all of the explicit
19851 instantiations you need into one big file; or you can create small files
19858 template class Foo<int>;
19859 template ostream& operator <<
19860 (ostream&, const Foo<int>&);
19864 for each of the instances you need, and create a template instantiation
19865 library from those.
19867 This is the simplest option, but also offers flexibility and
19868 fine-grained control when necessary. It is also the most portable
19869 alternative and programs using this approach will work with most modern
19874 Compile your template-using code with @option{-frepo}. The compiler
19875 generates files with the extension @samp{.rpo} listing all of the
19876 template instantiations used in the corresponding object files that
19877 could be instantiated there; the link wrapper, @samp{collect2},
19878 then updates the @samp{.rpo} files to tell the compiler where to place
19879 those instantiations and rebuild any affected object files. The
19880 link-time overhead is negligible after the first pass, as the compiler
19881 continues to place the instantiations in the same files.
19883 This can be a suitable option for application code written for the Borland
19884 model, as it usually just works. Code written for the Cfront model
19885 needs to be modified so that the template definitions are available at
19886 one or more points of instantiation; usually this is as simple as adding
19887 @code{#include <tmethods.cc>} to the end of each template header.
19889 For library code, if you want the library to provide all of the template
19890 instantiations it needs, just try to link all of its object files
19891 together; the link will fail, but cause the instantiations to be
19892 generated as a side effect. Be warned, however, that this may cause
19893 conflicts if multiple libraries try to provide the same instantiations.
19894 For greater control, use explicit instantiation as described in the next
19898 @opindex fno-implicit-templates
19899 Compile your code with @option{-fno-implicit-templates} to disable the
19900 implicit generation of template instances, and explicitly instantiate
19901 all the ones you use. This approach requires more knowledge of exactly
19902 which instances you need than do the others, but it's less
19903 mysterious and allows greater control if you want to ensure that only
19904 the intended instances are used.
19906 If you are using Cfront-model code, you can probably get away with not
19907 using @option{-fno-implicit-templates} when compiling files that don't
19908 @samp{#include} the member template definitions.
19910 If you use one big file to do the instantiations, you may want to
19911 compile it without @option{-fno-implicit-templates} so you get all of the
19912 instances required by your explicit instantiations (but not by any
19913 other files) without having to specify them as well.
19915 In addition to forward declaration of explicit instantiations
19916 (with @code{extern}), G++ has extended the template instantiation
19917 syntax to support instantiation of the compiler support data for a
19918 template class (i.e.@: the vtable) without instantiating any of its
19919 members (with @code{inline}), and instantiation of only the static data
19920 members of a template class, without the support data or member
19921 functions (with @code{static}):
19924 inline template class Foo<int>;
19925 static template class Foo<int>;
19929 @node Bound member functions
19930 @section Extracting the Function Pointer from a Bound Pointer to Member Function
19932 @cindex pointer to member function
19933 @cindex bound pointer to member function
19935 In C++, pointer to member functions (PMFs) are implemented using a wide
19936 pointer of sorts to handle all the possible call mechanisms; the PMF
19937 needs to store information about how to adjust the @samp{this} pointer,
19938 and if the function pointed to is virtual, where to find the vtable, and
19939 where in the vtable to look for the member function. If you are using
19940 PMFs in an inner loop, you should really reconsider that decision. If
19941 that is not an option, you can extract the pointer to the function that
19942 would be called for a given object/PMF pair and call it directly inside
19943 the inner loop, to save a bit of time.
19945 Note that you still pay the penalty for the call through a
19946 function pointer; on most modern architectures, such a call defeats the
19947 branch prediction features of the CPU@. This is also true of normal
19948 virtual function calls.
19950 The syntax for this extension is
19954 extern int (A::*fp)();
19955 typedef int (*fptr)(A *);
19957 fptr p = (fptr)(a.*fp);
19960 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
19961 no object is needed to obtain the address of the function. They can be
19962 converted to function pointers directly:
19965 fptr p1 = (fptr)(&A::foo);
19968 @opindex Wno-pmf-conversions
19969 You must specify @option{-Wno-pmf-conversions} to use this extension.
19971 @node C++ Attributes
19972 @section C++-Specific Variable, Function, and Type Attributes
19974 Some attributes only make sense for C++ programs.
19977 @item abi_tag ("@var{tag}", ...)
19978 @cindex @code{abi_tag} function attribute
19979 @cindex @code{abi_tag} variable attribute
19980 @cindex @code{abi_tag} type attribute
19981 The @code{abi_tag} attribute can be applied to a function, variable, or class
19982 declaration. It modifies the mangled name of the entity to
19983 incorporate the tag name, in order to distinguish the function or
19984 class from an earlier version with a different ABI; perhaps the class
19985 has changed size, or the function has a different return type that is
19986 not encoded in the mangled name.
19988 The attribute can also be applied to an inline namespace, but does not
19989 affect the mangled name of the namespace; in this case it is only used
19990 for @option{-Wabi-tag} warnings and automatic tagging of functions and
19991 variables. Tagging inline namespaces is generally preferable to
19992 tagging individual declarations, but the latter is sometimes
19993 necessary, such as when only certain members of a class need to be
19996 The argument can be a list of strings of arbitrary length. The
19997 strings are sorted on output, so the order of the list is
20000 A redeclaration of an entity must not add new ABI tags,
20001 since doing so would change the mangled name.
20003 The ABI tags apply to a name, so all instantiations and
20004 specializations of a template have the same tags. The attribute will
20005 be ignored if applied to an explicit specialization or instantiation.
20007 The @option{-Wabi-tag} flag enables a warning about a class which does
20008 not have all the ABI tags used by its subobjects and virtual functions; for users with code
20009 that needs to coexist with an earlier ABI, using this option can help
20010 to find all affected types that need to be tagged.
20012 When a type involving an ABI tag is used as the type of a variable or
20013 return type of a function where that tag is not already present in the
20014 signature of the function, the tag is automatically applied to the
20015 variable or function. @option{-Wabi-tag} also warns about this
20016 situation; this warning can be avoided by explicitly tagging the
20017 variable or function or moving it into a tagged inline namespace.
20019 @item init_priority (@var{priority})
20020 @cindex @code{init_priority} variable attribute
20022 In Standard C++, objects defined at namespace scope are guaranteed to be
20023 initialized in an order in strict accordance with that of their definitions
20024 @emph{in a given translation unit}. No guarantee is made for initializations
20025 across translation units. However, GNU C++ allows users to control the
20026 order of initialization of objects defined at namespace scope with the
20027 @code{init_priority} attribute by specifying a relative @var{priority},
20028 a constant integral expression currently bounded between 101 and 65535
20029 inclusive. Lower numbers indicate a higher priority.
20031 In the following example, @code{A} would normally be created before
20032 @code{B}, but the @code{init_priority} attribute reverses that order:
20035 Some_Class A __attribute__ ((init_priority (2000)));
20036 Some_Class B __attribute__ ((init_priority (543)));
20040 Note that the particular values of @var{priority} do not matter; only their
20043 @item java_interface
20044 @cindex @code{java_interface} type attribute
20046 This type attribute informs C++ that the class is a Java interface. It may
20047 only be applied to classes declared within an @code{extern "Java"} block.
20048 Calls to methods declared in this interface are dispatched using GCJ's
20049 interface table mechanism, instead of regular virtual table dispatch.
20052 @cindex @code{warn_unused} type attribute
20054 For C++ types with non-trivial constructors and/or destructors it is
20055 impossible for the compiler to determine whether a variable of this
20056 type is truly unused if it is not referenced. This type attribute
20057 informs the compiler that variables of this type should be warned
20058 about if they appear to be unused, just like variables of fundamental
20061 This attribute is appropriate for types which just represent a value,
20062 such as @code{std::string}; it is not appropriate for types which
20063 control a resource, such as @code{std::mutex}.
20065 This attribute is also accepted in C, but it is unnecessary because C
20066 does not have constructors or destructors.
20070 See also @ref{Namespace Association}.
20072 @node Function Multiversioning
20073 @section Function Multiversioning
20074 @cindex function versions
20076 With the GNU C++ front end, for x86 targets, you may specify multiple
20077 versions of a function, where each function is specialized for a
20078 specific target feature. At runtime, the appropriate version of the
20079 function is automatically executed depending on the characteristics of
20080 the execution platform. Here is an example.
20083 __attribute__ ((target ("default")))
20086 // The default version of foo.
20090 __attribute__ ((target ("sse4.2")))
20093 // foo version for SSE4.2
20097 __attribute__ ((target ("arch=atom")))
20100 // foo version for the Intel ATOM processor
20104 __attribute__ ((target ("arch=amdfam10")))
20107 // foo version for the AMD Family 0x10 processors.
20114 assert ((*p) () == foo ());
20119 In the above example, four versions of function foo are created. The
20120 first version of foo with the target attribute "default" is the default
20121 version. This version gets executed when no other target specific
20122 version qualifies for execution on a particular platform. A new version
20123 of foo is created by using the same function signature but with a
20124 different target string. Function foo is called or a pointer to it is
20125 taken just like a regular function. GCC takes care of doing the
20126 dispatching to call the right version at runtime. Refer to the
20127 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
20128 Function Multiversioning} for more details.
20130 @node Namespace Association
20131 @section Namespace Association
20133 @strong{Caution:} The semantics of this extension are equivalent
20134 to C++ 2011 inline namespaces. Users should use inline namespaces
20135 instead as this extension will be removed in future versions of G++.
20137 A using-directive with @code{__attribute ((strong))} is stronger
20138 than a normal using-directive in two ways:
20142 Templates from the used namespace can be specialized and explicitly
20143 instantiated as though they were members of the using namespace.
20146 The using namespace is considered an associated namespace of all
20147 templates in the used namespace for purposes of argument-dependent
20151 The used namespace must be nested within the using namespace so that
20152 normal unqualified lookup works properly.
20154 This is useful for composing a namespace transparently from
20155 implementation namespaces. For example:
20160 template <class T> struct A @{ @};
20162 using namespace debug __attribute ((__strong__));
20163 template <> struct A<int> @{ @}; // @r{OK to specialize}
20165 template <class T> void f (A<T>);
20170 f (std::A<float>()); // @r{lookup finds} std::f
20176 @section Type Traits
20178 The C++ front end implements syntactic extensions that allow
20179 compile-time determination of
20180 various characteristics of a type (or of a
20184 @item __has_nothrow_assign (type)
20185 If @code{type} is const qualified or is a reference type then the trait is
20186 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
20187 is true, else if @code{type} is a cv class or union type with copy assignment
20188 operators that are known not to throw an exception then the trait is true,
20189 else it is false. Requires: @code{type} shall be a complete type,
20190 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20192 @item __has_nothrow_copy (type)
20193 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
20194 @code{type} is a cv class or union type with copy constructors that
20195 are known not to throw an exception then the trait is true, else it is false.
20196 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
20197 @code{void}, or an array of unknown bound.
20199 @item __has_nothrow_constructor (type)
20200 If @code{__has_trivial_constructor (type)} is true then the trait is
20201 true, else if @code{type} is a cv class or union type (or array
20202 thereof) with a default constructor that is known not to throw an
20203 exception then the trait is true, else it is false. Requires:
20204 @code{type} shall be a complete type, (possibly cv-qualified)
20205 @code{void}, or an array of unknown bound.
20207 @item __has_trivial_assign (type)
20208 If @code{type} is const qualified or is a reference type then the trait is
20209 false. Otherwise if @code{__is_pod (type)} is true then the trait is
20210 true, else if @code{type} is a cv class or union type with a trivial
20211 copy assignment ([class.copy]) then the trait is true, else it is
20212 false. Requires: @code{type} shall be a complete type, (possibly
20213 cv-qualified) @code{void}, or an array of unknown bound.
20215 @item __has_trivial_copy (type)
20216 If @code{__is_pod (type)} is true or @code{type} is a reference type
20217 then the trait is true, else if @code{type} is a cv class or union type
20218 with a trivial copy constructor ([class.copy]) then the trait
20219 is true, else it is false. Requires: @code{type} shall be a complete
20220 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20222 @item __has_trivial_constructor (type)
20223 If @code{__is_pod (type)} is true then the trait is true, else if
20224 @code{type} is a cv class or union type (or array thereof) with a
20225 trivial default constructor ([class.ctor]) then the trait is true,
20226 else it is false. Requires: @code{type} shall be a complete
20227 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20229 @item __has_trivial_destructor (type)
20230 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20231 the trait is true, else if @code{type} is a cv class or union type (or
20232 array thereof) with a trivial destructor ([class.dtor]) then the trait
20233 is true, else it is false. Requires: @code{type} shall be a complete
20234 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20236 @item __has_virtual_destructor (type)
20237 If @code{type} is a class type with a virtual destructor
20238 ([class.dtor]) then the trait is true, else it is false. Requires:
20239 @code{type} shall be a complete type, (possibly cv-qualified)
20240 @code{void}, or an array of unknown bound.
20242 @item __is_abstract (type)
20243 If @code{type} is an abstract class ([class.abstract]) then the trait
20244 is true, else it is false. Requires: @code{type} shall be a complete
20245 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20247 @item __is_base_of (base_type, derived_type)
20248 If @code{base_type} is a base class of @code{derived_type}
20249 ([class.derived]) then the trait is true, otherwise it is false.
20250 Top-level cv qualifications of @code{base_type} and
20251 @code{derived_type} are ignored. For the purposes of this trait, a
20252 class type is considered is own base. Requires: if @code{__is_class
20253 (base_type)} and @code{__is_class (derived_type)} are true and
20254 @code{base_type} and @code{derived_type} are not the same type
20255 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20256 type. Diagnostic is produced if this requirement is not met.
20258 @item __is_class (type)
20259 If @code{type} is a cv class type, and not a union type
20260 ([basic.compound]) the trait is true, else it is false.
20262 @item __is_empty (type)
20263 If @code{__is_class (type)} is false then the trait is false.
20264 Otherwise @code{type} is considered empty if and only if: @code{type}
20265 has no non-static data members, or all non-static data members, if
20266 any, are bit-fields of length 0, and @code{type} has no virtual
20267 members, and @code{type} has no virtual base classes, and @code{type}
20268 has no base classes @code{base_type} for which
20269 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20270 be a complete type, (possibly cv-qualified) @code{void}, or an array
20273 @item __is_enum (type)
20274 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20275 true, else it is false.
20277 @item __is_literal_type (type)
20278 If @code{type} is a literal type ([basic.types]) the trait is
20279 true, else it is false. Requires: @code{type} shall be a complete type,
20280 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20282 @item __is_pod (type)
20283 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20284 else it is false. Requires: @code{type} shall be a complete type,
20285 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20287 @item __is_polymorphic (type)
20288 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20289 is true, else it is false. Requires: @code{type} shall be a complete
20290 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20292 @item __is_standard_layout (type)
20293 If @code{type} is a standard-layout type ([basic.types]) the trait is
20294 true, else it is false. Requires: @code{type} shall be a complete
20295 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20297 @item __is_trivial (type)
20298 If @code{type} is a trivial type ([basic.types]) the trait is
20299 true, else it is false. Requires: @code{type} shall be a complete
20300 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20302 @item __is_union (type)
20303 If @code{type} is a cv union type ([basic.compound]) the trait is
20304 true, else it is false.
20306 @item __underlying_type (type)
20307 The underlying type of @code{type}. Requires: @code{type} shall be
20308 an enumeration type ([dcl.enum]).
20314 @section C++ Concepts
20316 C++ concepts provide much-improved support for generic programming. In
20317 particular, they allow the specification of constraints on template arguments.
20318 The constraints are used to extend the usual overloading and partial
20319 specialization capabilities of the language, allowing generic data structures
20320 and algorithms to be ``refined'' based on their properties rather than their
20323 The following keywords are reserved for concepts.
20327 States an expression as an assumption, and if possible, verifies that the
20328 assumption is valid. For example, @code{assume(n > 0)}.
20331 Introduces an axiom definition. Axioms introduce requirements on values.
20334 Introduces a universally quantified object in an axiom. For example,
20335 @code{forall (int n) n + 0 == n}).
20338 Introduces a concept definition. Concepts are sets of syntactic and semantic
20339 requirements on types and their values.
20342 Introduces constraints on template arguments or requirements for a member
20343 function of a class template.
20347 The front end also exposes a number of internal mechanism that can be used
20348 to simplify the writing of type traits. Note that some of these traits are
20349 likely to be removed in the future.
20352 @item __is_same (type1, type2)
20353 A binary type trait: true whenever the type arguments are the same.
20358 @node Java Exceptions
20359 @section Java Exceptions
20361 The Java language uses a slightly different exception handling model
20362 from C++. Normally, GNU C++ automatically detects when you are
20363 writing C++ code that uses Java exceptions, and handle them
20364 appropriately. However, if C++ code only needs to execute destructors
20365 when Java exceptions are thrown through it, GCC guesses incorrectly.
20366 Sample problematic code is:
20369 struct S @{ ~S(); @};
20370 extern void bar(); // @r{is written in Java, and may throw exceptions}
20379 The usual effect of an incorrect guess is a link failure, complaining of
20380 a missing routine called @samp{__gxx_personality_v0}.
20382 You can inform the compiler that Java exceptions are to be used in a
20383 translation unit, irrespective of what it might think, by writing
20384 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20385 @samp{#pragma} must appear before any functions that throw or catch
20386 exceptions, or run destructors when exceptions are thrown through them.
20388 You cannot mix Java and C++ exceptions in the same translation unit. It
20389 is believed to be safe to throw a C++ exception from one file through
20390 another file compiled for the Java exception model, or vice versa, but
20391 there may be bugs in this area.
20393 @node Deprecated Features
20394 @section Deprecated Features
20396 In the past, the GNU C++ compiler was extended to experiment with new
20397 features, at a time when the C++ language was still evolving. Now that
20398 the C++ standard is complete, some of those features are superseded by
20399 superior alternatives. Using the old features might cause a warning in
20400 some cases that the feature will be dropped in the future. In other
20401 cases, the feature might be gone already.
20403 While the list below is not exhaustive, it documents some of the options
20404 that are now deprecated:
20407 @item -fexternal-templates
20408 @itemx -falt-external-templates
20409 These are two of the many ways for G++ to implement template
20410 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20411 defines how template definitions have to be organized across
20412 implementation units. G++ has an implicit instantiation mechanism that
20413 should work just fine for standard-conforming code.
20415 @item -fstrict-prototype
20416 @itemx -fno-strict-prototype
20417 Previously it was possible to use an empty prototype parameter list to
20418 indicate an unspecified number of parameters (like C), rather than no
20419 parameters, as C++ demands. This feature has been removed, except where
20420 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20423 G++ allows a virtual function returning @samp{void *} to be overridden
20424 by one returning a different pointer type. This extension to the
20425 covariant return type rules is now deprecated and will be removed from a
20428 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20429 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20430 and are now removed from G++. Code using these operators should be
20431 modified to use @code{std::min} and @code{std::max} instead.
20433 The named return value extension has been deprecated, and is now
20436 The use of initializer lists with new expressions has been deprecated,
20437 and is now removed from G++.
20439 Floating and complex non-type template parameters have been deprecated,
20440 and are now removed from G++.
20442 The implicit typename extension has been deprecated and is now
20445 The use of default arguments in function pointers, function typedefs
20446 and other places where they are not permitted by the standard is
20447 deprecated and will be removed from a future version of G++.
20449 G++ allows floating-point literals to appear in integral constant expressions,
20450 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20451 This extension is deprecated and will be removed from a future version.
20453 G++ allows static data members of const floating-point type to be declared
20454 with an initializer in a class definition. The standard only allows
20455 initializers for static members of const integral types and const
20456 enumeration types so this extension has been deprecated and will be removed
20457 from a future version.
20459 @node Backwards Compatibility
20460 @section Backwards Compatibility
20461 @cindex Backwards Compatibility
20462 @cindex ARM [Annotated C++ Reference Manual]
20464 Now that there is a definitive ISO standard C++, G++ has a specification
20465 to adhere to. The C++ language evolved over time, and features that
20466 used to be acceptable in previous drafts of the standard, such as the ARM
20467 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20468 compilation of C++ written to such drafts, G++ contains some backwards
20469 compatibilities. @emph{All such backwards compatibility features are
20470 liable to disappear in future versions of G++.} They should be considered
20471 deprecated. @xref{Deprecated Features}.
20475 If a variable is declared at for scope, it used to remain in scope until
20476 the end of the scope that contained the for statement (rather than just
20477 within the for scope). G++ retains this, but issues a warning, if such a
20478 variable is accessed outside the for scope.
20480 @item Implicit C language
20481 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20482 scope to set the language. On such systems, all header files are
20483 implicitly scoped inside a C language scope. Also, an empty prototype
20484 @code{()} is treated as an unspecified number of arguments, rather
20485 than no arguments, as C++ demands.
20488 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20489 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr