1 @c Copyright (C) 1988-2016 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Extensions to the C Language Family
8 @cindex extensions, C language
9 @cindex C language extensions
12 GNU C provides several language features not found in ISO standard C@.
13 (The @option{-pedantic} option directs GCC to print a warning message if
14 any of these features is used.) To test for the availability of these
15 features in conditional compilation, check for a predefined macro
16 @code{__GNUC__}, which is always defined under GCC@.
18 These extensions are available in C and Objective-C@. Most of them are
19 also available in C++. @xref{C++ Extensions,,Extensions to the
20 C++ Language}, for extensions that apply @emph{only} to C++.
22 Some features that are in ISO C99 but not C90 or C++ are also, as
23 extensions, accepted by GCC in C90 mode and in C++.
26 * Statement Exprs:: Putting statements and declarations inside expressions.
27 * Local Labels:: Labels local to a block.
28 * Labels as Values:: Getting pointers to labels, and computed gotos.
29 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
30 * Constructing Calls:: Dispatching a call to another function.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
33 * __int128:: 128-bit integers---@code{__int128}.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Floating Types:: Additional Floating Types.
37 * Half-Precision:: Half-Precision Floating Point.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Named Address Spaces::Named address spaces.
42 * Zero Length:: Zero-length arrays.
43 * Empty Structures:: Structures with no members.
44 * Variable Length:: Arrays whose length is computed at run time.
45 * Variadic Macros:: Macros with a variable number of arguments.
46 * Escaped Newlines:: Slightly looser rules for escaped newlines.
47 * Subscripting:: Any array can be subscripted, even if not an lvalue.
48 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
49 * Pointers to Arrays:: Pointers to arrays with qualifiers work as expected.
50 * Initializers:: Non-constant initializers.
51 * Compound Literals:: Compound literals give structures, unions
53 * Designated Inits:: Labeling elements of initializers.
54 * Case Ranges:: `case 1 ... 9' and such.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Mixed Declarations:: Mixing declarations and code.
57 * Function Attributes:: Declaring that functions have no side effects,
58 or that they can never return.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Label Attributes:: Specifying attributes on labels.
62 * Enumerator Attributes:: Specifying attributes on enumerators.
63 * Attribute Syntax:: Formal syntax for attributes.
64 * Function Prototypes:: Prototype declarations and old-style definitions.
65 * C++ Comments:: C++ comments are recognized.
66 * Dollar Signs:: Dollar sign is allowed in identifiers.
67 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
68 * Alignment:: Inquiring about the alignment of a type or variable.
69 * Inline:: Defining inline functions (as fast as macros).
70 * Volatiles:: What constitutes an access to a volatile object.
71 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
72 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
73 * Incomplete Enums:: @code{enum foo;}, with details to follow.
74 * Function Names:: Printable strings which are the name of the current
76 * Return Address:: Getting the return or frame address of a function.
77 * Vector Extensions:: Using vector instructions through built-in functions.
78 * Offsetof:: Special syntax for implementing @code{offsetof}.
79 * __sync Builtins:: Legacy built-in functions for atomic memory access.
80 * __atomic Builtins:: Atomic built-in functions with memory model.
81 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
82 arithmetic overflow checking.
83 * x86 specific memory model extensions for transactional memory:: x86 memory models.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
87 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
88 * Other Builtins:: Other built-in functions.
89 * Target Builtins:: Built-in functions specific to particular targets.
90 * Target Format Checks:: Format checks specific to particular targets.
91 * Pragmas:: Pragmas accepted by GCC.
92 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
93 * Thread-Local:: Per-thread variables.
94 * Binary constants:: Binary constants using the @samp{0b} prefix.
98 @section Statements and Declarations in Expressions
99 @cindex statements inside expressions
100 @cindex declarations inside expressions
101 @cindex expressions containing statements
102 @cindex macros, statements in expressions
104 @c the above section title wrapped and causes an underfull hbox.. i
105 @c changed it from "within" to "in". --mew 4feb93
106 A compound statement enclosed in parentheses may appear as an expression
107 in GNU C@. This allows you to use loops, switches, and local variables
108 within an expression.
110 Recall that a compound statement is a sequence of statements surrounded
111 by braces; in this construct, parentheses go around the braces. For
115 (@{ int y = foo (); int z;
122 is a valid (though slightly more complex than necessary) expression
123 for the absolute value of @code{foo ()}.
125 The last thing in the compound statement should be an expression
126 followed by a semicolon; the value of this subexpression serves as the
127 value of the entire construct. (If you use some other kind of statement
128 last within the braces, the construct has type @code{void}, and thus
129 effectively no value.)
131 This feature is especially useful in making macro definitions ``safe'' (so
132 that they evaluate each operand exactly once). For example, the
133 ``maximum'' function is commonly defined as a macro in standard C as
137 #define max(a,b) ((a) > (b) ? (a) : (b))
141 @cindex side effects, macro argument
142 But this definition computes either @var{a} or @var{b} twice, with bad
143 results if the operand has side effects. In GNU C, if you know the
144 type of the operands (here taken as @code{int}), you can define
145 the macro safely as follows:
148 #define maxint(a,b) \
149 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
152 Embedded statements are not allowed in constant expressions, such as
153 the value of an enumeration constant, the width of a bit-field, or
154 the initial value of a static variable.
156 If you don't know the type of the operand, you can still do this, but you
157 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
159 In G++, the result value of a statement expression undergoes array and
160 function pointer decay, and is returned by value to the enclosing
161 expression. For instance, if @code{A} is a class, then
170 constructs a temporary @code{A} object to hold the result of the
171 statement expression, and that is used to invoke @code{Foo}.
172 Therefore the @code{this} pointer observed by @code{Foo} is not the
175 In a statement expression, any temporaries created within a statement
176 are destroyed at that statement's end. This makes statement
177 expressions inside macros slightly different from function calls. In
178 the latter case temporaries introduced during argument evaluation are
179 destroyed at the end of the statement that includes the function
180 call. In the statement expression case they are destroyed during
181 the statement expression. For instance,
184 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
185 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
195 has different places where temporaries are destroyed. For the
196 @code{macro} case, the temporary @code{X} is destroyed just after
197 the initialization of @code{b}. In the @code{function} case that
198 temporary is destroyed when the function returns.
200 These considerations mean that it is probably a bad idea to use
201 statement expressions of this form in header files that are designed to
202 work with C++. (Note that some versions of the GNU C Library contained
203 header files using statement expressions that lead to precisely this
206 Jumping into a statement expression with @code{goto} or using a
207 @code{switch} statement outside the statement expression with a
208 @code{case} or @code{default} label inside the statement expression is
209 not permitted. Jumping into a statement expression with a computed
210 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
211 Jumping out of a statement expression is permitted, but if the
212 statement expression is part of a larger expression then it is
213 unspecified which other subexpressions of that expression have been
214 evaluated except where the language definition requires certain
215 subexpressions to be evaluated before or after the statement
216 expression. In any case, as with a function call, the evaluation of a
217 statement expression is not interleaved with the evaluation of other
218 parts of the containing expression. For example,
221 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
225 calls @code{foo} and @code{bar1} and does not call @code{baz} but
226 may or may not call @code{bar2}. If @code{bar2} is called, it is
227 called after @code{foo} and before @code{bar1}.
230 @section Locally Declared Labels
232 @cindex macros, local labels
234 GCC allows you to declare @dfn{local labels} in any nested block
235 scope. A local label is just like an ordinary label, but you can
236 only reference it (with a @code{goto} statement, or by taking its
237 address) within the block in which it is declared.
239 A local label declaration looks like this:
242 __label__ @var{label};
249 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
252 Local label declarations must come at the beginning of the block,
253 before any ordinary declarations or statements.
255 The label declaration defines the label @emph{name}, but does not define
256 the label itself. You must do this in the usual way, with
257 @code{@var{label}:}, within the statements of the statement expression.
259 The local label feature is useful for complex macros. If a macro
260 contains nested loops, a @code{goto} can be useful for breaking out of
261 them. However, an ordinary label whose scope is the whole function
262 cannot be used: if the macro can be expanded several times in one
263 function, the label is multiply defined in that function. A
264 local label avoids this problem. For example:
267 #define SEARCH(value, array, target) \
270 typeof (target) _SEARCH_target = (target); \
271 typeof (*(array)) *_SEARCH_array = (array); \
274 for (i = 0; i < max; i++) \
275 for (j = 0; j < max; j++) \
276 if (_SEARCH_array[i][j] == _SEARCH_target) \
277 @{ (value) = i; goto found; @} \
283 This could also be written using a statement expression:
286 #define SEARCH(array, target) \
289 typeof (target) _SEARCH_target = (target); \
290 typeof (*(array)) *_SEARCH_array = (array); \
293 for (i = 0; i < max; i++) \
294 for (j = 0; j < max; j++) \
295 if (_SEARCH_array[i][j] == _SEARCH_target) \
296 @{ value = i; goto found; @} \
303 Local label declarations also make the labels they declare visible to
304 nested functions, if there are any. @xref{Nested Functions}, for details.
306 @node Labels as Values
307 @section Labels as Values
308 @cindex labels as values
309 @cindex computed gotos
310 @cindex goto with computed label
311 @cindex address of a label
313 You can get the address of a label defined in the current function
314 (or a containing function) with the unary operator @samp{&&}. The
315 value has type @code{void *}. This value is a constant and can be used
316 wherever a constant of that type is valid. For example:
324 To use these values, you need to be able to jump to one. This is done
325 with the computed goto statement@footnote{The analogous feature in
326 Fortran is called an assigned goto, but that name seems inappropriate in
327 C, where one can do more than simply store label addresses in label
328 variables.}, @code{goto *@var{exp};}. For example,
335 Any expression of type @code{void *} is allowed.
337 One way of using these constants is in initializing a static array that
338 serves as a jump table:
341 static void *array[] = @{ &&foo, &&bar, &&hack @};
345 Then you can select a label with indexing, like this:
352 Note that this does not check whether the subscript is in bounds---array
353 indexing in C never does that.
355 Such an array of label values serves a purpose much like that of the
356 @code{switch} statement. The @code{switch} statement is cleaner, so
357 use that rather than an array unless the problem does not fit a
358 @code{switch} statement very well.
360 Another use of label values is in an interpreter for threaded code.
361 The labels within the interpreter function can be stored in the
362 threaded code for super-fast dispatching.
364 You may not use this mechanism to jump to code in a different function.
365 If you do that, totally unpredictable things happen. The best way to
366 avoid this is to store the label address only in automatic variables and
367 never pass it as an argument.
369 An alternate way to write the above example is
372 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
374 goto *(&&foo + array[i]);
378 This is more friendly to code living in shared libraries, as it reduces
379 the number of dynamic relocations that are needed, and by consequence,
380 allows the data to be read-only.
381 This alternative with label differences is not supported for the AVR target,
382 please use the first approach for AVR programs.
384 The @code{&&foo} expressions for the same label might have different
385 values if the containing function is inlined or cloned. If a program
386 relies on them being always the same,
387 @code{__attribute__((__noinline__,__noclone__))} should be used to
388 prevent inlining and cloning. If @code{&&foo} is used in a static
389 variable initializer, inlining and cloning is forbidden.
391 @node Nested Functions
392 @section Nested Functions
393 @cindex nested functions
394 @cindex downward funargs
397 A @dfn{nested function} is a function defined inside another function.
398 Nested functions are supported as an extension in GNU C, but are not
399 supported by GNU C++.
401 The nested function's name is local to the block where it is defined.
402 For example, here we define a nested function named @code{square}, and
407 foo (double a, double b)
409 double square (double z) @{ return z * z; @}
411 return square (a) + square (b);
416 The nested function can access all the variables of the containing
417 function that are visible at the point of its definition. This is
418 called @dfn{lexical scoping}. For example, here we show a nested
419 function which uses an inherited variable named @code{offset}:
423 bar (int *array, int offset, int size)
425 int access (int *array, int index)
426 @{ return array[index + offset]; @}
429 for (i = 0; i < size; i++)
430 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
435 Nested function definitions are permitted within functions in the places
436 where variable definitions are allowed; that is, in any block, mixed
437 with the other declarations and statements in the block.
439 It is possible to call the nested function from outside the scope of its
440 name by storing its address or passing the address to another function:
443 hack (int *array, int size)
445 void store (int index, int value)
446 @{ array[index] = value; @}
448 intermediate (store, size);
452 Here, the function @code{intermediate} receives the address of
453 @code{store} as an argument. If @code{intermediate} calls @code{store},
454 the arguments given to @code{store} are used to store into @code{array}.
455 But this technique works only so long as the containing function
456 (@code{hack}, in this example) does not exit.
458 If you try to call the nested function through its address after the
459 containing function exits, all hell breaks loose. If you try
460 to call it after a containing scope level exits, and if it refers
461 to some of the variables that are no longer in scope, you may be lucky,
462 but it's not wise to take the risk. If, however, the nested function
463 does not refer to anything that has gone out of scope, you should be
466 GCC implements taking the address of a nested function using a technique
467 called @dfn{trampolines}. This technique was described in
468 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
469 C++ Conference Proceedings, October 17-21, 1988).
471 A nested function can jump to a label inherited from a containing
472 function, provided the label is explicitly declared in the containing
473 function (@pxref{Local Labels}). Such a jump returns instantly to the
474 containing function, exiting the nested function that did the
475 @code{goto} and any intermediate functions as well. Here is an example:
479 bar (int *array, int offset, int size)
482 int access (int *array, int index)
486 return array[index + offset];
490 for (i = 0; i < size; i++)
491 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
495 /* @r{Control comes here from @code{access}
496 if it detects an error.} */
503 A nested function always has no linkage. Declaring one with
504 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
505 before its definition, use @code{auto} (which is otherwise meaningless
506 for function declarations).
509 bar (int *array, int offset, int size)
512 auto int access (int *, int);
514 int access (int *array, int index)
518 return array[index + offset];
524 @node Constructing Calls
525 @section Constructing Function Calls
526 @cindex constructing calls
527 @cindex forwarding calls
529 Using the built-in functions described below, you can record
530 the arguments a function received, and call another function
531 with the same arguments, without knowing the number or types
534 You can also record the return value of that function call,
535 and later return that value, without knowing what data type
536 the function tried to return (as long as your caller expects
539 However, these built-in functions may interact badly with some
540 sophisticated features or other extensions of the language. It
541 is, therefore, not recommended to use them outside very simple
542 functions acting as mere forwarders for their arguments.
544 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
545 This built-in function returns a pointer to data
546 describing how to perform a call with the same arguments as are passed
547 to the current function.
549 The function saves the arg pointer register, structure value address,
550 and all registers that might be used to pass arguments to a function
551 into a block of memory allocated on the stack. Then it returns the
552 address of that block.
555 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
556 This built-in function invokes @var{function}
557 with a copy of the parameters described by @var{arguments}
560 The value of @var{arguments} should be the value returned by
561 @code{__builtin_apply_args}. The argument @var{size} specifies the size
562 of the stack argument data, in bytes.
564 This function returns a pointer to data describing
565 how to return whatever value is returned by @var{function}. The data
566 is saved in a block of memory allocated on the stack.
568 It is not always simple to compute the proper value for @var{size}. The
569 value is used by @code{__builtin_apply} to compute the amount of data
570 that should be pushed on the stack and copied from the incoming argument
574 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
575 This built-in function returns the value described by @var{result} from
576 the containing function. You should specify, for @var{result}, a value
577 returned by @code{__builtin_apply}.
580 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
581 This built-in function represents all anonymous arguments of an inline
582 function. It can be used only in inline functions that are always
583 inlined, never compiled as a separate function, such as those using
584 @code{__attribute__ ((__always_inline__))} or
585 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
586 It must be only passed as last argument to some other function
587 with variable arguments. This is useful for writing small wrapper
588 inlines for variable argument functions, when using preprocessor
589 macros is undesirable. For example:
591 extern int myprintf (FILE *f, const char *format, ...);
592 extern inline __attribute__ ((__gnu_inline__)) int
593 myprintf (FILE *f, const char *format, ...)
595 int r = fprintf (f, "myprintf: ");
598 int s = fprintf (f, format, __builtin_va_arg_pack ());
606 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
607 This built-in function returns the number of anonymous arguments of
608 an inline function. It can be used only in inline functions that
609 are always inlined, never compiled as a separate function, such
610 as those using @code{__attribute__ ((__always_inline__))} or
611 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
612 For example following does link- or run-time checking of open
613 arguments for optimized code:
616 extern inline __attribute__((__gnu_inline__)) int
617 myopen (const char *path, int oflag, ...)
619 if (__builtin_va_arg_pack_len () > 1)
620 warn_open_too_many_arguments ();
622 if (__builtin_constant_p (oflag))
624 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
626 warn_open_missing_mode ();
627 return __open_2 (path, oflag);
629 return open (path, oflag, __builtin_va_arg_pack ());
632 if (__builtin_va_arg_pack_len () < 1)
633 return __open_2 (path, oflag);
635 return open (path, oflag, __builtin_va_arg_pack ());
642 @section Referring to a Type with @code{typeof}
645 @cindex macros, types of arguments
647 Another way to refer to the type of an expression is with @code{typeof}.
648 The syntax of using of this keyword looks like @code{sizeof}, but the
649 construct acts semantically like a type name defined with @code{typedef}.
651 There are two ways of writing the argument to @code{typeof}: with an
652 expression or with a type. Here is an example with an expression:
659 This assumes that @code{x} is an array of pointers to functions;
660 the type described is that of the values of the functions.
662 Here is an example with a typename as the argument:
669 Here the type described is that of pointers to @code{int}.
671 If you are writing a header file that must work when included in ISO C
672 programs, write @code{__typeof__} instead of @code{typeof}.
673 @xref{Alternate Keywords}.
675 A @code{typeof} construct can be used anywhere a typedef name can be
676 used. For example, you can use it in a declaration, in a cast, or inside
677 of @code{sizeof} or @code{typeof}.
679 The operand of @code{typeof} is evaluated for its side effects if and
680 only if it is an expression of variably modified type or the name of
683 @code{typeof} is often useful in conjunction with
684 statement expressions (@pxref{Statement Exprs}).
685 Here is how the two together can
686 be used to define a safe ``maximum'' macro which operates on any
687 arithmetic type and evaluates each of its arguments exactly once:
691 (@{ typeof (a) _a = (a); \
692 typeof (b) _b = (b); \
693 _a > _b ? _a : _b; @})
696 @cindex underscores in variables in macros
697 @cindex @samp{_} in variables in macros
698 @cindex local variables in macros
699 @cindex variables, local, in macros
700 @cindex macros, local variables in
702 The reason for using names that start with underscores for the local
703 variables is to avoid conflicts with variable names that occur within the
704 expressions that are substituted for @code{a} and @code{b}. Eventually we
705 hope to design a new form of declaration syntax that allows you to declare
706 variables whose scopes start only after their initializers; this will be a
707 more reliable way to prevent such conflicts.
710 Some more examples of the use of @code{typeof}:
714 This declares @code{y} with the type of what @code{x} points to.
721 This declares @code{y} as an array of such values.
728 This declares @code{y} as an array of pointers to characters:
731 typeof (typeof (char *)[4]) y;
735 It is equivalent to the following traditional C declaration:
741 To see the meaning of the declaration using @code{typeof}, and why it
742 might be a useful way to write, rewrite it with these macros:
745 #define pointer(T) typeof(T *)
746 #define array(T, N) typeof(T [N])
750 Now the declaration can be rewritten this way:
753 array (pointer (char), 4) y;
757 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
758 pointers to @code{char}.
761 In GNU C, but not GNU C++, you may also declare the type of a variable
762 as @code{__auto_type}. In that case, the declaration must declare
763 only one variable, whose declarator must just be an identifier, the
764 declaration must be initialized, and the type of the variable is
765 determined by the initializer; the name of the variable is not in
766 scope until after the initializer. (In C++, you should use C++11
767 @code{auto} for this purpose.) Using @code{__auto_type}, the
768 ``maximum'' macro above could be written as:
772 (@{ __auto_type _a = (a); \
773 __auto_type _b = (b); \
774 _a > _b ? _a : _b; @})
777 Using @code{__auto_type} instead of @code{typeof} has two advantages:
780 @item Each argument to the macro appears only once in the expansion of
781 the macro. This prevents the size of the macro expansion growing
782 exponentially when calls to such macros are nested inside arguments of
785 @item If the argument to the macro has variably modified type, it is
786 evaluated only once when using @code{__auto_type}, but twice if
787 @code{typeof} is used.
791 @section Conditionals with Omitted Operands
792 @cindex conditional expressions, extensions
793 @cindex omitted middle-operands
794 @cindex middle-operands, omitted
795 @cindex extensions, @code{?:}
796 @cindex @code{?:} extensions
798 The middle operand in a conditional expression may be omitted. Then
799 if the first operand is nonzero, its value is the value of the conditional
802 Therefore, the expression
809 has the value of @code{x} if that is nonzero; otherwise, the value of
812 This example is perfectly equivalent to
818 @cindex side effect in @code{?:}
819 @cindex @code{?:} side effect
821 In this simple case, the ability to omit the middle operand is not
822 especially useful. When it becomes useful is when the first operand does,
823 or may (if it is a macro argument), contain a side effect. Then repeating
824 the operand in the middle would perform the side effect twice. Omitting
825 the middle operand uses the value already computed without the undesirable
826 effects of recomputing it.
829 @section 128-bit Integers
830 @cindex @code{__int128} data types
832 As an extension the integer scalar type @code{__int128} is supported for
833 targets which have an integer mode wide enough to hold 128 bits.
834 Simply write @code{__int128} for a signed 128-bit integer, or
835 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
836 support in GCC for expressing an integer constant of type @code{__int128}
837 for targets with @code{long long} integer less than 128 bits wide.
840 @section Double-Word Integers
841 @cindex @code{long long} data types
842 @cindex double-word arithmetic
843 @cindex multiprecision arithmetic
844 @cindex @code{LL} integer suffix
845 @cindex @code{ULL} integer suffix
847 ISO C99 supports data types for integers that are at least 64 bits wide,
848 and as an extension GCC supports them in C90 mode and in C++.
849 Simply write @code{long long int} for a signed integer, or
850 @code{unsigned long long int} for an unsigned integer. To make an
851 integer constant of type @code{long long int}, add the suffix @samp{LL}
852 to the integer. To make an integer constant of type @code{unsigned long
853 long int}, add the suffix @samp{ULL} to the integer.
855 You can use these types in arithmetic like any other integer types.
856 Addition, subtraction, and bitwise boolean operations on these types
857 are open-coded on all types of machines. Multiplication is open-coded
858 if the machine supports a fullword-to-doubleword widening multiply
859 instruction. Division and shifts are open-coded only on machines that
860 provide special support. The operations that are not open-coded use
861 special library routines that come with GCC@.
863 There may be pitfalls when you use @code{long long} types for function
864 arguments without function prototypes. If a function
865 expects type @code{int} for its argument, and you pass a value of type
866 @code{long long int}, confusion results because the caller and the
867 subroutine disagree about the number of bytes for the argument.
868 Likewise, if the function expects @code{long long int} and you pass
869 @code{int}. The best way to avoid such problems is to use prototypes.
872 @section Complex Numbers
873 @cindex complex numbers
874 @cindex @code{_Complex} keyword
875 @cindex @code{__complex__} keyword
877 ISO C99 supports complex floating data types, and as an extension GCC
878 supports them in C90 mode and in C++. GCC also supports complex integer data
879 types which are not part of ISO C99. You can declare complex types
880 using the keyword @code{_Complex}. As an extension, the older GNU
881 keyword @code{__complex__} is also supported.
883 For example, @samp{_Complex double x;} declares @code{x} as a
884 variable whose real part and imaginary part are both of type
885 @code{double}. @samp{_Complex short int y;} declares @code{y} to
886 have real and imaginary parts of type @code{short int}; this is not
887 likely to be useful, but it shows that the set of complex types is
890 To write a constant with a complex data type, use the suffix @samp{i} or
891 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
892 has type @code{_Complex float} and @code{3i} has type
893 @code{_Complex int}. Such a constant always has a pure imaginary
894 value, but you can form any complex value you like by adding one to a
895 real constant. This is a GNU extension; if you have an ISO C99
896 conforming C library (such as the GNU C Library), and want to construct complex
897 constants of floating type, you should include @code{<complex.h>} and
898 use the macros @code{I} or @code{_Complex_I} instead.
900 @cindex @code{__real__} keyword
901 @cindex @code{__imag__} keyword
902 To extract the real part of a complex-valued expression @var{exp}, write
903 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
904 extract the imaginary part. This is a GNU extension; for values of
905 floating type, you should use the ISO C99 functions @code{crealf},
906 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
907 @code{cimagl}, declared in @code{<complex.h>} and also provided as
908 built-in functions by GCC@.
910 @cindex complex conjugation
911 The operator @samp{~} performs complex conjugation when used on a value
912 with a complex type. This is a GNU extension; for values of
913 floating type, you should use the ISO C99 functions @code{conjf},
914 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
915 provided as built-in functions by GCC@.
917 GCC can allocate complex automatic variables in a noncontiguous
918 fashion; it's even possible for the real part to be in a register while
919 the imaginary part is on the stack (or vice versa). Only the DWARF
920 debug info format can represent this, so use of DWARF is recommended.
921 If you are using the stabs debug info format, GCC describes a noncontiguous
922 complex variable as if it were two separate variables of noncomplex type.
923 If the variable's actual name is @code{foo}, the two fictitious
924 variables are named @code{foo$real} and @code{foo$imag}. You can
925 examine and set these two fictitious variables with your debugger.
928 @section Additional Floating Types
929 @cindex additional floating types
930 @cindex @code{__float80} data type
931 @cindex @code{__float128} data type
932 @cindex @code{__ibm128} data type
933 @cindex @code{w} floating point suffix
934 @cindex @code{q} floating point suffix
935 @cindex @code{W} floating point suffix
936 @cindex @code{Q} floating point suffix
938 As an extension, GNU C supports additional floating
939 types, @code{__float80} and @code{__float128} to support 80-bit
940 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
941 Support for additional types includes the arithmetic operators:
942 add, subtract, multiply, divide; unary arithmetic operators;
943 relational operators; equality operators; and conversions to and from
944 integer and other floating types. Use a suffix @samp{w} or @samp{W}
945 in a literal constant of type @code{__float80} or type
946 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
948 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
949 types using the corresponding internal complex type, @code{XCmode} for
950 @code{__float80} type and @code{TCmode} for @code{__float128} type:
953 typedef _Complex float __attribute__((mode(TC))) _Complex128;
954 typedef _Complex float __attribute__((mode(XC))) _Complex80;
957 On PowerPC 64-bit Linux systems there are currently problems in using
958 the complex @code{__float128} type. When these problems are fixed,
962 typedef _Complex float __attribute__((mode(KC))) _Complex128;
965 Not all targets support additional floating-point types.
966 @code{__float80} and @code{__float128} types are supported on x86 and
967 IA-64 targets. The @code{__float128} type is supported on hppa HP-UX.
968 The @code{__float128} type is supported on PowerPC 64-bit Linux
969 systems by default if the vector scalar instruction set (VSX) is
972 On the PowerPC, @code{__ibm128} provides access to the IBM extended
973 double format, and it is intended to be used by the library functions
974 that handle conversions if/when long double is changed to be IEEE
975 128-bit floating point.
978 @section Half-Precision Floating Point
979 @cindex half-precision floating point
980 @cindex @code{__fp16} data type
982 On ARM targets, GCC supports half-precision (16-bit) floating point via
983 the @code{__fp16} type. You must enable this type explicitly
984 with the @option{-mfp16-format} command-line option in order to use it.
986 ARM supports two incompatible representations for half-precision
987 floating-point values. You must choose one of the representations and
988 use it consistently in your program.
990 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
991 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
992 There are 11 bits of significand precision, approximately 3
995 Specifying @option{-mfp16-format=alternative} selects the ARM
996 alternative format. This representation is similar to the IEEE
997 format, but does not support infinities or NaNs. Instead, the range
998 of exponents is extended, so that this format can represent normalized
999 values in the range of @math{2^{-14}} to 131008.
1001 The @code{__fp16} type is a storage format only. For purposes
1002 of arithmetic and other operations, @code{__fp16} values in C or C++
1003 expressions are automatically promoted to @code{float}. In addition,
1004 you cannot declare a function with a return value or parameters
1005 of type @code{__fp16}.
1007 Note that conversions from @code{double} to @code{__fp16}
1008 involve an intermediate conversion to @code{float}. Because
1009 of rounding, this can sometimes produce a different result than a
1012 ARM provides hardware support for conversions between
1013 @code{__fp16} and @code{float} values
1014 as an extension to VFP and NEON (Advanced SIMD). GCC generates
1015 code using these hardware instructions if you compile with
1016 options to select an FPU that provides them;
1017 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1018 in addition to the @option{-mfp16-format} option to select
1019 a half-precision format.
1021 Language-level support for the @code{__fp16} data type is
1022 independent of whether GCC generates code using hardware floating-point
1023 instructions. In cases where hardware support is not specified, GCC
1024 implements conversions between @code{__fp16} and @code{float} values
1028 @section Decimal Floating Types
1029 @cindex decimal floating types
1030 @cindex @code{_Decimal32} data type
1031 @cindex @code{_Decimal64} data type
1032 @cindex @code{_Decimal128} data type
1033 @cindex @code{df} integer suffix
1034 @cindex @code{dd} integer suffix
1035 @cindex @code{dl} integer suffix
1036 @cindex @code{DF} integer suffix
1037 @cindex @code{DD} integer suffix
1038 @cindex @code{DL} integer suffix
1040 As an extension, GNU C supports decimal floating types as
1041 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1042 floating types in GCC will evolve as the draft technical report changes.
1043 Calling conventions for any target might also change. Not all targets
1044 support decimal floating types.
1046 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1047 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1048 @code{float}, @code{double}, and @code{long double} whose radix is not
1049 specified by the C standard but is usually two.
1051 Support for decimal floating types includes the arithmetic operators
1052 add, subtract, multiply, divide; unary arithmetic operators;
1053 relational operators; equality operators; and conversions to and from
1054 integer and other floating types. Use a suffix @samp{df} or
1055 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1056 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1059 GCC support of decimal float as specified by the draft technical report
1064 When the value of a decimal floating type cannot be represented in the
1065 integer type to which it is being converted, the result is undefined
1066 rather than the result value specified by the draft technical report.
1069 GCC does not provide the C library functionality associated with
1070 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1071 @file{wchar.h}, which must come from a separate C library implementation.
1072 Because of this the GNU C compiler does not define macro
1073 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1074 the technical report.
1077 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1078 are supported by the DWARF debug information format.
1084 ISO C99 supports floating-point numbers written not only in the usual
1085 decimal notation, such as @code{1.55e1}, but also numbers such as
1086 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1087 supports this in C90 mode (except in some cases when strictly
1088 conforming) and in C++. In that format the
1089 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1090 mandatory. The exponent is a decimal number that indicates the power of
1091 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1098 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1099 is the same as @code{1.55e1}.
1101 Unlike for floating-point numbers in the decimal notation the exponent
1102 is always required in the hexadecimal notation. Otherwise the compiler
1103 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1104 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1105 extension for floating-point constants of type @code{float}.
1108 @section Fixed-Point Types
1109 @cindex fixed-point types
1110 @cindex @code{_Fract} data type
1111 @cindex @code{_Accum} data type
1112 @cindex @code{_Sat} data type
1113 @cindex @code{hr} fixed-suffix
1114 @cindex @code{r} fixed-suffix
1115 @cindex @code{lr} fixed-suffix
1116 @cindex @code{llr} fixed-suffix
1117 @cindex @code{uhr} fixed-suffix
1118 @cindex @code{ur} fixed-suffix
1119 @cindex @code{ulr} fixed-suffix
1120 @cindex @code{ullr} fixed-suffix
1121 @cindex @code{hk} fixed-suffix
1122 @cindex @code{k} fixed-suffix
1123 @cindex @code{lk} fixed-suffix
1124 @cindex @code{llk} fixed-suffix
1125 @cindex @code{uhk} fixed-suffix
1126 @cindex @code{uk} fixed-suffix
1127 @cindex @code{ulk} fixed-suffix
1128 @cindex @code{ullk} fixed-suffix
1129 @cindex @code{HR} fixed-suffix
1130 @cindex @code{R} fixed-suffix
1131 @cindex @code{LR} fixed-suffix
1132 @cindex @code{LLR} fixed-suffix
1133 @cindex @code{UHR} fixed-suffix
1134 @cindex @code{UR} fixed-suffix
1135 @cindex @code{ULR} fixed-suffix
1136 @cindex @code{ULLR} fixed-suffix
1137 @cindex @code{HK} fixed-suffix
1138 @cindex @code{K} fixed-suffix
1139 @cindex @code{LK} fixed-suffix
1140 @cindex @code{LLK} fixed-suffix
1141 @cindex @code{UHK} fixed-suffix
1142 @cindex @code{UK} fixed-suffix
1143 @cindex @code{ULK} fixed-suffix
1144 @cindex @code{ULLK} fixed-suffix
1146 As an extension, GNU C supports fixed-point types as
1147 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1148 types in GCC will evolve as the draft technical report changes.
1149 Calling conventions for any target might also change. Not all targets
1150 support fixed-point types.
1152 The fixed-point types are
1153 @code{short _Fract},
1156 @code{long long _Fract},
1157 @code{unsigned short _Fract},
1158 @code{unsigned _Fract},
1159 @code{unsigned long _Fract},
1160 @code{unsigned long long _Fract},
1161 @code{_Sat short _Fract},
1163 @code{_Sat long _Fract},
1164 @code{_Sat long long _Fract},
1165 @code{_Sat unsigned short _Fract},
1166 @code{_Sat unsigned _Fract},
1167 @code{_Sat unsigned long _Fract},
1168 @code{_Sat unsigned long long _Fract},
1169 @code{short _Accum},
1172 @code{long long _Accum},
1173 @code{unsigned short _Accum},
1174 @code{unsigned _Accum},
1175 @code{unsigned long _Accum},
1176 @code{unsigned long long _Accum},
1177 @code{_Sat short _Accum},
1179 @code{_Sat long _Accum},
1180 @code{_Sat long long _Accum},
1181 @code{_Sat unsigned short _Accum},
1182 @code{_Sat unsigned _Accum},
1183 @code{_Sat unsigned long _Accum},
1184 @code{_Sat unsigned long long _Accum}.
1186 Fixed-point data values contain fractional and optional integral parts.
1187 The format of fixed-point data varies and depends on the target machine.
1189 Support for fixed-point types includes:
1192 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1194 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1196 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1198 binary shift operators (@code{<<}, @code{>>})
1200 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1202 equality operators (@code{==}, @code{!=})
1204 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1205 @code{<<=}, @code{>>=})
1207 conversions to and from integer, floating-point, or fixed-point types
1210 Use a suffix in a fixed-point literal constant:
1212 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1213 @code{_Sat short _Fract}
1214 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1215 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1216 @code{_Sat long _Fract}
1217 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1218 @code{_Sat long long _Fract}
1219 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1220 @code{_Sat unsigned short _Fract}
1221 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1222 @code{_Sat unsigned _Fract}
1223 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1224 @code{_Sat unsigned long _Fract}
1225 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1226 and @code{_Sat unsigned long long _Fract}
1227 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1228 @code{_Sat short _Accum}
1229 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1230 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1231 @code{_Sat long _Accum}
1232 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1233 @code{_Sat long long _Accum}
1234 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1235 @code{_Sat unsigned short _Accum}
1236 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1237 @code{_Sat unsigned _Accum}
1238 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1239 @code{_Sat unsigned long _Accum}
1240 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1241 and @code{_Sat unsigned long long _Accum}
1244 GCC support of fixed-point types as specified by the draft technical report
1249 Pragmas to control overflow and rounding behaviors are not implemented.
1252 Fixed-point types are supported by the DWARF debug information format.
1254 @node Named Address Spaces
1255 @section Named Address Spaces
1256 @cindex Named Address Spaces
1258 As an extension, GNU C supports named address spaces as
1259 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1260 address spaces in GCC will evolve as the draft technical report
1261 changes. Calling conventions for any target might also change. At
1262 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1263 address spaces other than the generic address space.
1265 Address space identifiers may be used exactly like any other C type
1266 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1267 document for more details.
1269 @anchor{AVR Named Address Spaces}
1270 @subsection AVR Named Address Spaces
1272 On the AVR target, there are several address spaces that can be used
1273 in order to put read-only data into the flash memory and access that
1274 data by means of the special instructions @code{LPM} or @code{ELPM}
1275 needed to read from flash.
1277 Per default, any data including read-only data is located in RAM
1278 (the generic address space) so that non-generic address spaces are
1279 needed to locate read-only data in flash memory
1280 @emph{and} to generate the right instructions to access this data
1281 without using (inline) assembler code.
1285 @cindex @code{__flash} AVR Named Address Spaces
1286 The @code{__flash} qualifier locates data in the
1287 @code{.progmem.data} section. Data is read using the @code{LPM}
1288 instruction. Pointers to this address space are 16 bits wide.
1295 @cindex @code{__flash1} AVR Named Address Spaces
1296 @cindex @code{__flash2} AVR Named Address Spaces
1297 @cindex @code{__flash3} AVR Named Address Spaces
1298 @cindex @code{__flash4} AVR Named Address Spaces
1299 @cindex @code{__flash5} AVR Named Address Spaces
1300 These are 16-bit address spaces locating data in section
1301 @code{.progmem@var{N}.data} where @var{N} refers to
1302 address space @code{__flash@var{N}}.
1303 The compiler sets the @code{RAMPZ} segment register appropriately
1304 before reading data by means of the @code{ELPM} instruction.
1307 @cindex @code{__memx} AVR Named Address Spaces
1308 This is a 24-bit address space that linearizes flash and RAM:
1309 If the high bit of the address is set, data is read from
1310 RAM using the lower two bytes as RAM address.
1311 If the high bit of the address is clear, data is read from flash
1312 with @code{RAMPZ} set according to the high byte of the address.
1313 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1315 Objects in this address space are located in @code{.progmemx.data}.
1321 char my_read (const __flash char ** p)
1323 /* p is a pointer to RAM that points to a pointer to flash.
1324 The first indirection of p reads that flash pointer
1325 from RAM and the second indirection reads a char from this
1331 /* Locate array[] in flash memory */
1332 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1338 /* Return 17 by reading from flash memory */
1339 return array[array[i]];
1344 For each named address space supported by avr-gcc there is an equally
1345 named but uppercase built-in macro defined.
1346 The purpose is to facilitate testing if respective address space
1347 support is available or not:
1351 const __flash int var = 1;
1358 #include <avr/pgmspace.h> /* From AVR-LibC */
1360 const int var PROGMEM = 1;
1364 return (int) pgm_read_word (&var);
1366 #endif /* __FLASH */
1370 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1371 locates data in flash but
1372 accesses to these data read from generic address space, i.e.@:
1374 so that you need special accessors like @code{pgm_read_byte}
1375 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1376 together with attribute @code{progmem}.
1379 @b{Limitations and caveats}
1383 Reading across the 64@tie{}KiB section boundary of
1384 the @code{__flash} or @code{__flash@var{N}} address spaces
1385 shows undefined behavior. The only address space that
1386 supports reading across the 64@tie{}KiB flash segment boundaries is
1390 If you use one of the @code{__flash@var{N}} address spaces
1391 you must arrange your linker script to locate the
1392 @code{.progmem@var{N}.data} sections according to your needs.
1395 Any data or pointers to the non-generic address spaces must
1396 be qualified as @code{const}, i.e.@: as read-only data.
1397 This still applies if the data in one of these address
1398 spaces like software version number or calibration lookup table are intended to
1399 be changed after load time by, say, a boot loader. In this case
1400 the right qualification is @code{const} @code{volatile} so that the compiler
1401 must not optimize away known values or insert them
1402 as immediates into operands of instructions.
1405 The following code initializes a variable @code{pfoo}
1406 located in static storage with a 24-bit address:
1408 extern const __memx char foo;
1409 const __memx void *pfoo = &foo;
1413 Such code requires at least binutils 2.23, see
1414 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1418 @subsection M32C Named Address Spaces
1419 @cindex @code{__far} M32C Named Address Spaces
1421 On the M32C target, with the R8C and M16C CPU variants, variables
1422 qualified with @code{__far} are accessed using 32-bit addresses in
1423 order to access memory beyond the first 64@tie{}Ki bytes. If
1424 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1427 @subsection RL78 Named Address Spaces
1428 @cindex @code{__far} RL78 Named Address Spaces
1430 On the RL78 target, variables qualified with @code{__far} are accessed
1431 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1432 addresses. Non-far variables are assumed to appear in the topmost
1433 64@tie{}KiB of the address space.
1435 @subsection SPU Named Address Spaces
1436 @cindex @code{__ea} SPU Named Address Spaces
1438 On the SPU target variables may be declared as
1439 belonging to another address space by qualifying the type with the
1440 @code{__ea} address space identifier:
1447 The compiler generates special code to access the variable @code{i}.
1448 It may use runtime library
1449 support, or generate special machine instructions to access that address
1452 @subsection x86 Named Address Spaces
1453 @cindex x86 named address spaces
1455 On the x86 target, variables may be declared as being relative
1456 to the @code{%fs} or @code{%gs} segments.
1461 @cindex @code{__seg_fs} x86 named address space
1462 @cindex @code{__seg_gs} x86 named address space
1463 The object is accessed with the respective segment override prefix.
1465 The respective segment base must be set via some method specific to
1466 the operating system. Rather than require an expensive system call
1467 to retrieve the segment base, these address spaces are not considered
1468 to be subspaces of the generic (flat) address space. This means that
1469 explicit casts are required to convert pointers between these address
1470 spaces and the generic address space. In practice the application
1471 should cast to @code{uintptr_t} and apply the segment base offset
1472 that it installed previously.
1474 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1475 defined when these address spaces are supported.
1478 @cindex @code{__seg_tls} x86 named address space
1479 Some operating systems define either the @code{%fs} or @code{%gs}
1480 segment as the thread-local storage base for each thread. Objects
1481 within this address space are accessed with the appropriate
1482 segment override prefix.
1484 The pointer located at address 0 within the segment contains the
1485 offset of the segment within the generic address space. Thus this
1486 address space is considered a subspace of the generic address space,
1487 and the known segment offset is applied when converting addresses
1488 to and from the generic address space.
1490 The preprocessor symbol @code{__SEG_TLS} is defined when this
1491 address space is supported.
1496 @section Arrays of Length Zero
1497 @cindex arrays of length zero
1498 @cindex zero-length arrays
1499 @cindex length-zero arrays
1500 @cindex flexible array members
1502 Zero-length arrays are allowed in GNU C@. They are very useful as the
1503 last element of a structure that is really a header for a variable-length
1512 struct line *thisline = (struct line *)
1513 malloc (sizeof (struct line) + this_length);
1514 thisline->length = this_length;
1517 In ISO C90, you would have to give @code{contents} a length of 1, which
1518 means either you waste space or complicate the argument to @code{malloc}.
1520 In ISO C99, you would use a @dfn{flexible array member}, which is
1521 slightly different in syntax and semantics:
1525 Flexible array members are written as @code{contents[]} without
1529 Flexible array members have incomplete type, and so the @code{sizeof}
1530 operator may not be applied. As a quirk of the original implementation
1531 of zero-length arrays, @code{sizeof} evaluates to zero.
1534 Flexible array members may only appear as the last member of a
1535 @code{struct} that is otherwise non-empty.
1538 A structure containing a flexible array member, or a union containing
1539 such a structure (possibly recursively), may not be a member of a
1540 structure or an element of an array. (However, these uses are
1541 permitted by GCC as extensions.)
1544 Non-empty initialization of zero-length
1545 arrays is treated like any case where there are more initializer
1546 elements than the array holds, in that a suitable warning about ``excess
1547 elements in array'' is given, and the excess elements (all of them, in
1548 this case) are ignored.
1550 GCC allows static initialization of flexible array members.
1551 This is equivalent to defining a new structure containing the original
1552 structure followed by an array of sufficient size to contain the data.
1553 E.g.@: in the following, @code{f1} is constructed as if it were declared
1559 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1562 struct f1 f1; int data[3];
1563 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1567 The convenience of this extension is that @code{f1} has the desired
1568 type, eliminating the need to consistently refer to @code{f2.f1}.
1570 This has symmetry with normal static arrays, in that an array of
1571 unknown size is also written with @code{[]}.
1573 Of course, this extension only makes sense if the extra data comes at
1574 the end of a top-level object, as otherwise we would be overwriting
1575 data at subsequent offsets. To avoid undue complication and confusion
1576 with initialization of deeply nested arrays, we simply disallow any
1577 non-empty initialization except when the structure is the top-level
1578 object. For example:
1581 struct foo @{ int x; int y[]; @};
1582 struct bar @{ struct foo z; @};
1584 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1585 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1586 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1587 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1590 @node Empty Structures
1591 @section Structures with No Members
1592 @cindex empty structures
1593 @cindex zero-size structures
1595 GCC permits a C structure to have no members:
1602 The structure has size zero. In C++, empty structures are part
1603 of the language. G++ treats empty structures as if they had a single
1604 member of type @code{char}.
1606 @node Variable Length
1607 @section Arrays of Variable Length
1608 @cindex variable-length arrays
1609 @cindex arrays of variable length
1612 Variable-length automatic arrays are allowed in ISO C99, and as an
1613 extension GCC accepts them in C90 mode and in C++. These arrays are
1614 declared like any other automatic arrays, but with a length that is not
1615 a constant expression. The storage is allocated at the point of
1616 declaration and deallocated when the block scope containing the declaration
1622 concat_fopen (char *s1, char *s2, char *mode)
1624 char str[strlen (s1) + strlen (s2) + 1];
1627 return fopen (str, mode);
1631 @cindex scope of a variable length array
1632 @cindex variable-length array scope
1633 @cindex deallocating variable length arrays
1634 Jumping or breaking out of the scope of the array name deallocates the
1635 storage. Jumping into the scope is not allowed; you get an error
1638 @cindex variable-length array in a structure
1639 As an extension, GCC accepts variable-length arrays as a member of
1640 a structure or a union. For example:
1646 struct S @{ int x[n]; @};
1650 @cindex @code{alloca} vs variable-length arrays
1651 You can use the function @code{alloca} to get an effect much like
1652 variable-length arrays. The function @code{alloca} is available in
1653 many other C implementations (but not in all). On the other hand,
1654 variable-length arrays are more elegant.
1656 There are other differences between these two methods. Space allocated
1657 with @code{alloca} exists until the containing @emph{function} returns.
1658 The space for a variable-length array is deallocated as soon as the array
1659 name's scope ends, unless you also use @code{alloca} in this scope.
1661 You can also use variable-length arrays as arguments to functions:
1665 tester (int len, char data[len][len])
1671 The length of an array is computed once when the storage is allocated
1672 and is remembered for the scope of the array in case you access it with
1675 If you want to pass the array first and the length afterward, you can
1676 use a forward declaration in the parameter list---another GNU extension.
1680 tester (int len; char data[len][len], int len)
1686 @cindex parameter forward declaration
1687 The @samp{int len} before the semicolon is a @dfn{parameter forward
1688 declaration}, and it serves the purpose of making the name @code{len}
1689 known when the declaration of @code{data} is parsed.
1691 You can write any number of such parameter forward declarations in the
1692 parameter list. They can be separated by commas or semicolons, but the
1693 last one must end with a semicolon, which is followed by the ``real''
1694 parameter declarations. Each forward declaration must match a ``real''
1695 declaration in parameter name and data type. ISO C99 does not support
1696 parameter forward declarations.
1698 @node Variadic Macros
1699 @section Macros with a Variable Number of Arguments.
1700 @cindex variable number of arguments
1701 @cindex macro with variable arguments
1702 @cindex rest argument (in macro)
1703 @cindex variadic macros
1705 In the ISO C standard of 1999, a macro can be declared to accept a
1706 variable number of arguments much as a function can. The syntax for
1707 defining the macro is similar to that of a function. Here is an
1711 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1715 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1716 such a macro, it represents the zero or more tokens until the closing
1717 parenthesis that ends the invocation, including any commas. This set of
1718 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1719 wherever it appears. See the CPP manual for more information.
1721 GCC has long supported variadic macros, and used a different syntax that
1722 allowed you to give a name to the variable arguments just like any other
1723 argument. Here is an example:
1726 #define debug(format, args...) fprintf (stderr, format, args)
1730 This is in all ways equivalent to the ISO C example above, but arguably
1731 more readable and descriptive.
1733 GNU CPP has two further variadic macro extensions, and permits them to
1734 be used with either of the above forms of macro definition.
1736 In standard C, you are not allowed to leave the variable argument out
1737 entirely; but you are allowed to pass an empty argument. For example,
1738 this invocation is invalid in ISO C, because there is no comma after
1745 GNU CPP permits you to completely omit the variable arguments in this
1746 way. In the above examples, the compiler would complain, though since
1747 the expansion of the macro still has the extra comma after the format
1750 To help solve this problem, CPP behaves specially for variable arguments
1751 used with the token paste operator, @samp{##}. If instead you write
1754 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1758 and if the variable arguments are omitted or empty, the @samp{##}
1759 operator causes the preprocessor to remove the comma before it. If you
1760 do provide some variable arguments in your macro invocation, GNU CPP
1761 does not complain about the paste operation and instead places the
1762 variable arguments after the comma. Just like any other pasted macro
1763 argument, these arguments are not macro expanded.
1765 @node Escaped Newlines
1766 @section Slightly Looser Rules for Escaped Newlines
1767 @cindex escaped newlines
1768 @cindex newlines (escaped)
1770 The preprocessor treatment of escaped newlines is more relaxed
1771 than that specified by the C90 standard, which requires the newline
1772 to immediately follow a backslash.
1773 GCC's implementation allows whitespace in the form
1774 of spaces, horizontal and vertical tabs, and form feeds between the
1775 backslash and the subsequent newline. The preprocessor issues a
1776 warning, but treats it as a valid escaped newline and combines the two
1777 lines to form a single logical line. This works within comments and
1778 tokens, as well as between tokens. Comments are @emph{not} treated as
1779 whitespace for the purposes of this relaxation, since they have not
1780 yet been replaced with spaces.
1783 @section Non-Lvalue Arrays May Have Subscripts
1784 @cindex subscripting
1785 @cindex arrays, non-lvalue
1787 @cindex subscripting and function values
1788 In ISO C99, arrays that are not lvalues still decay to pointers, and
1789 may be subscripted, although they may not be modified or used after
1790 the next sequence point and the unary @samp{&} operator may not be
1791 applied to them. As an extension, GNU C allows such arrays to be
1792 subscripted in C90 mode, though otherwise they do not decay to
1793 pointers outside C99 mode. For example,
1794 this is valid in GNU C though not valid in C90:
1798 struct foo @{int a[4];@};
1804 return f().a[index];
1810 @section Arithmetic on @code{void}- and Function-Pointers
1811 @cindex void pointers, arithmetic
1812 @cindex void, size of pointer to
1813 @cindex function pointers, arithmetic
1814 @cindex function, size of pointer to
1816 In GNU C, addition and subtraction operations are supported on pointers to
1817 @code{void} and on pointers to functions. This is done by treating the
1818 size of a @code{void} or of a function as 1.
1820 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1821 and on function types, and returns 1.
1823 @opindex Wpointer-arith
1824 The option @option{-Wpointer-arith} requests a warning if these extensions
1827 @node Pointers to Arrays
1828 @section Pointers to Arrays with Qualifiers Work as Expected
1829 @cindex pointers to arrays
1830 @cindex const qualifier
1832 In GNU C, pointers to arrays with qualifiers work similar to pointers
1833 to other qualified types. For example, a value of type @code{int (*)[5]}
1834 can be used to initialize a variable of type @code{const int (*)[5]}.
1835 These types are incompatible in ISO C because the @code{const} qualifier
1836 is formally attached to the element type of the array and not the
1841 transpose (int N, int M, double out[M][N], const double in[N][M]);
1845 transpose(3, 2, y, x);
1849 @section Non-Constant Initializers
1850 @cindex initializers, non-constant
1851 @cindex non-constant initializers
1853 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1854 automatic variable are not required to be constant expressions in GNU C@.
1855 Here is an example of an initializer with run-time varying elements:
1858 foo (float f, float g)
1860 float beat_freqs[2] = @{ f-g, f+g @};
1865 @node Compound Literals
1866 @section Compound Literals
1867 @cindex constructor expressions
1868 @cindex initializations in expressions
1869 @cindex structures, constructor expression
1870 @cindex expressions, constructor
1871 @cindex compound literals
1872 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1874 ISO C99 supports compound literals. A compound literal looks like
1875 a cast containing an initializer. Its value is an object of the
1876 type specified in the cast, containing the elements specified in
1877 the initializer; it is an lvalue. As an extension, GCC supports
1878 compound literals in C90 mode and in C++, though the semantics are
1879 somewhat different in C++.
1881 Usually, the specified type is a structure. Assume that
1882 @code{struct foo} and @code{structure} are declared as shown:
1885 struct foo @{int a; char b[2];@} structure;
1889 Here is an example of constructing a @code{struct foo} with a compound literal:
1892 structure = ((struct foo) @{x + y, 'a', 0@});
1896 This is equivalent to writing the following:
1900 struct foo temp = @{x + y, 'a', 0@};
1905 You can also construct an array, though this is dangerous in C++, as
1906 explained below. If all the elements of the compound literal are
1907 (made up of) simple constant expressions, suitable for use in
1908 initializers of objects of static storage duration, then the compound
1909 literal can be coerced to a pointer to its first element and used in
1910 such an initializer, as shown here:
1913 char **foo = (char *[]) @{ "x", "y", "z" @};
1916 Compound literals for scalar types and union types are
1917 also allowed, but then the compound literal is equivalent
1920 As a GNU extension, GCC allows initialization of objects with static storage
1921 duration by compound literals (which is not possible in ISO C99, because
1922 the initializer is not a constant).
1923 It is handled as if the object is initialized only with the bracket
1924 enclosed list if the types of the compound literal and the object match.
1925 The initializer list of the compound literal must be constant.
1926 If the object being initialized has array type of unknown size, the size is
1927 determined by compound literal size.
1930 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1931 static int y[] = (int []) @{1, 2, 3@};
1932 static int z[] = (int [3]) @{1@};
1936 The above lines are equivalent to the following:
1938 static struct foo x = @{1, 'a', 'b'@};
1939 static int y[] = @{1, 2, 3@};
1940 static int z[] = @{1, 0, 0@};
1943 In C, a compound literal designates an unnamed object with static or
1944 automatic storage duration. In C++, a compound literal designates a
1945 temporary object, which only lives until the end of its
1946 full-expression. As a result, well-defined C code that takes the
1947 address of a subobject of a compound literal can be undefined in C++,
1948 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1949 For instance, if the array compound literal example above appeared
1950 inside a function, any subsequent use of @samp{foo} in C++ has
1951 undefined behavior because the lifetime of the array ends after the
1952 declaration of @samp{foo}.
1954 As an optimization, the C++ compiler sometimes gives array compound
1955 literals longer lifetimes: when the array either appears outside a
1956 function or has const-qualified type. If @samp{foo} and its
1957 initializer had elements of @samp{char *const} type rather than
1958 @samp{char *}, or if @samp{foo} were a global variable, the array
1959 would have static storage duration. But it is probably safest just to
1960 avoid the use of array compound literals in code compiled as C++.
1962 @node Designated Inits
1963 @section Designated Initializers
1964 @cindex initializers with labeled elements
1965 @cindex labeled elements in initializers
1966 @cindex case labels in initializers
1967 @cindex designated initializers
1969 Standard C90 requires the elements of an initializer to appear in a fixed
1970 order, the same as the order of the elements in the array or structure
1973 In ISO C99 you can give the elements in any order, specifying the array
1974 indices or structure field names they apply to, and GNU C allows this as
1975 an extension in C90 mode as well. This extension is not
1976 implemented in GNU C++.
1978 To specify an array index, write
1979 @samp{[@var{index}] =} before the element value. For example,
1982 int a[6] = @{ [4] = 29, [2] = 15 @};
1989 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1993 The index values must be constant expressions, even if the array being
1994 initialized is automatic.
1996 An alternative syntax for this that has been obsolete since GCC 2.5 but
1997 GCC still accepts is to write @samp{[@var{index}]} before the element
1998 value, with no @samp{=}.
2000 To initialize a range of elements to the same value, write
2001 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2002 extension. For example,
2005 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2009 If the value in it has side-effects, the side-effects happen only once,
2010 not for each initialized field by the range initializer.
2013 Note that the length of the array is the highest value specified
2016 In a structure initializer, specify the name of a field to initialize
2017 with @samp{.@var{fieldname} =} before the element value. For example,
2018 given the following structure,
2021 struct point @{ int x, y; @};
2025 the following initialization
2028 struct point p = @{ .y = yvalue, .x = xvalue @};
2035 struct point p = @{ xvalue, yvalue @};
2038 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2039 @samp{@var{fieldname}:}, as shown here:
2042 struct point p = @{ y: yvalue, x: xvalue @};
2045 Omitted field members are implicitly initialized the same as objects
2046 that have static storage duration.
2049 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2050 @dfn{designator}. You can also use a designator (or the obsolete colon
2051 syntax) when initializing a union, to specify which element of the union
2052 should be used. For example,
2055 union foo @{ int i; double d; @};
2057 union foo f = @{ .d = 4 @};
2061 converts 4 to a @code{double} to store it in the union using
2062 the second element. By contrast, casting 4 to type @code{union foo}
2063 stores it into the union as the integer @code{i}, since it is
2064 an integer. (@xref{Cast to Union}.)
2066 You can combine this technique of naming elements with ordinary C
2067 initialization of successive elements. Each initializer element that
2068 does not have a designator applies to the next consecutive element of the
2069 array or structure. For example,
2072 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2079 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2082 Labeling the elements of an array initializer is especially useful
2083 when the indices are characters or belong to an @code{enum} type.
2088 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2089 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2092 @cindex designator lists
2093 You can also write a series of @samp{.@var{fieldname}} and
2094 @samp{[@var{index}]} designators before an @samp{=} to specify a
2095 nested subobject to initialize; the list is taken relative to the
2096 subobject corresponding to the closest surrounding brace pair. For
2097 example, with the @samp{struct point} declaration above:
2100 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2104 If the same field is initialized multiple times, it has the value from
2105 the last initialization. If any such overridden initialization has
2106 side-effect, it is unspecified whether the side-effect happens or not.
2107 Currently, GCC discards them and issues a warning.
2110 @section Case Ranges
2112 @cindex ranges in case statements
2114 You can specify a range of consecutive values in a single @code{case} label,
2118 case @var{low} ... @var{high}:
2122 This has the same effect as the proper number of individual @code{case}
2123 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2125 This feature is especially useful for ranges of ASCII character codes:
2131 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2132 it may be parsed wrong when you use it with integer values. For example,
2147 @section Cast to a Union Type
2148 @cindex cast to a union
2149 @cindex union, casting to a
2151 A cast to union type is similar to other casts, except that the type
2152 specified is a union type. You can specify the type either with
2153 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2154 a constructor, not a cast, and hence does not yield an lvalue like
2155 normal casts. (@xref{Compound Literals}.)
2157 The types that may be cast to the union type are those of the members
2158 of the union. Thus, given the following union and variables:
2161 union foo @{ int i; double d; @};
2167 both @code{x} and @code{y} can be cast to type @code{union foo}.
2169 Using the cast as the right-hand side of an assignment to a variable of
2170 union type is equivalent to storing in a member of the union:
2175 u = (union foo) x @equiv{} u.i = x
2176 u = (union foo) y @equiv{} u.d = y
2179 You can also use the union cast as a function argument:
2182 void hack (union foo);
2184 hack ((union foo) x);
2187 @node Mixed Declarations
2188 @section Mixed Declarations and Code
2189 @cindex mixed declarations and code
2190 @cindex declarations, mixed with code
2191 @cindex code, mixed with declarations
2193 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2194 within compound statements. As an extension, GNU C also allows this in
2195 C90 mode. For example, you could do:
2204 Each identifier is visible from where it is declared until the end of
2205 the enclosing block.
2207 @node Function Attributes
2208 @section Declaring Attributes of Functions
2209 @cindex function attributes
2210 @cindex declaring attributes of functions
2211 @cindex @code{volatile} applied to function
2212 @cindex @code{const} applied to function
2214 In GNU C, you can use function attributes to declare certain things
2215 about functions called in your program which help the compiler
2216 optimize calls and check your code more carefully. For example, you
2217 can use attributes to declare that a function never returns
2218 (@code{noreturn}), returns a value depending only on its arguments
2219 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2221 You can also use attributes to control memory placement, code
2222 generation options or call/return conventions within the function
2223 being annotated. Many of these attributes are target-specific. For
2224 example, many targets support attributes for defining interrupt
2225 handler functions, which typically must follow special register usage
2226 and return conventions.
2228 Function attributes are introduced by the @code{__attribute__} keyword
2229 on a declaration, followed by an attribute specification inside double
2230 parentheses. You can specify multiple attributes in a declaration by
2231 separating them by commas within the double parentheses or by
2232 immediately following an attribute declaration with another attribute
2233 declaration. @xref{Attribute Syntax}, for the exact rules on
2234 attribute syntax and placement.
2236 GCC also supports attributes on
2237 variable declarations (@pxref{Variable Attributes}),
2238 labels (@pxref{Label Attributes}),
2239 enumerators (@pxref{Enumerator Attributes}),
2240 and types (@pxref{Type Attributes}).
2242 There is some overlap between the purposes of attributes and pragmas
2243 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2244 found convenient to use @code{__attribute__} to achieve a natural
2245 attachment of attributes to their corresponding declarations, whereas
2246 @code{#pragma} is of use for compatibility with other compilers
2247 or constructs that do not naturally form part of the grammar.
2249 In addition to the attributes documented here,
2250 GCC plugins may provide their own attributes.
2253 * Common Function Attributes::
2254 * AArch64 Function Attributes::
2255 * ARC Function Attributes::
2256 * ARM Function Attributes::
2257 * AVR Function Attributes::
2258 * Blackfin Function Attributes::
2259 * CR16 Function Attributes::
2260 * Epiphany Function Attributes::
2261 * H8/300 Function Attributes::
2262 * IA-64 Function Attributes::
2263 * M32C Function Attributes::
2264 * M32R/D Function Attributes::
2265 * m68k Function Attributes::
2266 * MCORE Function Attributes::
2267 * MeP Function Attributes::
2268 * MicroBlaze Function Attributes::
2269 * Microsoft Windows Function Attributes::
2270 * MIPS Function Attributes::
2271 * MSP430 Function Attributes::
2272 * NDS32 Function Attributes::
2273 * Nios II Function Attributes::
2274 * Nvidia PTX Function Attributes::
2275 * PowerPC Function Attributes::
2276 * RL78 Function Attributes::
2277 * RX Function Attributes::
2278 * S/390 Function Attributes::
2279 * SH Function Attributes::
2280 * SPU Function Attributes::
2281 * Symbian OS Function Attributes::
2282 * V850 Function Attributes::
2283 * Visium Function Attributes::
2284 * x86 Function Attributes::
2285 * Xstormy16 Function Attributes::
2288 @node Common Function Attributes
2289 @subsection Common Function Attributes
2291 The following attributes are supported on most targets.
2294 @c Keep this table alphabetized by attribute name. Treat _ as space.
2296 @item alias ("@var{target}")
2297 @cindex @code{alias} function attribute
2298 The @code{alias} attribute causes the declaration to be emitted as an
2299 alias for another symbol, which must be specified. For instance,
2302 void __f () @{ /* @r{Do something.} */; @}
2303 void f () __attribute__ ((weak, alias ("__f")));
2307 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2308 mangled name for the target must be used. It is an error if @samp{__f}
2309 is not defined in the same translation unit.
2311 This attribute requires assembler and object file support,
2312 and may not be available on all targets.
2314 @item aligned (@var{alignment})
2315 @cindex @code{aligned} function attribute
2316 This attribute specifies a minimum alignment for the function,
2319 You cannot use this attribute to decrease the alignment of a function,
2320 only to increase it. However, when you explicitly specify a function
2321 alignment this overrides the effect of the
2322 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2325 Note that the effectiveness of @code{aligned} attributes may be
2326 limited by inherent limitations in your linker. On many systems, the
2327 linker is only able to arrange for functions to be aligned up to a
2328 certain maximum alignment. (For some linkers, the maximum supported
2329 alignment may be very very small.) See your linker documentation for
2330 further information.
2332 The @code{aligned} attribute can also be used for variables and fields
2333 (@pxref{Variable Attributes}.)
2336 @cindex @code{alloc_align} function attribute
2337 The @code{alloc_align} attribute is used to tell the compiler that the
2338 function return value points to memory, where the returned pointer minimum
2339 alignment is given by one of the functions parameters. GCC uses this
2340 information to improve pointer alignment analysis.
2342 The function parameter denoting the allocated alignment is specified by
2343 one integer argument, whose number is the argument of the attribute.
2344 Argument numbering starts at one.
2349 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2353 declares that @code{my_memalign} returns memory with minimum alignment
2354 given by parameter 1.
2357 @cindex @code{alloc_size} function attribute
2358 The @code{alloc_size} attribute is used to tell the compiler that the
2359 function return value points to memory, where the size is given by
2360 one or two of the functions parameters. GCC uses this
2361 information to improve the correctness of @code{__builtin_object_size}.
2363 The function parameter(s) denoting the allocated size are specified by
2364 one or two integer arguments supplied to the attribute. The allocated size
2365 is either the value of the single function argument specified or the product
2366 of the two function arguments specified. Argument numbering starts at
2372 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2373 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2377 declares that @code{my_calloc} returns memory of the size given by
2378 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2379 of the size given by parameter 2.
2382 @cindex @code{always_inline} function attribute
2383 Generally, functions are not inlined unless optimization is specified.
2384 For functions declared inline, this attribute inlines the function
2385 independent of any restrictions that otherwise apply to inlining.
2386 Failure to inline such a function is diagnosed as an error.
2387 Note that if such a function is called indirectly the compiler may
2388 or may not inline it depending on optimization level and a failure
2389 to inline an indirect call may or may not be diagnosed.
2392 @cindex @code{artificial} function attribute
2393 This attribute is useful for small inline wrappers that if possible
2394 should appear during debugging as a unit. Depending on the debug
2395 info format it either means marking the function as artificial
2396 or using the caller location for all instructions within the inlined
2399 @item assume_aligned
2400 @cindex @code{assume_aligned} function attribute
2401 The @code{assume_aligned} attribute is used to tell the compiler that the
2402 function return value points to memory, where the returned pointer minimum
2403 alignment is given by the first argument.
2404 If the attribute has two arguments, the second argument is misalignment offset.
2409 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2410 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2414 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2415 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2418 @item bnd_instrument
2419 @cindex @code{bnd_instrument} function attribute
2420 The @code{bnd_instrument} attribute on functions is used to inform the
2421 compiler that the function should be instrumented when compiled
2422 with the @option{-fchkp-instrument-marked-only} option.
2425 @cindex @code{bnd_legacy} function attribute
2426 @cindex Pointer Bounds Checker attributes
2427 The @code{bnd_legacy} attribute on functions is used to inform the
2428 compiler that the function should not be instrumented when compiled
2429 with the @option{-fcheck-pointer-bounds} option.
2432 @cindex @code{cold} function attribute
2433 The @code{cold} attribute on functions is used to inform the compiler that
2434 the function is unlikely to be executed. The function is optimized for
2435 size rather than speed and on many targets it is placed into a special
2436 subsection of the text section so all cold functions appear close together,
2437 improving code locality of non-cold parts of program. The paths leading
2438 to calls of cold functions within code are marked as unlikely by the branch
2439 prediction mechanism. It is thus useful to mark functions used to handle
2440 unlikely conditions, such as @code{perror}, as cold to improve optimization
2441 of hot functions that do call marked functions in rare occasions.
2443 When profile feedback is available, via @option{-fprofile-use}, cold functions
2444 are automatically detected and this attribute is ignored.
2447 @cindex @code{const} function attribute
2448 @cindex functions that have no side effects
2449 Many functions do not examine any values except their arguments, and
2450 have no effects except the return value. Basically this is just slightly
2451 more strict class than the @code{pure} attribute below, since function is not
2452 allowed to read global memory.
2454 @cindex pointer arguments
2455 Note that a function that has pointer arguments and examines the data
2456 pointed to must @emph{not} be declared @code{const}. Likewise, a
2457 function that calls a non-@code{const} function usually must not be
2458 @code{const}. It does not make sense for a @code{const} function to
2463 @itemx constructor (@var{priority})
2464 @itemx destructor (@var{priority})
2465 @cindex @code{constructor} function attribute
2466 @cindex @code{destructor} function attribute
2467 The @code{constructor} attribute causes the function to be called
2468 automatically before execution enters @code{main ()}. Similarly, the
2469 @code{destructor} attribute causes the function to be called
2470 automatically after @code{main ()} completes or @code{exit ()} is
2471 called. Functions with these attributes are useful for
2472 initializing data that is used implicitly during the execution of
2475 You may provide an optional integer priority to control the order in
2476 which constructor and destructor functions are run. A constructor
2477 with a smaller priority number runs before a constructor with a larger
2478 priority number; the opposite relationship holds for destructors. So,
2479 if you have a constructor that allocates a resource and a destructor
2480 that deallocates the same resource, both functions typically have the
2481 same priority. The priorities for constructor and destructor
2482 functions are the same as those specified for namespace-scope C++
2483 objects (@pxref{C++ Attributes}).
2485 These attributes are not currently implemented for Objective-C@.
2488 @itemx deprecated (@var{msg})
2489 @cindex @code{deprecated} function attribute
2490 The @code{deprecated} attribute results in a warning if the function
2491 is used anywhere in the source file. This is useful when identifying
2492 functions that are expected to be removed in a future version of a
2493 program. The warning also includes the location of the declaration
2494 of the deprecated function, to enable users to easily find further
2495 information about why the function is deprecated, or what they should
2496 do instead. Note that the warnings only occurs for uses:
2499 int old_fn () __attribute__ ((deprecated));
2501 int (*fn_ptr)() = old_fn;
2505 results in a warning on line 3 but not line 2. The optional @var{msg}
2506 argument, which must be a string, is printed in the warning if
2509 The @code{deprecated} attribute can also be used for variables and
2510 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2512 @item error ("@var{message}")
2513 @itemx warning ("@var{message}")
2514 @cindex @code{error} function attribute
2515 @cindex @code{warning} function attribute
2516 If the @code{error} or @code{warning} attribute
2517 is used on a function declaration and a call to such a function
2518 is not eliminated through dead code elimination or other optimizations,
2519 an error or warning (respectively) that includes @var{message} is diagnosed.
2521 for compile-time checking, especially together with @code{__builtin_constant_p}
2522 and inline functions where checking the inline function arguments is not
2523 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2525 While it is possible to leave the function undefined and thus invoke
2526 a link failure (to define the function with
2527 a message in @code{.gnu.warning*} section),
2528 when using these attributes the problem is diagnosed
2529 earlier and with exact location of the call even in presence of inline
2530 functions or when not emitting debugging information.
2532 @item externally_visible
2533 @cindex @code{externally_visible} function attribute
2534 This attribute, attached to a global variable or function, nullifies
2535 the effect of the @option{-fwhole-program} command-line option, so the
2536 object remains visible outside the current compilation unit.
2538 If @option{-fwhole-program} is used together with @option{-flto} and
2539 @command{gold} is used as the linker plugin,
2540 @code{externally_visible} attributes are automatically added to functions
2541 (not variable yet due to a current @command{gold} issue)
2542 that are accessed outside of LTO objects according to resolution file
2543 produced by @command{gold}.
2544 For other linkers that cannot generate resolution file,
2545 explicit @code{externally_visible} attributes are still necessary.
2548 @cindex @code{flatten} function attribute
2549 Generally, inlining into a function is limited. For a function marked with
2550 this attribute, every call inside this function is inlined, if possible.
2551 Whether the function itself is considered for inlining depends on its size and
2552 the current inlining parameters.
2554 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2555 @cindex @code{format} function attribute
2556 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2558 The @code{format} attribute specifies that a function takes @code{printf},
2559 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2560 should be type-checked against a format string. For example, the
2565 my_printf (void *my_object, const char *my_format, ...)
2566 __attribute__ ((format (printf, 2, 3)));
2570 causes the compiler to check the arguments in calls to @code{my_printf}
2571 for consistency with the @code{printf} style format string argument
2574 The parameter @var{archetype} determines how the format string is
2575 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2576 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2577 @code{strfmon}. (You can also use @code{__printf__},
2578 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2579 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2580 @code{ms_strftime} are also present.
2581 @var{archetype} values such as @code{printf} refer to the formats accepted
2582 by the system's C runtime library,
2583 while values prefixed with @samp{gnu_} always refer
2584 to the formats accepted by the GNU C Library. On Microsoft Windows
2585 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2586 @file{msvcrt.dll} library.
2587 The parameter @var{string-index}
2588 specifies which argument is the format string argument (starting
2589 from 1), while @var{first-to-check} is the number of the first
2590 argument to check against the format string. For functions
2591 where the arguments are not available to be checked (such as
2592 @code{vprintf}), specify the third parameter as zero. In this case the
2593 compiler only checks the format string for consistency. For
2594 @code{strftime} formats, the third parameter is required to be zero.
2595 Since non-static C++ methods have an implicit @code{this} argument, the
2596 arguments of such methods should be counted from two, not one, when
2597 giving values for @var{string-index} and @var{first-to-check}.
2599 In the example above, the format string (@code{my_format}) is the second
2600 argument of the function @code{my_print}, and the arguments to check
2601 start with the third argument, so the correct parameters for the format
2602 attribute are 2 and 3.
2604 @opindex ffreestanding
2605 @opindex fno-builtin
2606 The @code{format} attribute allows you to identify your own functions
2607 that take format strings as arguments, so that GCC can check the
2608 calls to these functions for errors. The compiler always (unless
2609 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2610 for the standard library functions @code{printf}, @code{fprintf},
2611 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2612 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2613 warnings are requested (using @option{-Wformat}), so there is no need to
2614 modify the header file @file{stdio.h}. In C99 mode, the functions
2615 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2616 @code{vsscanf} are also checked. Except in strictly conforming C
2617 standard modes, the X/Open function @code{strfmon} is also checked as
2618 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2619 @xref{C Dialect Options,,Options Controlling C Dialect}.
2621 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2622 recognized in the same context. Declarations including these format attributes
2623 are parsed for correct syntax, however the result of checking of such format
2624 strings is not yet defined, and is not carried out by this version of the
2627 The target may also provide additional types of format checks.
2628 @xref{Target Format Checks,,Format Checks Specific to Particular
2631 @item format_arg (@var{string-index})
2632 @cindex @code{format_arg} function attribute
2633 @opindex Wformat-nonliteral
2634 The @code{format_arg} attribute specifies that a function takes a format
2635 string for a @code{printf}, @code{scanf}, @code{strftime} or
2636 @code{strfmon} style function and modifies it (for example, to translate
2637 it into another language), so the result can be passed to a
2638 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2639 function (with the remaining arguments to the format function the same
2640 as they would have been for the unmodified string). For example, the
2645 my_dgettext (char *my_domain, const char *my_format)
2646 __attribute__ ((format_arg (2)));
2650 causes the compiler to check the arguments in calls to a @code{printf},
2651 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2652 format string argument is a call to the @code{my_dgettext} function, for
2653 consistency with the format string argument @code{my_format}. If the
2654 @code{format_arg} attribute had not been specified, all the compiler
2655 could tell in such calls to format functions would be that the format
2656 string argument is not constant; this would generate a warning when
2657 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2658 without the attribute.
2660 The parameter @var{string-index} specifies which argument is the format
2661 string argument (starting from one). Since non-static C++ methods have
2662 an implicit @code{this} argument, the arguments of such methods should
2663 be counted from two.
2665 The @code{format_arg} attribute allows you to identify your own
2666 functions that modify format strings, so that GCC can check the
2667 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2668 type function whose operands are a call to one of your own function.
2669 The compiler always treats @code{gettext}, @code{dgettext}, and
2670 @code{dcgettext} in this manner except when strict ISO C support is
2671 requested by @option{-ansi} or an appropriate @option{-std} option, or
2672 @option{-ffreestanding} or @option{-fno-builtin}
2673 is used. @xref{C Dialect Options,,Options
2674 Controlling C Dialect}.
2676 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2677 @code{NSString} reference for compatibility with the @code{format} attribute
2680 The target may also allow additional types in @code{format-arg} attributes.
2681 @xref{Target Format Checks,,Format Checks Specific to Particular
2685 @cindex @code{gnu_inline} function attribute
2686 This attribute should be used with a function that is also declared
2687 with the @code{inline} keyword. It directs GCC to treat the function
2688 as if it were defined in gnu90 mode even when compiling in C99 or
2691 If the function is declared @code{extern}, then this definition of the
2692 function is used only for inlining. In no case is the function
2693 compiled as a standalone function, not even if you take its address
2694 explicitly. Such an address becomes an external reference, as if you
2695 had only declared the function, and had not defined it. This has
2696 almost the effect of a macro. The way to use this is to put a
2697 function definition in a header file with this attribute, and put
2698 another copy of the function, without @code{extern}, in a library
2699 file. The definition in the header file causes most calls to the
2700 function to be inlined. If any uses of the function remain, they
2701 refer to the single copy in the library. Note that the two
2702 definitions of the functions need not be precisely the same, although
2703 if they do not have the same effect your program may behave oddly.
2705 In C, if the function is neither @code{extern} nor @code{static}, then
2706 the function is compiled as a standalone function, as well as being
2707 inlined where possible.
2709 This is how GCC traditionally handled functions declared
2710 @code{inline}. Since ISO C99 specifies a different semantics for
2711 @code{inline}, this function attribute is provided as a transition
2712 measure and as a useful feature in its own right. This attribute is
2713 available in GCC 4.1.3 and later. It is available if either of the
2714 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2715 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2716 Function is As Fast As a Macro}.
2718 In C++, this attribute does not depend on @code{extern} in any way,
2719 but it still requires the @code{inline} keyword to enable its special
2723 @cindex @code{hot} function attribute
2724 The @code{hot} attribute on a function is used to inform the compiler that
2725 the function is a hot spot of the compiled program. The function is
2726 optimized more aggressively and on many targets it is placed into a special
2727 subsection of the text section so all hot functions appear close together,
2730 When profile feedback is available, via @option{-fprofile-use}, hot functions
2731 are automatically detected and this attribute is ignored.
2733 @item ifunc ("@var{resolver}")
2734 @cindex @code{ifunc} function attribute
2735 @cindex indirect functions
2736 @cindex functions that are dynamically resolved
2737 The @code{ifunc} attribute is used to mark a function as an indirect
2738 function using the STT_GNU_IFUNC symbol type extension to the ELF
2739 standard. This allows the resolution of the symbol value to be
2740 determined dynamically at load time, and an optimized version of the
2741 routine can be selected for the particular processor or other system
2742 characteristics determined then. To use this attribute, first define
2743 the implementation functions available, and a resolver function that
2744 returns a pointer to the selected implementation function. The
2745 implementation functions' declarations must match the API of the
2746 function being implemented, the resolver's declaration is be a
2747 function returning pointer to void function returning void:
2750 void *my_memcpy (void *dst, const void *src, size_t len)
2755 static void (*resolve_memcpy (void)) (void)
2757 return my_memcpy; // we'll just always select this routine
2762 The exported header file declaring the function the user calls would
2766 extern void *memcpy (void *, const void *, size_t);
2770 allowing the user to call this as a regular function, unaware of the
2771 implementation. Finally, the indirect function needs to be defined in
2772 the same translation unit as the resolver function:
2775 void *memcpy (void *, const void *, size_t)
2776 __attribute__ ((ifunc ("resolve_memcpy")));
2779 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2780 and GNU C Library version 2.11.1 are required to use this feature.
2783 @itemx interrupt_handler
2784 Many GCC back ends support attributes to indicate that a function is
2785 an interrupt handler, which tells the compiler to generate function
2786 entry and exit sequences that differ from those from regular
2787 functions. The exact syntax and behavior are target-specific;
2788 refer to the following subsections for details.
2791 @cindex @code{leaf} function attribute
2792 Calls to external functions with this attribute must return to the current
2793 compilation unit only by return or by exception handling. In particular, leaf
2794 functions are not allowed to call callback function passed to it from the current
2795 compilation unit or directly call functions exported by the unit or longjmp
2796 into the unit. Leaf function might still call functions from other compilation
2797 units and thus they are not necessarily leaf in the sense that they contain no
2798 function calls at all.
2800 The attribute is intended for library functions to improve dataflow analysis.
2801 The compiler takes the hint that any data not escaping the current compilation unit can
2802 not be used or modified by the leaf function. For example, the @code{sin} function
2803 is a leaf function, but @code{qsort} is not.
2805 Note that leaf functions might invoke signals and signal handlers might be
2806 defined in the current compilation unit and use static variables. The only
2807 compliant way to write such a signal handler is to declare such variables
2810 The attribute has no effect on functions defined within the current compilation
2811 unit. This is to allow easy merging of multiple compilation units into one,
2812 for example, by using the link-time optimization. For this reason the
2813 attribute is not allowed on types to annotate indirect calls.
2817 @cindex @code{malloc} function attribute
2818 @cindex functions that behave like malloc
2819 This tells the compiler that a function is @code{malloc}-like, i.e.,
2820 that the pointer @var{P} returned by the function cannot alias any
2821 other pointer valid when the function returns, and moreover no
2822 pointers to valid objects occur in any storage addressed by @var{P}.
2824 Using this attribute can improve optimization. Functions like
2825 @code{malloc} and @code{calloc} have this property because they return
2826 a pointer to uninitialized or zeroed-out storage. However, functions
2827 like @code{realloc} do not have this property, as they can return a
2828 pointer to storage containing pointers.
2831 @cindex @code{no_icf} function attribute
2832 This function attribute prevents a functions from being merged with another
2833 semantically equivalent function.
2835 @item no_instrument_function
2836 @cindex @code{no_instrument_function} function attribute
2837 @opindex finstrument-functions
2838 If @option{-finstrument-functions} is given, profiling function calls are
2839 generated at entry and exit of most user-compiled functions.
2840 Functions with this attribute are not so instrumented.
2843 @cindex @code{no_reorder} function attribute
2844 Do not reorder functions or variables marked @code{no_reorder}
2845 against each other or top level assembler statements the executable.
2846 The actual order in the program will depend on the linker command
2847 line. Static variables marked like this are also not removed.
2848 This has a similar effect
2849 as the @option{-fno-toplevel-reorder} option, but only applies to the
2852 @item no_sanitize_address
2853 @itemx no_address_safety_analysis
2854 @cindex @code{no_sanitize_address} function attribute
2855 The @code{no_sanitize_address} attribute on functions is used
2856 to inform the compiler that it should not instrument memory accesses
2857 in the function when compiling with the @option{-fsanitize=address} option.
2858 The @code{no_address_safety_analysis} is a deprecated alias of the
2859 @code{no_sanitize_address} attribute, new code should use
2860 @code{no_sanitize_address}.
2862 @item no_sanitize_thread
2863 @cindex @code{no_sanitize_thread} function attribute
2864 The @code{no_sanitize_thread} attribute on functions is used
2865 to inform the compiler that it should not instrument memory accesses
2866 in the function when compiling with the @option{-fsanitize=thread} option.
2868 @item no_sanitize_undefined
2869 @cindex @code{no_sanitize_undefined} function attribute
2870 The @code{no_sanitize_undefined} attribute on functions is used
2871 to inform the compiler that it should not check for undefined behavior
2872 in the function when compiling with the @option{-fsanitize=undefined} option.
2874 @item no_split_stack
2875 @cindex @code{no_split_stack} function attribute
2876 @opindex fsplit-stack
2877 If @option{-fsplit-stack} is given, functions have a small
2878 prologue which decides whether to split the stack. Functions with the
2879 @code{no_split_stack} attribute do not have that prologue, and thus
2880 may run with only a small amount of stack space available.
2882 @item no_stack_limit
2883 @cindex @code{no_stack_limit} function attribute
2884 This attribute locally overrides the @option{-fstack-limit-register}
2885 and @option{-fstack-limit-symbol} command-line options; it has the effect
2886 of disabling stack limit checking in the function it applies to.
2889 @cindex @code{noclone} function attribute
2890 This function attribute prevents a function from being considered for
2891 cloning---a mechanism that produces specialized copies of functions
2892 and which is (currently) performed by interprocedural constant
2896 @cindex @code{noinline} function attribute
2897 This function attribute prevents a function from being considered for
2899 @c Don't enumerate the optimizations by name here; we try to be
2900 @c future-compatible with this mechanism.
2901 If the function does not have side-effects, there are optimizations
2902 other than inlining that cause function calls to be optimized away,
2903 although the function call is live. To keep such calls from being
2910 (@pxref{Extended Asm}) in the called function, to serve as a special
2913 @item nonnull (@var{arg-index}, @dots{})
2914 @cindex @code{nonnull} function attribute
2915 @cindex functions with non-null pointer arguments
2916 The @code{nonnull} attribute specifies that some function parameters should
2917 be non-null pointers. For instance, the declaration:
2921 my_memcpy (void *dest, const void *src, size_t len)
2922 __attribute__((nonnull (1, 2)));
2926 causes the compiler to check that, in calls to @code{my_memcpy},
2927 arguments @var{dest} and @var{src} are non-null. If the compiler
2928 determines that a null pointer is passed in an argument slot marked
2929 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2930 is issued. The compiler may also choose to make optimizations based
2931 on the knowledge that certain function arguments will never be null.
2933 If no argument index list is given to the @code{nonnull} attribute,
2934 all pointer arguments are marked as non-null. To illustrate, the
2935 following declaration is equivalent to the previous example:
2939 my_memcpy (void *dest, const void *src, size_t len)
2940 __attribute__((nonnull));
2944 @cindex @code{noplt} function attribute
2945 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
2946 Calls to functions marked with this attribute in position-independent code
2951 /* Externally defined function foo. */
2952 int foo () __attribute__ ((noplt));
2955 main (/* @r{@dots{}} */)
2964 The @code{noplt} attribute on function @code{foo}
2965 tells the compiler to assume that
2966 the function @code{foo} is externally defined and that the call to
2967 @code{foo} must avoid the PLT
2968 in position-independent code.
2970 In position-dependent code, a few targets also convert calls to
2971 functions that are marked to not use the PLT to use the GOT instead.
2974 @cindex @code{noreturn} function attribute
2975 @cindex functions that never return
2976 A few standard library functions, such as @code{abort} and @code{exit},
2977 cannot return. GCC knows this automatically. Some programs define
2978 their own functions that never return. You can declare them
2979 @code{noreturn} to tell the compiler this fact. For example,
2983 void fatal () __attribute__ ((noreturn));
2986 fatal (/* @r{@dots{}} */)
2988 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2994 The @code{noreturn} keyword tells the compiler to assume that
2995 @code{fatal} cannot return. It can then optimize without regard to what
2996 would happen if @code{fatal} ever did return. This makes slightly
2997 better code. More importantly, it helps avoid spurious warnings of
2998 uninitialized variables.
3000 The @code{noreturn} keyword does not affect the exceptional path when that
3001 applies: a @code{noreturn}-marked function may still return to the caller
3002 by throwing an exception or calling @code{longjmp}.
3004 Do not assume that registers saved by the calling function are
3005 restored before calling the @code{noreturn} function.
3007 It does not make sense for a @code{noreturn} function to have a return
3008 type other than @code{void}.
3011 @cindex @code{nothrow} function attribute
3012 The @code{nothrow} attribute is used to inform the compiler that a
3013 function cannot throw an exception. For example, most functions in
3014 the standard C library can be guaranteed not to throw an exception
3015 with the notable exceptions of @code{qsort} and @code{bsearch} that
3016 take function pointer arguments.
3019 @cindex @code{optimize} function attribute
3020 The @code{optimize} attribute is used to specify that a function is to
3021 be compiled with different optimization options than specified on the
3022 command line. Arguments can either be numbers or strings. Numbers
3023 are assumed to be an optimization level. Strings that begin with
3024 @code{O} are assumed to be an optimization option, while other options
3025 are assumed to be used with a @code{-f} prefix. You can also use the
3026 @samp{#pragma GCC optimize} pragma to set the optimization options
3027 that affect more than one function.
3028 @xref{Function Specific Option Pragmas}, for details about the
3029 @samp{#pragma GCC optimize} pragma.
3031 This can be used for instance to have frequently-executed functions
3032 compiled with more aggressive optimization options that produce faster
3033 and larger code, while other functions can be compiled with less
3037 @cindex @code{pure} function attribute
3038 @cindex functions that have no side effects
3039 Many functions have no effects except the return value and their
3040 return value depends only on the parameters and/or global variables.
3041 Such a function can be subject
3042 to common subexpression elimination and loop optimization just as an
3043 arithmetic operator would be. These functions should be declared
3044 with the attribute @code{pure}. For example,
3047 int square (int) __attribute__ ((pure));
3051 says that the hypothetical function @code{square} is safe to call
3052 fewer times than the program says.
3054 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3055 Interesting non-pure functions are functions with infinite loops or those
3056 depending on volatile memory or other system resource, that may change between
3057 two consecutive calls (such as @code{feof} in a multithreading environment).
3059 @item returns_nonnull
3060 @cindex @code{returns_nonnull} function attribute
3061 The @code{returns_nonnull} attribute specifies that the function
3062 return value should be a non-null pointer. For instance, the declaration:
3066 mymalloc (size_t len) __attribute__((returns_nonnull));
3070 lets the compiler optimize callers based on the knowledge
3071 that the return value will never be null.
3074 @cindex @code{returns_twice} function attribute
3075 @cindex functions that return more than once
3076 The @code{returns_twice} attribute tells the compiler that a function may
3077 return more than one time. The compiler ensures that all registers
3078 are dead before calling such a function and emits a warning about
3079 the variables that may be clobbered after the second return from the
3080 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3081 The @code{longjmp}-like counterpart of such function, if any, might need
3082 to be marked with the @code{noreturn} attribute.
3084 @item section ("@var{section-name}")
3085 @cindex @code{section} function attribute
3086 @cindex functions in arbitrary sections
3087 Normally, the compiler places the code it generates in the @code{text} section.
3088 Sometimes, however, you need additional sections, or you need certain
3089 particular functions to appear in special sections. The @code{section}
3090 attribute specifies that a function lives in a particular section.
3091 For example, the declaration:
3094 extern void foobar (void) __attribute__ ((section ("bar")));
3098 puts the function @code{foobar} in the @code{bar} section.
3100 Some file formats do not support arbitrary sections so the @code{section}
3101 attribute is not available on all platforms.
3102 If you need to map the entire contents of a module to a particular
3103 section, consider using the facilities of the linker instead.
3106 @cindex @code{sentinel} function attribute
3107 This function attribute ensures that a parameter in a function call is
3108 an explicit @code{NULL}. The attribute is only valid on variadic
3109 functions. By default, the sentinel is located at position zero, the
3110 last parameter of the function call. If an optional integer position
3111 argument P is supplied to the attribute, the sentinel must be located at
3112 position P counting backwards from the end of the argument list.
3115 __attribute__ ((sentinel))
3117 __attribute__ ((sentinel(0)))
3120 The attribute is automatically set with a position of 0 for the built-in
3121 functions @code{execl} and @code{execlp}. The built-in function
3122 @code{execle} has the attribute set with a position of 1.
3124 A valid @code{NULL} in this context is defined as zero with any pointer
3125 type. If your system defines the @code{NULL} macro with an integer type
3126 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3127 with a copy that redefines NULL appropriately.
3129 The warnings for missing or incorrect sentinels are enabled with
3133 @itemx simd("@var{mask}")
3134 @cindex @code{simd} function attribute
3135 This attribute enables creation of one or more function versions that
3136 can process multiple arguments using SIMD instructions from a
3137 single invocation. Specifying this attribute allows compiler to
3138 assume that such versions are available at link time (provided
3139 in the same or another translation unit). Generated versions are
3140 target-dependent and described in the corresponding Vector ABI document. For
3141 x86_64 target this document can be found
3142 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3144 The optional argument @var{mask} may have the value
3145 @code{notinbranch} or @code{inbranch},
3146 and instructs the compiler to generate non-masked or masked
3147 clones correspondingly. By default, all clones are generated.
3149 The attribute should not be used together with Cilk Plus @code{vector}
3150 attribute on the same function.
3152 If the attribute is specified and @code{#pragma omp declare simd} is
3153 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3154 switch is specified, then the attribute is ignored.
3157 @cindex @code{stack_protect} function attribute
3158 This attribute adds stack protection code to the function if
3159 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3160 or @option{-fstack-protector-explicit} are set.
3162 @item target (@var{options})
3163 @cindex @code{target} function attribute
3164 Multiple target back ends implement the @code{target} attribute
3165 to specify that a function is to
3166 be compiled with different target options than specified on the
3167 command line. This can be used for instance to have functions
3168 compiled with a different ISA (instruction set architecture) than the
3169 default. You can also use the @samp{#pragma GCC target} pragma to set
3170 more than one function to be compiled with specific target options.
3171 @xref{Function Specific Option Pragmas}, for details about the
3172 @samp{#pragma GCC target} pragma.
3174 For instance, on an x86, you could declare one function with the
3175 @code{target("sse4.1,arch=core2")} attribute and another with
3176 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3177 compiling the first function with @option{-msse4.1} and
3178 @option{-march=core2} options, and the second function with
3179 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3180 to make sure that a function is only invoked on a machine that
3181 supports the particular ISA it is compiled for (for example by using
3182 @code{cpuid} on x86 to determine what feature bits and architecture
3186 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3187 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3190 You can either use multiple
3191 strings separated by commas to specify multiple options,
3192 or separate the options with a comma (@samp{,}) within a single string.
3194 The options supported are specific to each target; refer to @ref{x86
3195 Function Attributes}, @ref{PowerPC Function Attributes},
3196 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3199 @item target_clones (@var{options})
3200 @cindex @code{target_clones} function attribute
3201 The @code{target_clones} attribute is used to specify that a function
3202 be cloned into multiple versions compiled with different target options
3203 than specified on the command line. The supported options and restrictions
3204 are the same as for @code{target} attribute.
3206 For instance, on an x86, you could compile a function with
3207 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3208 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3209 It also creates a resolver function (see the @code{ifunc} attribute
3210 above) that dynamically selects a clone suitable for current architecture.
3213 @cindex @code{unused} function attribute
3214 This attribute, attached to a function, means that the function is meant
3215 to be possibly unused. GCC does not produce a warning for this
3219 @cindex @code{used} function attribute
3220 This attribute, attached to a function, means that code must be emitted
3221 for the function even if it appears that the function is not referenced.
3222 This is useful, for example, when the function is referenced only in
3225 When applied to a member function of a C++ class template, the
3226 attribute also means that the function is instantiated if the
3227 class itself is instantiated.
3229 @item visibility ("@var{visibility_type}")
3230 @cindex @code{visibility} function attribute
3231 This attribute affects the linkage of the declaration to which it is attached.
3232 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3233 (@pxref{Common Type Attributes}) as well as functions.
3235 There are four supported @var{visibility_type} values: default,
3236 hidden, protected or internal visibility.
3239 void __attribute__ ((visibility ("protected")))
3240 f () @{ /* @r{Do something.} */; @}
3241 int i __attribute__ ((visibility ("hidden")));
3244 The possible values of @var{visibility_type} correspond to the
3245 visibility settings in the ELF gABI.
3248 @c keep this list of visibilities in alphabetical order.
3251 Default visibility is the normal case for the object file format.
3252 This value is available for the visibility attribute to override other
3253 options that may change the assumed visibility of entities.
3255 On ELF, default visibility means that the declaration is visible to other
3256 modules and, in shared libraries, means that the declared entity may be
3259 On Darwin, default visibility means that the declaration is visible to
3262 Default visibility corresponds to ``external linkage'' in the language.
3265 Hidden visibility indicates that the entity declared has a new
3266 form of linkage, which we call ``hidden linkage''. Two
3267 declarations of an object with hidden linkage refer to the same object
3268 if they are in the same shared object.
3271 Internal visibility is like hidden visibility, but with additional
3272 processor specific semantics. Unless otherwise specified by the
3273 psABI, GCC defines internal visibility to mean that a function is
3274 @emph{never} called from another module. Compare this with hidden
3275 functions which, while they cannot be referenced directly by other
3276 modules, can be referenced indirectly via function pointers. By
3277 indicating that a function cannot be called from outside the module,
3278 GCC may for instance omit the load of a PIC register since it is known
3279 that the calling function loaded the correct value.
3282 Protected visibility is like default visibility except that it
3283 indicates that references within the defining module bind to the
3284 definition in that module. That is, the declared entity cannot be
3285 overridden by another module.
3289 All visibilities are supported on many, but not all, ELF targets
3290 (supported when the assembler supports the @samp{.visibility}
3291 pseudo-op). Default visibility is supported everywhere. Hidden
3292 visibility is supported on Darwin targets.
3294 The visibility attribute should be applied only to declarations that
3295 would otherwise have external linkage. The attribute should be applied
3296 consistently, so that the same entity should not be declared with
3297 different settings of the attribute.
3299 In C++, the visibility attribute applies to types as well as functions
3300 and objects, because in C++ types have linkage. A class must not have
3301 greater visibility than its non-static data member types and bases,
3302 and class members default to the visibility of their class. Also, a
3303 declaration without explicit visibility is limited to the visibility
3306 In C++, you can mark member functions and static member variables of a
3307 class with the visibility attribute. This is useful if you know a
3308 particular method or static member variable should only be used from
3309 one shared object; then you can mark it hidden while the rest of the
3310 class has default visibility. Care must be taken to avoid breaking
3311 the One Definition Rule; for example, it is usually not useful to mark
3312 an inline method as hidden without marking the whole class as hidden.
3314 A C++ namespace declaration can also have the visibility attribute.
3317 namespace nspace1 __attribute__ ((visibility ("protected")))
3318 @{ /* @r{Do something.} */; @}
3321 This attribute applies only to the particular namespace body, not to
3322 other definitions of the same namespace; it is equivalent to using
3323 @samp{#pragma GCC visibility} before and after the namespace
3324 definition (@pxref{Visibility Pragmas}).
3326 In C++, if a template argument has limited visibility, this
3327 restriction is implicitly propagated to the template instantiation.
3328 Otherwise, template instantiations and specializations default to the
3329 visibility of their template.
3331 If both the template and enclosing class have explicit visibility, the
3332 visibility from the template is used.
3334 @item warn_unused_result
3335 @cindex @code{warn_unused_result} function attribute
3336 The @code{warn_unused_result} attribute causes a warning to be emitted
3337 if a caller of the function with this attribute does not use its
3338 return value. This is useful for functions where not checking
3339 the result is either a security problem or always a bug, such as
3343 int fn () __attribute__ ((warn_unused_result));
3346 if (fn () < 0) return -1;
3353 results in warning on line 5.
3356 @cindex @code{weak} function attribute
3357 The @code{weak} attribute causes the declaration to be emitted as a weak
3358 symbol rather than a global. This is primarily useful in defining
3359 library functions that can be overridden in user code, though it can
3360 also be used with non-function declarations. Weak symbols are supported
3361 for ELF targets, and also for a.out targets when using the GNU assembler
3365 @itemx weakref ("@var{target}")
3366 @cindex @code{weakref} function attribute
3367 The @code{weakref} attribute marks a declaration as a weak reference.
3368 Without arguments, it should be accompanied by an @code{alias} attribute
3369 naming the target symbol. Optionally, the @var{target} may be given as
3370 an argument to @code{weakref} itself. In either case, @code{weakref}
3371 implicitly marks the declaration as @code{weak}. Without a
3372 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3373 @code{weakref} is equivalent to @code{weak}.
3376 static int x() __attribute__ ((weakref ("y")));
3377 /* is equivalent to... */
3378 static int x() __attribute__ ((weak, weakref, alias ("y")));
3380 static int x() __attribute__ ((weakref));
3381 static int x() __attribute__ ((alias ("y")));
3384 A weak reference is an alias that does not by itself require a
3385 definition to be given for the target symbol. If the target symbol is
3386 only referenced through weak references, then it becomes a @code{weak}
3387 undefined symbol. If it is directly referenced, however, then such
3388 strong references prevail, and a definition is required for the
3389 symbol, not necessarily in the same translation unit.
3391 The effect is equivalent to moving all references to the alias to a
3392 separate translation unit, renaming the alias to the aliased symbol,
3393 declaring it as weak, compiling the two separate translation units and
3394 performing a reloadable link on them.
3396 At present, a declaration to which @code{weakref} is attached can
3397 only be @code{static}.
3402 @c This is the end of the target-independent attribute table
3404 @node AArch64 Function Attributes
3405 @subsection AArch64 Function Attributes
3407 The following target-specific function attributes are available for the
3408 AArch64 target. For the most part, these options mirror the behavior of
3409 similar command-line options (@pxref{AArch64 Options}), but on a
3413 @item general-regs-only
3414 @cindex @code{general-regs-only} function attribute, AArch64
3415 Indicates that no floating-point or Advanced SIMD registers should be
3416 used when generating code for this function. If the function explicitly
3417 uses floating-point code, then the compiler gives an error. This is
3418 the same behavior as that of the command-line option
3419 @option{-mgeneral-regs-only}.
3421 @item fix-cortex-a53-835769
3422 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3423 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3424 applied to this function. To explicitly disable the workaround for this
3425 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3426 This corresponds to the behavior of the command line options
3427 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3430 @cindex @code{cmodel=} function attribute, AArch64
3431 Indicates that code should be generated for a particular code model for
3432 this function. The behavior and permissible arguments are the same as
3433 for the command line option @option{-mcmodel=}.
3436 @cindex @code{strict-align} function attribute, AArch64
3437 Indicates that the compiler should not assume that unaligned memory references
3438 are handled by the system. The behavior is the same as for the command-line
3439 option @option{-mstrict-align}.
3441 @item omit-leaf-frame-pointer
3442 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3443 Indicates that the frame pointer should be omitted for a leaf function call.
3444 To keep the frame pointer, the inverse attribute
3445 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3446 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3447 and @option{-mno-omit-leaf-frame-pointer}.
3450 @cindex @code{tls-dialect=} function attribute, AArch64
3451 Specifies the TLS dialect to use for this function. The behavior and
3452 permissible arguments are the same as for the command-line option
3453 @option{-mtls-dialect=}.
3456 @cindex @code{arch=} function attribute, AArch64
3457 Specifies the architecture version and architectural extensions to use
3458 for this function. The behavior and permissible arguments are the same as
3459 for the @option{-march=} command-line option.
3462 @cindex @code{tune=} function attribute, AArch64
3463 Specifies the core for which to tune the performance of this function.
3464 The behavior and permissible arguments are the same as for the @option{-mtune=}
3465 command-line option.
3468 @cindex @code{cpu=} function attribute, AArch64
3469 Specifies the core for which to tune the performance of this function and also
3470 whose architectural features to use. The behavior and valid arguments are the
3471 same as for the @option{-mcpu=} command-line option.
3475 The above target attributes can be specified as follows:
3478 __attribute__((target("@var{attr-string}")))
3486 where @code{@var{attr-string}} is one of the attribute strings specified above.
3488 Additionally, the architectural extension string may be specified on its
3489 own. This can be used to turn on and off particular architectural extensions
3490 without having to specify a particular architecture version or core. Example:
3493 __attribute__((target("+crc+nocrypto")))
3501 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3502 extension and disables the @code{crypto} extension for the function @code{foo}
3503 without modifying an existing @option{-march=} or @option{-mcpu} option.
3505 Multiple target function attributes can be specified by separating them with
3506 a comma. For example:
3508 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3516 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3517 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3519 @subsubsection Inlining rules
3520 Specifying target attributes on individual functions or performing link-time
3521 optimization across translation units compiled with different target options
3522 can affect function inlining rules:
3524 In particular, a caller function can inline a callee function only if the
3525 architectural features available to the callee are a subset of the features
3526 available to the caller.
3527 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3528 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3529 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3530 because the all the architectural features that function @code{bar} requires
3531 are available to function @code{foo}. Conversely, function @code{bar} cannot
3532 inline function @code{foo}.
3534 Additionally inlining a function compiled with @option{-mstrict-align} into a
3535 function compiled without @code{-mstrict-align} is not allowed.
3536 However, inlining a function compiled without @option{-mstrict-align} into a
3537 function compiled with @option{-mstrict-align} is allowed.
3539 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3540 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3541 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3542 architectural feature rules specified above.
3544 @node ARC Function Attributes
3545 @subsection ARC Function Attributes
3547 These function attributes are supported by the ARC back end:
3551 @cindex @code{interrupt} function attribute, ARC
3552 Use this attribute to indicate
3553 that the specified function is an interrupt handler. The compiler generates
3554 function entry and exit sequences suitable for use in an interrupt handler
3555 when this attribute is present.
3557 On the ARC, you must specify the kind of interrupt to be handled
3558 in a parameter to the interrupt attribute like this:
3561 void f () __attribute__ ((interrupt ("ilink1")));
3564 Permissible values for this parameter are: @w{@code{ilink1}} and
3570 @cindex @code{long_call} function attribute, ARC
3571 @cindex @code{medium_call} function attribute, ARC
3572 @cindex @code{short_call} function attribute, ARC
3573 @cindex indirect calls, ARC
3574 These attributes specify how a particular function is called.
3575 These attributes override the
3576 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3577 command-line switches and @code{#pragma long_calls} settings.
3579 For ARC, a function marked with the @code{long_call} attribute is
3580 always called using register-indirect jump-and-link instructions,
3581 thereby enabling the called function to be placed anywhere within the
3582 32-bit address space. A function marked with the @code{medium_call}
3583 attribute will always be close enough to be called with an unconditional
3584 branch-and-link instruction, which has a 25-bit offset from
3585 the call site. A function marked with the @code{short_call}
3586 attribute will always be close enough to be called with a conditional
3587 branch-and-link instruction, which has a 21-bit offset from
3591 @node ARM Function Attributes
3592 @subsection ARM Function Attributes
3594 These function attributes are supported for ARM targets:
3598 @cindex @code{interrupt} function attribute, ARM
3599 Use this attribute to indicate
3600 that the specified function is an interrupt handler. The compiler generates
3601 function entry and exit sequences suitable for use in an interrupt handler
3602 when this attribute is present.
3604 You can specify the kind of interrupt to be handled by
3605 adding an optional parameter to the interrupt attribute like this:
3608 void f () __attribute__ ((interrupt ("IRQ")));
3612 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3613 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3615 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3616 may be called with a word-aligned stack pointer.
3619 @cindex @code{isr} function attribute, ARM
3620 Use this attribute on ARM to write Interrupt Service Routines. This is an
3621 alias to the @code{interrupt} attribute above.
3625 @cindex @code{long_call} function attribute, ARM
3626 @cindex @code{short_call} function attribute, ARM
3627 @cindex indirect calls, ARM
3628 These attributes specify how a particular function is called.
3629 These attributes override the
3630 @option{-mlong-calls} (@pxref{ARM Options})
3631 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3632 @code{long_call} attribute indicates that the function might be far
3633 away from the call site and require a different (more expensive)
3634 calling sequence. The @code{short_call} attribute always places
3635 the offset to the function from the call site into the @samp{BL}
3636 instruction directly.
3639 @cindex @code{naked} function attribute, ARM
3640 This attribute allows the compiler to construct the
3641 requisite function declaration, while allowing the body of the
3642 function to be assembly code. The specified function will not have
3643 prologue/epilogue sequences generated by the compiler. Only basic
3644 @code{asm} statements can safely be included in naked functions
3645 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3646 basic @code{asm} and C code may appear to work, they cannot be
3647 depended upon to work reliably and are not supported.
3650 @cindex @code{pcs} function attribute, ARM
3652 The @code{pcs} attribute can be used to control the calling convention
3653 used for a function on ARM. The attribute takes an argument that specifies
3654 the calling convention to use.
3656 When compiling using the AAPCS ABI (or a variant of it) then valid
3657 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3658 order to use a variant other than @code{"aapcs"} then the compiler must
3659 be permitted to use the appropriate co-processor registers (i.e., the
3660 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3664 /* Argument passed in r0, and result returned in r0+r1. */
3665 double f2d (float) __attribute__((pcs("aapcs")));
3668 Variadic functions always use the @code{"aapcs"} calling convention and
3669 the compiler rejects attempts to specify an alternative.
3671 @item target (@var{options})
3672 @cindex @code{target} function attribute
3673 As discussed in @ref{Common Function Attributes}, this attribute
3674 allows specification of target-specific compilation options.
3676 On ARM, the following options are allowed:
3680 @cindex @code{target("thumb")} function attribute, ARM
3681 Force code generation in the Thumb (T16/T32) ISA, depending on the
3685 @cindex @code{target("arm")} function attribute, ARM
3686 Force code generation in the ARM (A32) ISA.
3688 Functions from different modes can be inlined in the caller's mode.
3691 @cindex @code{target("fpu=")} function attribute, ARM
3692 Specifies the fpu for which to tune the performance of this function.
3693 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3694 command-line option.
3700 @node AVR Function Attributes
3701 @subsection AVR Function Attributes
3703 These function attributes are supported by the AVR back end:
3707 @cindex @code{interrupt} function attribute, AVR
3708 Use this attribute to indicate
3709 that the specified function is an interrupt handler. The compiler generates
3710 function entry and exit sequences suitable for use in an interrupt handler
3711 when this attribute is present.
3713 On the AVR, the hardware globally disables interrupts when an
3714 interrupt is executed. The first instruction of an interrupt handler
3715 declared with this attribute is a @code{SEI} instruction to
3716 re-enable interrupts. See also the @code{signal} function attribute
3717 that does not insert a @code{SEI} instruction. If both @code{signal} and
3718 @code{interrupt} are specified for the same function, @code{signal}
3719 is silently ignored.
3722 @cindex @code{naked} function attribute, AVR
3723 This attribute allows the compiler to construct the
3724 requisite function declaration, while allowing the body of the
3725 function to be assembly code. The specified function will not have
3726 prologue/epilogue sequences generated by the compiler. Only basic
3727 @code{asm} statements can safely be included in naked functions
3728 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3729 basic @code{asm} and C code may appear to work, they cannot be
3730 depended upon to work reliably and are not supported.
3734 @cindex @code{OS_main} function attribute, AVR
3735 @cindex @code{OS_task} function attribute, AVR
3736 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3737 do not save/restore any call-saved register in their prologue/epilogue.
3739 The @code{OS_main} attribute can be used when there @emph{is
3740 guarantee} that interrupts are disabled at the time when the function
3741 is entered. This saves resources when the stack pointer has to be
3742 changed to set up a frame for local variables.
3744 The @code{OS_task} attribute can be used when there is @emph{no
3745 guarantee} that interrupts are disabled at that time when the function
3746 is entered like for, e@.g@. task functions in a multi-threading operating
3747 system. In that case, changing the stack pointer register is
3748 guarded by save/clear/restore of the global interrupt enable flag.
3750 The differences to the @code{naked} function attribute are:
3752 @item @code{naked} functions do not have a return instruction whereas
3753 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3754 @code{RETI} return instruction.
3755 @item @code{naked} functions do not set up a frame for local variables
3756 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3761 @cindex @code{signal} function attribute, AVR
3762 Use this attribute on the AVR to indicate that the specified
3763 function is an interrupt handler. The compiler generates function
3764 entry and exit sequences suitable for use in an interrupt handler when this
3765 attribute is present.
3767 See also the @code{interrupt} function attribute.
3769 The AVR hardware globally disables interrupts when an interrupt is executed.
3770 Interrupt handler functions defined with the @code{signal} attribute
3771 do not re-enable interrupts. It is save to enable interrupts in a
3772 @code{signal} handler. This ``save'' only applies to the code
3773 generated by the compiler and not to the IRQ layout of the
3774 application which is responsibility of the application.
3776 If both @code{signal} and @code{interrupt} are specified for the same
3777 function, @code{signal} is silently ignored.
3780 @node Blackfin Function Attributes
3781 @subsection Blackfin Function Attributes
3783 These function attributes are supported by the Blackfin back end:
3787 @item exception_handler
3788 @cindex @code{exception_handler} function attribute
3789 @cindex exception handler functions, Blackfin
3790 Use this attribute on the Blackfin to indicate that the specified function
3791 is an exception handler. The compiler generates function entry and
3792 exit sequences suitable for use in an exception handler when this
3793 attribute is present.
3795 @item interrupt_handler
3796 @cindex @code{interrupt_handler} function attribute, Blackfin
3797 Use this attribute to
3798 indicate that the specified function is an interrupt handler. The compiler
3799 generates function entry and exit sequences suitable for use in an
3800 interrupt handler when this attribute is present.
3803 @cindex @code{kspisusp} function attribute, Blackfin
3804 @cindex User stack pointer in interrupts on the Blackfin
3805 When used together with @code{interrupt_handler}, @code{exception_handler}
3806 or @code{nmi_handler}, code is generated to load the stack pointer
3807 from the USP register in the function prologue.
3810 @cindex @code{l1_text} function attribute, Blackfin
3811 This attribute specifies a function to be placed into L1 Instruction
3812 SRAM@. The function is put into a specific section named @code{.l1.text}.
3813 With @option{-mfdpic}, function calls with a such function as the callee
3814 or caller uses inlined PLT.
3817 @cindex @code{l2} function attribute, Blackfin
3818 This attribute specifies a function to be placed into L2
3819 SRAM. The function is put into a specific section named
3820 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3825 @cindex indirect calls, Blackfin
3826 @cindex @code{longcall} function attribute, Blackfin
3827 @cindex @code{shortcall} function attribute, Blackfin
3828 The @code{longcall} attribute
3829 indicates that the function might be far away from the call site and
3830 require a different (more expensive) calling sequence. The
3831 @code{shortcall} attribute indicates that the function is always close
3832 enough for the shorter calling sequence to be used. These attributes
3833 override the @option{-mlongcall} switch.
3836 @cindex @code{nesting} function attribute, Blackfin
3837 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3838 Use this attribute together with @code{interrupt_handler},
3839 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3840 entry code should enable nested interrupts or exceptions.
3843 @cindex @code{nmi_handler} function attribute, Blackfin
3844 @cindex NMI handler functions on the Blackfin processor
3845 Use this attribute on the Blackfin to indicate that the specified function
3846 is an NMI handler. The compiler generates function entry and
3847 exit sequences suitable for use in an NMI handler when this
3848 attribute is present.
3851 @cindex @code{saveall} function attribute, Blackfin
3852 @cindex save all registers on the Blackfin
3853 Use this attribute to indicate that
3854 all registers except the stack pointer should be saved in the prologue
3855 regardless of whether they are used or not.
3858 @node CR16 Function Attributes
3859 @subsection CR16 Function Attributes
3861 These function attributes are supported by the CR16 back end:
3865 @cindex @code{interrupt} function attribute, CR16
3866 Use this attribute to indicate
3867 that the specified function is an interrupt handler. The compiler generates
3868 function entry and exit sequences suitable for use in an interrupt handler
3869 when this attribute is present.
3872 @node Epiphany Function Attributes
3873 @subsection Epiphany Function Attributes
3875 These function attributes are supported by the Epiphany back end:
3879 @cindex @code{disinterrupt} function attribute, Epiphany
3880 This attribute causes the compiler to emit
3881 instructions to disable interrupts for the duration of the given
3884 @item forwarder_section
3885 @cindex @code{forwarder_section} function attribute, Epiphany
3886 This attribute modifies the behavior of an interrupt handler.
3887 The interrupt handler may be in external memory which cannot be
3888 reached by a branch instruction, so generate a local memory trampoline
3889 to transfer control. The single parameter identifies the section where
3890 the trampoline is placed.
3893 @cindex @code{interrupt} function attribute, Epiphany
3894 Use this attribute to indicate
3895 that the specified function is an interrupt handler. The compiler generates
3896 function entry and exit sequences suitable for use in an interrupt handler
3897 when this attribute is present. It may also generate
3898 a special section with code to initialize the interrupt vector table.
3900 On Epiphany targets one or more optional parameters can be added like this:
3903 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3906 Permissible values for these parameters are: @w{@code{reset}},
3907 @w{@code{software_exception}}, @w{@code{page_miss}},
3908 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3909 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3910 Multiple parameters indicate that multiple entries in the interrupt
3911 vector table should be initialized for this function, i.e.@: for each
3912 parameter @w{@var{name}}, a jump to the function is emitted in
3913 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3914 entirely, in which case no interrupt vector table entry is provided.
3916 Note that interrupts are enabled inside the function
3917 unless the @code{disinterrupt} attribute is also specified.
3919 The following examples are all valid uses of these attributes on
3922 void __attribute__ ((interrupt)) universal_handler ();
3923 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3924 void __attribute__ ((interrupt ("dma0, dma1")))
3925 universal_dma_handler ();
3926 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3927 fast_timer_handler ();
3928 void __attribute__ ((interrupt ("dma0, dma1"),
3929 forwarder_section ("tramp")))
3930 external_dma_handler ();
3935 @cindex @code{long_call} function attribute, Epiphany
3936 @cindex @code{short_call} function attribute, Epiphany
3937 @cindex indirect calls, Epiphany
3938 These attributes specify how a particular function is called.
3939 These attributes override the
3940 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3941 command-line switch and @code{#pragma long_calls} settings.
3945 @node H8/300 Function Attributes
3946 @subsection H8/300 Function Attributes
3948 These function attributes are available for H8/300 targets:
3951 @item function_vector
3952 @cindex @code{function_vector} function attribute, H8/300
3953 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3954 that the specified function should be called through the function vector.
3955 Calling a function through the function vector reduces code size; however,
3956 the function vector has a limited size (maximum 128 entries on the H8/300
3957 and 64 entries on the H8/300H and H8S)
3958 and shares space with the interrupt vector.
3960 @item interrupt_handler
3961 @cindex @code{interrupt_handler} function attribute, H8/300
3962 Use this attribute on the H8/300, H8/300H, and H8S to
3963 indicate that the specified function is an interrupt handler. The compiler
3964 generates function entry and exit sequences suitable for use in an
3965 interrupt handler when this attribute is present.
3968 @cindex @code{saveall} function attribute, H8/300
3969 @cindex save all registers on the H8/300, H8/300H, and H8S
3970 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3971 all registers except the stack pointer should be saved in the prologue
3972 regardless of whether they are used or not.
3975 @node IA-64 Function Attributes
3976 @subsection IA-64 Function Attributes
3978 These function attributes are supported on IA-64 targets:
3981 @item syscall_linkage
3982 @cindex @code{syscall_linkage} function attribute, IA-64
3983 This attribute is used to modify the IA-64 calling convention by marking
3984 all input registers as live at all function exits. This makes it possible
3985 to restart a system call after an interrupt without having to save/restore
3986 the input registers. This also prevents kernel data from leaking into
3990 @cindex @code{version_id} function attribute, IA-64
3991 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
3992 symbol to contain a version string, thus allowing for function level
3993 versioning. HP-UX system header files may use function level versioning
3994 for some system calls.
3997 extern int foo () __attribute__((version_id ("20040821")));
4001 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4004 @node M32C Function Attributes
4005 @subsection M32C Function Attributes
4007 These function attributes are supported by the M32C back end:
4011 @cindex @code{bank_switch} function attribute, M32C
4012 When added to an interrupt handler with the M32C port, causes the
4013 prologue and epilogue to use bank switching to preserve the registers
4014 rather than saving them on the stack.
4016 @item fast_interrupt
4017 @cindex @code{fast_interrupt} function attribute, M32C
4018 Use this attribute on the M32C port to indicate that the specified
4019 function is a fast interrupt handler. This is just like the
4020 @code{interrupt} attribute, except that @code{freit} is used to return
4021 instead of @code{reit}.
4023 @item function_vector
4024 @cindex @code{function_vector} function attribute, M16C/M32C
4025 On M16C/M32C targets, the @code{function_vector} attribute declares a
4026 special page subroutine call function. Use of this attribute reduces
4027 the code size by 2 bytes for each call generated to the
4028 subroutine. The argument to the attribute is the vector number entry
4029 from the special page vector table which contains the 16 low-order
4030 bits of the subroutine's entry address. Each vector table has special
4031 page number (18 to 255) that is used in @code{jsrs} instructions.
4032 Jump addresses of the routines are generated by adding 0x0F0000 (in
4033 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4034 2-byte addresses set in the vector table. Therefore you need to ensure
4035 that all the special page vector routines should get mapped within the
4036 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4039 In the following example 2 bytes are saved for each call to
4040 function @code{foo}.
4043 void foo (void) __attribute__((function_vector(0x18)));
4054 If functions are defined in one file and are called in another file,
4055 then be sure to write this declaration in both files.
4057 This attribute is ignored for R8C target.
4060 @cindex @code{interrupt} function attribute, M32C
4061 Use this attribute to indicate
4062 that the specified function is an interrupt handler. The compiler generates
4063 function entry and exit sequences suitable for use in an interrupt handler
4064 when this attribute is present.
4067 @node M32R/D Function Attributes
4068 @subsection M32R/D Function Attributes
4070 These function attributes are supported by the M32R/D back end:
4074 @cindex @code{interrupt} function attribute, M32R/D
4075 Use this attribute to indicate
4076 that the specified function is an interrupt handler. The compiler generates
4077 function entry and exit sequences suitable for use in an interrupt handler
4078 when this attribute is present.
4080 @item model (@var{model-name})
4081 @cindex @code{model} function attribute, M32R/D
4082 @cindex function addressability on the M32R/D
4084 On the M32R/D, use this attribute to set the addressability of an
4085 object, and of the code generated for a function. The identifier
4086 @var{model-name} is one of @code{small}, @code{medium}, or
4087 @code{large}, representing each of the code models.
4089 Small model objects live in the lower 16MB of memory (so that their
4090 addresses can be loaded with the @code{ld24} instruction), and are
4091 callable with the @code{bl} instruction.
4093 Medium model objects may live anywhere in the 32-bit address space (the
4094 compiler generates @code{seth/add3} instructions to load their addresses),
4095 and are callable with the @code{bl} instruction.
4097 Large model objects may live anywhere in the 32-bit address space (the
4098 compiler generates @code{seth/add3} instructions to load their addresses),
4099 and may not be reachable with the @code{bl} instruction (the compiler
4100 generates the much slower @code{seth/add3/jl} instruction sequence).
4103 @node m68k Function Attributes
4104 @subsection m68k Function Attributes
4106 These function attributes are supported by the m68k back end:
4110 @itemx interrupt_handler
4111 @cindex @code{interrupt} function attribute, m68k
4112 @cindex @code{interrupt_handler} function attribute, m68k
4113 Use this attribute to
4114 indicate that the specified function is an interrupt handler. The compiler
4115 generates function entry and exit sequences suitable for use in an
4116 interrupt handler when this attribute is present. Either name may be used.
4118 @item interrupt_thread
4119 @cindex @code{interrupt_thread} function attribute, fido
4120 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4121 that the specified function is an interrupt handler that is designed
4122 to run as a thread. The compiler omits generate prologue/epilogue
4123 sequences and replaces the return instruction with a @code{sleep}
4124 instruction. This attribute is available only on fido.
4127 @node MCORE Function Attributes
4128 @subsection MCORE Function Attributes
4130 These function attributes are supported by the MCORE back end:
4134 @cindex @code{naked} function attribute, MCORE
4135 This attribute allows the compiler to construct the
4136 requisite function declaration, while allowing the body of the
4137 function to be assembly code. The specified function will not have
4138 prologue/epilogue sequences generated by the compiler. Only basic
4139 @code{asm} statements can safely be included in naked functions
4140 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4141 basic @code{asm} and C code may appear to work, they cannot be
4142 depended upon to work reliably and are not supported.
4145 @node MeP Function Attributes
4146 @subsection MeP Function Attributes
4148 These function attributes are supported by the MeP back end:
4152 @cindex @code{disinterrupt} function attribute, MeP
4153 On MeP targets, this attribute causes the compiler to emit
4154 instructions to disable interrupts for the duration of the given
4158 @cindex @code{interrupt} function attribute, MeP
4159 Use this attribute to indicate
4160 that the specified function is an interrupt handler. The compiler generates
4161 function entry and exit sequences suitable for use in an interrupt handler
4162 when this attribute is present.
4165 @cindex @code{near} function attribute, MeP
4166 This attribute causes the compiler to assume the called
4167 function is close enough to use the normal calling convention,
4168 overriding the @option{-mtf} command-line option.
4171 @cindex @code{far} function attribute, MeP
4172 On MeP targets this causes the compiler to use a calling convention
4173 that assumes the called function is too far away for the built-in
4177 @cindex @code{vliw} function attribute, MeP
4178 The @code{vliw} attribute tells the compiler to emit
4179 instructions in VLIW mode instead of core mode. Note that this
4180 attribute is not allowed unless a VLIW coprocessor has been configured
4181 and enabled through command-line options.
4184 @node MicroBlaze Function Attributes
4185 @subsection MicroBlaze Function Attributes
4187 These function attributes are supported on MicroBlaze targets:
4190 @item save_volatiles
4191 @cindex @code{save_volatiles} function attribute, MicroBlaze
4192 Use this attribute to indicate that the function is
4193 an interrupt handler. All volatile registers (in addition to non-volatile
4194 registers) are saved in the function prologue. If the function is a leaf
4195 function, only volatiles used by the function are saved. A normal function
4196 return is generated instead of a return from interrupt.
4199 @cindex @code{break_handler} function attribute, MicroBlaze
4200 @cindex break handler functions
4201 Use this attribute to indicate that
4202 the specified function is a break handler. The compiler generates function
4203 entry and exit sequences suitable for use in an break handler when this
4204 attribute is present. The return from @code{break_handler} is done through
4205 the @code{rtbd} instead of @code{rtsd}.
4208 void f () __attribute__ ((break_handler));
4211 @item interrupt_handler
4212 @itemx fast_interrupt
4213 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4214 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4215 These attributes indicate that the specified function is an interrupt
4216 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4217 used in low-latency interrupt mode, and @code{interrupt_handler} for
4218 interrupts that do not use low-latency handlers. In both cases, GCC
4219 emits appropriate prologue code and generates a return from the handler
4220 using @code{rtid} instead of @code{rtsd}.
4223 @node Microsoft Windows Function Attributes
4224 @subsection Microsoft Windows Function Attributes
4226 The following attributes are available on Microsoft Windows and Symbian OS
4231 @cindex @code{dllexport} function attribute
4232 @cindex @code{__declspec(dllexport)}
4233 On Microsoft Windows targets and Symbian OS targets the
4234 @code{dllexport} attribute causes the compiler to provide a global
4235 pointer to a pointer in a DLL, so that it can be referenced with the
4236 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4237 name is formed by combining @code{_imp__} and the function or variable
4240 You can use @code{__declspec(dllexport)} as a synonym for
4241 @code{__attribute__ ((dllexport))} for compatibility with other
4244 On systems that support the @code{visibility} attribute, this
4245 attribute also implies ``default'' visibility. It is an error to
4246 explicitly specify any other visibility.
4248 GCC's default behavior is to emit all inline functions with the
4249 @code{dllexport} attribute. Since this can cause object file-size bloat,
4250 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4251 ignore the attribute for inlined functions unless the
4252 @option{-fkeep-inline-functions} flag is used instead.
4254 The attribute is ignored for undefined symbols.
4256 When applied to C++ classes, the attribute marks defined non-inlined
4257 member functions and static data members as exports. Static consts
4258 initialized in-class are not marked unless they are also defined
4261 For Microsoft Windows targets there are alternative methods for
4262 including the symbol in the DLL's export table such as using a
4263 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4264 the @option{--export-all} linker flag.
4267 @cindex @code{dllimport} function attribute
4268 @cindex @code{__declspec(dllimport)}
4269 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4270 attribute causes the compiler to reference a function or variable via
4271 a global pointer to a pointer that is set up by the DLL exporting the
4272 symbol. The attribute implies @code{extern}. On Microsoft Windows
4273 targets, the pointer name is formed by combining @code{_imp__} and the
4274 function or variable name.
4276 You can use @code{__declspec(dllimport)} as a synonym for
4277 @code{__attribute__ ((dllimport))} for compatibility with other
4280 On systems that support the @code{visibility} attribute, this
4281 attribute also implies ``default'' visibility. It is an error to
4282 explicitly specify any other visibility.
4284 Currently, the attribute is ignored for inlined functions. If the
4285 attribute is applied to a symbol @emph{definition}, an error is reported.
4286 If a symbol previously declared @code{dllimport} is later defined, the
4287 attribute is ignored in subsequent references, and a warning is emitted.
4288 The attribute is also overridden by a subsequent declaration as
4291 When applied to C++ classes, the attribute marks non-inlined
4292 member functions and static data members as imports. However, the
4293 attribute is ignored for virtual methods to allow creation of vtables
4296 On the SH Symbian OS target the @code{dllimport} attribute also has
4297 another affect---it can cause the vtable and run-time type information
4298 for a class to be exported. This happens when the class has a
4299 dllimported constructor or a non-inline, non-pure virtual function
4300 and, for either of those two conditions, the class also has an inline
4301 constructor or destructor and has a key function that is defined in
4302 the current translation unit.
4304 For Microsoft Windows targets the use of the @code{dllimport}
4305 attribute on functions is not necessary, but provides a small
4306 performance benefit by eliminating a thunk in the DLL@. The use of the
4307 @code{dllimport} attribute on imported variables can be avoided by passing the
4308 @option{--enable-auto-import} switch to the GNU linker. As with
4309 functions, using the attribute for a variable eliminates a thunk in
4312 One drawback to using this attribute is that a pointer to a
4313 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4314 address. However, a pointer to a @emph{function} with the
4315 @code{dllimport} attribute can be used as a constant initializer; in
4316 this case, the address of a stub function in the import lib is
4317 referenced. On Microsoft Windows targets, the attribute can be disabled
4318 for functions by setting the @option{-mnop-fun-dllimport} flag.
4321 @node MIPS Function Attributes
4322 @subsection MIPS Function Attributes
4324 These function attributes are supported by the MIPS back end:
4328 @cindex @code{interrupt} function attribute, MIPS
4329 Use this attribute to indicate that the specified function is an interrupt
4330 handler. The compiler generates function entry and exit sequences suitable
4331 for use in an interrupt handler when this attribute is present.
4332 An optional argument is supported for the interrupt attribute which allows
4333 the interrupt mode to be described. By default GCC assumes the external
4334 interrupt controller (EIC) mode is in use, this can be explicitly set using
4335 @code{eic}. When interrupts are non-masked then the requested Interrupt
4336 Priority Level (IPL) is copied to the current IPL which has the effect of only
4337 enabling higher priority interrupts. To use vectored interrupt mode use
4338 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4339 the behaviour of the non-masked interrupt support and GCC will arrange to mask
4340 all interrupts from sw0 up to and including the specified interrupt vector.
4342 You can use the following attributes to modify the behavior
4343 of an interrupt handler:
4345 @item use_shadow_register_set
4346 @cindex @code{use_shadow_register_set} function attribute, MIPS
4347 Assume that the handler uses a shadow register set, instead of
4348 the main general-purpose registers. An optional argument @code{intstack} is
4349 supported to indicate that the shadow register set contains a valid stack
4352 @item keep_interrupts_masked
4353 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4354 Keep interrupts masked for the whole function. Without this attribute,
4355 GCC tries to reenable interrupts for as much of the function as it can.
4357 @item use_debug_exception_return
4358 @cindex @code{use_debug_exception_return} function attribute, MIPS
4359 Return using the @code{deret} instruction. Interrupt handlers that don't
4360 have this attribute return using @code{eret} instead.
4363 You can use any combination of these attributes, as shown below:
4365 void __attribute__ ((interrupt)) v0 ();
4366 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4367 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4368 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4369 void __attribute__ ((interrupt, use_shadow_register_set,
4370 keep_interrupts_masked)) v4 ();
4371 void __attribute__ ((interrupt, use_shadow_register_set,
4372 use_debug_exception_return)) v5 ();
4373 void __attribute__ ((interrupt, keep_interrupts_masked,
4374 use_debug_exception_return)) v6 ();
4375 void __attribute__ ((interrupt, use_shadow_register_set,
4376 keep_interrupts_masked,
4377 use_debug_exception_return)) v7 ();
4378 void __attribute__ ((interrupt("eic"))) v8 ();
4379 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4385 @cindex indirect calls, MIPS
4386 @cindex @code{long_call} function attribute, MIPS
4387 @cindex @code{near} function attribute, MIPS
4388 @cindex @code{far} function attribute, MIPS
4389 These attributes specify how a particular function is called on MIPS@.
4390 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4391 command-line switch. The @code{long_call} and @code{far} attributes are
4392 synonyms, and cause the compiler to always call
4393 the function by first loading its address into a register, and then using
4394 the contents of that register. The @code{near} attribute has the opposite
4395 effect; it specifies that non-PIC calls should be made using the more
4396 efficient @code{jal} instruction.
4400 @cindex @code{mips16} function attribute, MIPS
4401 @cindex @code{nomips16} function attribute, MIPS
4403 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4404 function attributes to locally select or turn off MIPS16 code generation.
4405 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4406 while MIPS16 code generation is disabled for functions with the
4407 @code{nomips16} attribute. These attributes override the
4408 @option{-mips16} and @option{-mno-mips16} options on the command line
4409 (@pxref{MIPS Options}).
4411 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4412 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4413 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4414 may interact badly with some GCC extensions such as @code{__builtin_apply}
4415 (@pxref{Constructing Calls}).
4417 @item micromips, MIPS
4418 @itemx nomicromips, MIPS
4419 @cindex @code{micromips} function attribute
4420 @cindex @code{nomicromips} function attribute
4422 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4423 function attributes to locally select or turn off microMIPS code generation.
4424 A function with the @code{micromips} attribute is emitted as microMIPS code,
4425 while microMIPS code generation is disabled for functions with the
4426 @code{nomicromips} attribute. These attributes override the
4427 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4428 (@pxref{MIPS Options}).
4430 When compiling files containing mixed microMIPS and non-microMIPS code, the
4431 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4433 not that within individual functions. Mixed microMIPS and non-microMIPS code
4434 may interact badly with some GCC extensions such as @code{__builtin_apply}
4435 (@pxref{Constructing Calls}).
4438 @cindex @code{nocompression} function attribute, MIPS
4439 On MIPS targets, you can use the @code{nocompression} function attribute
4440 to locally turn off MIPS16 and microMIPS code generation. This attribute
4441 overrides the @option{-mips16} and @option{-mmicromips} options on the
4442 command line (@pxref{MIPS Options}).
4445 @node MSP430 Function Attributes
4446 @subsection MSP430 Function Attributes
4448 These function attributes are supported by the MSP430 back end:
4452 @cindex @code{critical} function attribute, MSP430
4453 Critical functions disable interrupts upon entry and restore the
4454 previous interrupt state upon exit. Critical functions cannot also
4455 have the @code{naked} or @code{reentrant} attributes. They can have
4456 the @code{interrupt} attribute.
4459 @cindex @code{interrupt} function attribute, MSP430
4460 Use this attribute to indicate
4461 that the specified function is an interrupt handler. The compiler generates
4462 function entry and exit sequences suitable for use in an interrupt handler
4463 when this attribute is present.
4465 You can provide an argument to the interrupt
4466 attribute which specifies a name or number. If the argument is a
4467 number it indicates the slot in the interrupt vector table (0 - 31) to
4468 which this handler should be assigned. If the argument is a name it
4469 is treated as a symbolic name for the vector slot. These names should
4470 match up with appropriate entries in the linker script. By default
4471 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4472 @code{reset} for vector 31 are recognized.
4475 @cindex @code{naked} function attribute, MSP430
4476 This attribute allows the compiler to construct the
4477 requisite function declaration, while allowing the body of the
4478 function to be assembly code. The specified function will not have
4479 prologue/epilogue sequences generated by the compiler. Only basic
4480 @code{asm} statements can safely be included in naked functions
4481 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4482 basic @code{asm} and C code may appear to work, they cannot be
4483 depended upon to work reliably and are not supported.
4486 @cindex @code{reentrant} function attribute, MSP430
4487 Reentrant functions disable interrupts upon entry and enable them
4488 upon exit. Reentrant functions cannot also have the @code{naked}
4489 or @code{critical} attributes. They can have the @code{interrupt}
4493 @cindex @code{wakeup} function attribute, MSP430
4494 This attribute only applies to interrupt functions. It is silently
4495 ignored if applied to a non-interrupt function. A wakeup interrupt
4496 function will rouse the processor from any low-power state that it
4497 might be in when the function exits.
4502 @cindex @code{lower} function attribute, MSP430
4503 @cindex @code{upper} function attribute, MSP430
4504 @cindex @code{either} function attribute, MSP430
4505 On the MSP430 target these attributes can be used to specify whether
4506 the function or variable should be placed into low memory, high
4507 memory, or the placement should be left to the linker to decide. The
4508 attributes are only significant if compiling for the MSP430X
4511 The attributes work in conjunction with a linker script that has been
4512 augmented to specify where to place sections with a @code{.lower} and
4513 a @code{.upper} prefix. So, for example, as well as placing the
4514 @code{.data} section, the script also specifies the placement of a
4515 @code{.lower.data} and a @code{.upper.data} section. The intention
4516 is that @code{lower} sections are placed into a small but easier to
4517 access memory region and the upper sections are placed into a larger, but
4518 slower to access, region.
4520 The @code{either} attribute is special. It tells the linker to place
4521 the object into the corresponding @code{lower} section if there is
4522 room for it. If there is insufficient room then the object is placed
4523 into the corresponding @code{upper} section instead. Note that the
4524 placement algorithm is not very sophisticated. It does not attempt to
4525 find an optimal packing of the @code{lower} sections. It just makes
4526 one pass over the objects and does the best that it can. Using the
4527 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4528 options can help the packing, however, since they produce smaller,
4529 easier to pack regions.
4532 @node NDS32 Function Attributes
4533 @subsection NDS32 Function Attributes
4535 These function attributes are supported by the NDS32 back end:
4539 @cindex @code{exception} function attribute
4540 @cindex exception handler functions, NDS32
4541 Use this attribute on the NDS32 target to indicate that the specified function
4542 is an exception handler. The compiler will generate corresponding sections
4543 for use in an exception handler.
4546 @cindex @code{interrupt} function attribute, NDS32
4547 On NDS32 target, this attribute indicates that the specified function
4548 is an interrupt handler. The compiler generates corresponding sections
4549 for use in an interrupt handler. You can use the following attributes
4550 to modify the behavior:
4553 @cindex @code{nested} function attribute, NDS32
4554 This interrupt service routine is interruptible.
4556 @cindex @code{not_nested} function attribute, NDS32
4557 This interrupt service routine is not interruptible.
4559 @cindex @code{nested_ready} function attribute, NDS32
4560 This interrupt service routine is interruptible after @code{PSW.GIE}
4561 (global interrupt enable) is set. This allows interrupt service routine to
4562 finish some short critical code before enabling interrupts.
4564 @cindex @code{save_all} function attribute, NDS32
4565 The system will help save all registers into stack before entering
4568 @cindex @code{partial_save} function attribute, NDS32
4569 The system will help save caller registers into stack before entering
4574 @cindex @code{naked} function attribute, NDS32
4575 This attribute allows the compiler to construct the
4576 requisite function declaration, while allowing the body of the
4577 function to be assembly code. The specified function will not have
4578 prologue/epilogue sequences generated by the compiler. Only basic
4579 @code{asm} statements can safely be included in naked functions
4580 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4581 basic @code{asm} and C code may appear to work, they cannot be
4582 depended upon to work reliably and are not supported.
4585 @cindex @code{reset} function attribute, NDS32
4586 @cindex reset handler functions
4587 Use this attribute on the NDS32 target to indicate that the specified function
4588 is a reset handler. The compiler will generate corresponding sections
4589 for use in a reset handler. You can use the following attributes
4590 to provide extra exception handling:
4593 @cindex @code{nmi} function attribute, NDS32
4594 Provide a user-defined function to handle NMI exception.
4596 @cindex @code{warm} function attribute, NDS32
4597 Provide a user-defined function to handle warm reset exception.
4601 @node Nios II Function Attributes
4602 @subsection Nios II Function Attributes
4604 These function attributes are supported by the Nios II back end:
4607 @item target (@var{options})
4608 @cindex @code{target} function attribute
4609 As discussed in @ref{Common Function Attributes}, this attribute
4610 allows specification of target-specific compilation options.
4612 When compiling for Nios II, the following options are allowed:
4615 @item custom-@var{insn}=@var{N}
4616 @itemx no-custom-@var{insn}
4617 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4618 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4619 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4620 custom instruction with encoding @var{N} when generating code that uses
4621 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4622 the custom instruction @var{insn}.
4623 These target attributes correspond to the
4624 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4625 command-line options, and support the same set of @var{insn} keywords.
4626 @xref{Nios II Options}, for more information.
4628 @item custom-fpu-cfg=@var{name}
4629 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4630 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4631 command-line option, to select a predefined set of custom instructions
4633 @xref{Nios II Options}, for more information.
4637 @node Nvidia PTX Function Attributes
4638 @subsection Nvidia PTX Function Attributes
4640 These function attributes are supported by the Nvidia PTX back end:
4644 @cindex @code{kernel} attribute, Nvidia PTX
4645 This attribute indicates that the corresponding function should be compiled
4646 as a kernel function, which can be invoked from the host via the CUDA RT
4648 By default functions are only callable only from other PTX functions.
4650 Kernel functions must have @code{void} return type.
4653 @node PowerPC Function Attributes
4654 @subsection PowerPC Function Attributes
4656 These function attributes are supported by the PowerPC back end:
4661 @cindex indirect calls, PowerPC
4662 @cindex @code{longcall} function attribute, PowerPC
4663 @cindex @code{shortcall} function attribute, PowerPC
4664 The @code{longcall} attribute
4665 indicates that the function might be far away from the call site and
4666 require a different (more expensive) calling sequence. The
4667 @code{shortcall} attribute indicates that the function is always close
4668 enough for the shorter calling sequence to be used. These attributes
4669 override both the @option{-mlongcall} switch and
4670 the @code{#pragma longcall} setting.
4672 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4673 calls are necessary.
4675 @item target (@var{options})
4676 @cindex @code{target} function attribute
4677 As discussed in @ref{Common Function Attributes}, this attribute
4678 allows specification of target-specific compilation options.
4680 On the PowerPC, the following options are allowed:
4685 @cindex @code{target("altivec")} function attribute, PowerPC
4686 Generate code that uses (does not use) AltiVec instructions. In
4687 32-bit code, you cannot enable AltiVec instructions unless
4688 @option{-mabi=altivec} is used on the command line.
4692 @cindex @code{target("cmpb")} function attribute, PowerPC
4693 Generate code that uses (does not use) the compare bytes instruction
4694 implemented on the POWER6 processor and other processors that support
4695 the PowerPC V2.05 architecture.
4699 @cindex @code{target("dlmzb")} function attribute, PowerPC
4700 Generate code that uses (does not use) the string-search @samp{dlmzb}
4701 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4702 generated by default when targeting those processors.
4706 @cindex @code{target("fprnd")} function attribute, PowerPC
4707 Generate code that uses (does not use) the FP round to integer
4708 instructions implemented on the POWER5+ processor and other processors
4709 that support the PowerPC V2.03 architecture.
4713 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4714 Generate code that uses (does not use) the decimal floating-point
4715 instructions implemented on some POWER processors.
4719 @cindex @code{target("isel")} function attribute, PowerPC
4720 Generate code that uses (does not use) ISEL instruction.
4724 @cindex @code{target("mfcrf")} function attribute, PowerPC
4725 Generate code that uses (does not use) the move from condition
4726 register field instruction implemented on the POWER4 processor and
4727 other processors that support the PowerPC V2.01 architecture.
4731 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4732 Generate code that uses (does not use) the FP move to/from general
4733 purpose register instructions implemented on the POWER6X processor and
4734 other processors that support the extended PowerPC V2.05 architecture.
4738 @cindex @code{target("mulhw")} function attribute, PowerPC
4739 Generate code that uses (does not use) the half-word multiply and
4740 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4741 These instructions are generated by default when targeting those
4746 @cindex @code{target("multiple")} function attribute, PowerPC
4747 Generate code that uses (does not use) the load multiple word
4748 instructions and the store multiple word instructions.
4752 @cindex @code{target("update")} function attribute, PowerPC
4753 Generate code that uses (does not use) the load or store instructions
4754 that update the base register to the address of the calculated memory
4759 @cindex @code{target("popcntb")} function attribute, PowerPC
4760 Generate code that uses (does not use) the popcount and double-precision
4761 FP reciprocal estimate instruction implemented on the POWER5
4762 processor and other processors that support the PowerPC V2.02
4767 @cindex @code{target("popcntd")} function attribute, PowerPC
4768 Generate code that uses (does not use) the popcount instruction
4769 implemented on the POWER7 processor and other processors that support
4770 the PowerPC V2.06 architecture.
4772 @item powerpc-gfxopt
4773 @itemx no-powerpc-gfxopt
4774 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4775 Generate code that uses (does not use) the optional PowerPC
4776 architecture instructions in the Graphics group, including
4777 floating-point select.
4780 @itemx no-powerpc-gpopt
4781 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4782 Generate code that uses (does not use) the optional PowerPC
4783 architecture instructions in the General Purpose group, including
4784 floating-point square root.
4786 @item recip-precision
4787 @itemx no-recip-precision
4788 @cindex @code{target("recip-precision")} function attribute, PowerPC
4789 Assume (do not assume) that the reciprocal estimate instructions
4790 provide higher-precision estimates than is mandated by the PowerPC
4795 @cindex @code{target("string")} function attribute, PowerPC
4796 Generate code that uses (does not use) the load string instructions
4797 and the store string word instructions to save multiple registers and
4798 do small block moves.
4802 @cindex @code{target("vsx")} function attribute, PowerPC
4803 Generate code that uses (does not use) vector/scalar (VSX)
4804 instructions, and also enable the use of built-in functions that allow
4805 more direct access to the VSX instruction set. In 32-bit code, you
4806 cannot enable VSX or AltiVec instructions unless
4807 @option{-mabi=altivec} is used on the command line.
4811 @cindex @code{target("friz")} function attribute, PowerPC
4812 Generate (do not generate) the @code{friz} instruction when the
4813 @option{-funsafe-math-optimizations} option is used to optimize
4814 rounding a floating-point value to 64-bit integer and back to floating
4815 point. The @code{friz} instruction does not return the same value if
4816 the floating-point number is too large to fit in an integer.
4818 @item avoid-indexed-addresses
4819 @itemx no-avoid-indexed-addresses
4820 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4821 Generate code that tries to avoid (not avoid) the use of indexed load
4822 or store instructions.
4826 @cindex @code{target("paired")} function attribute, PowerPC
4827 Generate code that uses (does not use) the generation of PAIRED simd
4832 @cindex @code{target("longcall")} function attribute, PowerPC
4833 Generate code that assumes (does not assume) that all calls are far
4834 away so that a longer more expensive calling sequence is required.
4837 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4838 Specify the architecture to generate code for when compiling the
4839 function. If you select the @code{target("cpu=power7")} attribute when
4840 generating 32-bit code, VSX and AltiVec instructions are not generated
4841 unless you use the @option{-mabi=altivec} option on the command line.
4843 @item tune=@var{TUNE}
4844 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4845 Specify the architecture to tune for when compiling the function. If
4846 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4847 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4848 compilation tunes for the @var{CPU} architecture, and not the
4849 default tuning specified on the command line.
4852 On the PowerPC, the inliner does not inline a
4853 function that has different target options than the caller, unless the
4854 callee has a subset of the target options of the caller.
4857 @node RL78 Function Attributes
4858 @subsection RL78 Function Attributes
4860 These function attributes are supported by the RL78 back end:
4864 @itemx brk_interrupt
4865 @cindex @code{interrupt} function attribute, RL78
4866 @cindex @code{brk_interrupt} function attribute, RL78
4867 These attributes indicate
4868 that the specified function is an interrupt handler. The compiler generates
4869 function entry and exit sequences suitable for use in an interrupt handler
4870 when this attribute is present.
4872 Use @code{brk_interrupt} instead of @code{interrupt} for
4873 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4874 that must end with @code{RETB} instead of @code{RETI}).
4877 @cindex @code{naked} function attribute, RL78
4878 This attribute allows the compiler to construct the
4879 requisite function declaration, while allowing the body of the
4880 function to be assembly code. The specified function will not have
4881 prologue/epilogue sequences generated by the compiler. Only basic
4882 @code{asm} statements can safely be included in naked functions
4883 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4884 basic @code{asm} and C code may appear to work, they cannot be
4885 depended upon to work reliably and are not supported.
4888 @node RX Function Attributes
4889 @subsection RX Function Attributes
4891 These function attributes are supported by the RX back end:
4894 @item fast_interrupt
4895 @cindex @code{fast_interrupt} function attribute, RX
4896 Use this attribute on the RX port to indicate that the specified
4897 function is a fast interrupt handler. This is just like the
4898 @code{interrupt} attribute, except that @code{freit} is used to return
4899 instead of @code{reit}.
4902 @cindex @code{interrupt} function attribute, RX
4903 Use this attribute to indicate
4904 that the specified function is an interrupt handler. The compiler generates
4905 function entry and exit sequences suitable for use in an interrupt handler
4906 when this attribute is present.
4908 On RX targets, you may specify one or more vector numbers as arguments
4909 to the attribute, as well as naming an alternate table name.
4910 Parameters are handled sequentially, so one handler can be assigned to
4911 multiple entries in multiple tables. One may also pass the magic
4912 string @code{"$default"} which causes the function to be used for any
4913 unfilled slots in the current table.
4915 This example shows a simple assignment of a function to one vector in
4916 the default table (note that preprocessor macros may be used for
4917 chip-specific symbolic vector names):
4919 void __attribute__ ((interrupt (5))) txd1_handler ();
4922 This example assigns a function to two slots in the default table
4923 (using preprocessor macros defined elsewhere) and makes it the default
4924 for the @code{dct} table:
4926 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4931 @cindex @code{naked} function attribute, RX
4932 This attribute allows the compiler to construct the
4933 requisite function declaration, while allowing the body of the
4934 function to be assembly code. The specified function will not have
4935 prologue/epilogue sequences generated by the compiler. Only basic
4936 @code{asm} statements can safely be included in naked functions
4937 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4938 basic @code{asm} and C code may appear to work, they cannot be
4939 depended upon to work reliably and are not supported.
4942 @cindex @code{vector} function attribute, RX
4943 This RX attribute is similar to the @code{interrupt} attribute, including its
4944 parameters, but does not make the function an interrupt-handler type
4945 function (i.e. it retains the normal C function calling ABI). See the
4946 @code{interrupt} attribute for a description of its arguments.
4949 @node S/390 Function Attributes
4950 @subsection S/390 Function Attributes
4952 These function attributes are supported on the S/390:
4955 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4956 @cindex @code{hotpatch} function attribute, S/390
4958 On S/390 System z targets, you can use this function attribute to
4959 make GCC generate a ``hot-patching'' function prologue. If the
4960 @option{-mhotpatch=} command-line option is used at the same time,
4961 the @code{hotpatch} attribute takes precedence. The first of the
4962 two arguments specifies the number of halfwords to be added before
4963 the function label. A second argument can be used to specify the
4964 number of halfwords to be added after the function label. For
4965 both arguments the maximum allowed value is 1000000.
4967 If both arguments are zero, hotpatching is disabled.
4969 @item target (@var{options})
4970 @cindex @code{target} function attribute
4971 As discussed in @ref{Common Function Attributes}, this attribute
4972 allows specification of target-specific compilation options.
4974 On S/390, the following options are supported:
4982 @item warn-framesize=
4994 @itemx no-packed-stack
4996 @itemx no-small-exec
4999 @item warn-dynamicstack
5000 @itemx no-warn-dynamicstack
5003 The options work exactly like the S/390 specific command line
5004 options (without the prefix @option{-m}) except that they do not
5005 change any feature macros. For example,
5008 @code{target("no-vx")}
5011 does not undefine the @code{__VEC__} macro.
5014 @node SH Function Attributes
5015 @subsection SH Function Attributes
5017 These function attributes are supported on the SH family of processors:
5020 @item function_vector
5021 @cindex @code{function_vector} function attribute, SH
5022 @cindex calling functions through the function vector on SH2A
5023 On SH2A targets, this attribute declares a function to be called using the
5024 TBR relative addressing mode. The argument to this attribute is the entry
5025 number of the same function in a vector table containing all the TBR
5026 relative addressable functions. For correct operation the TBR must be setup
5027 accordingly to point to the start of the vector table before any functions with
5028 this attribute are invoked. Usually a good place to do the initialization is
5029 the startup routine. The TBR relative vector table can have at max 256 function
5030 entries. The jumps to these functions are generated using a SH2A specific,
5031 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5032 from GNU binutils version 2.7 or later for this attribute to work correctly.
5034 In an application, for a function being called once, this attribute
5035 saves at least 8 bytes of code; and if other successive calls are being
5036 made to the same function, it saves 2 bytes of code per each of these
5039 @item interrupt_handler
5040 @cindex @code{interrupt_handler} function attribute, SH
5041 Use this attribute to
5042 indicate that the specified function is an interrupt handler. The compiler
5043 generates function entry and exit sequences suitable for use in an
5044 interrupt handler when this attribute is present.
5046 @item nosave_low_regs
5047 @cindex @code{nosave_low_regs} function attribute, SH
5048 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5049 function should not save and restore registers R0..R7. This can be used on SH3*
5050 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5054 @cindex @code{renesas} function attribute, SH
5055 On SH targets this attribute specifies that the function or struct follows the
5059 @cindex @code{resbank} function attribute, SH
5060 On the SH2A target, this attribute enables the high-speed register
5061 saving and restoration using a register bank for @code{interrupt_handler}
5062 routines. Saving to the bank is performed automatically after the CPU
5063 accepts an interrupt that uses a register bank.
5065 The nineteen 32-bit registers comprising general register R0 to R14,
5066 control register GBR, and system registers MACH, MACL, and PR and the
5067 vector table address offset are saved into a register bank. Register
5068 banks are stacked in first-in last-out (FILO) sequence. Restoration
5069 from the bank is executed by issuing a RESBANK instruction.
5072 @cindex @code{sp_switch} function attribute, SH
5073 Use this attribute on the SH to indicate an @code{interrupt_handler}
5074 function should switch to an alternate stack. It expects a string
5075 argument that names a global variable holding the address of the
5080 void f () __attribute__ ((interrupt_handler,
5081 sp_switch ("alt_stack")));
5085 @cindex @code{trap_exit} function attribute, SH
5086 Use this attribute on the SH for an @code{interrupt_handler} to return using
5087 @code{trapa} instead of @code{rte}. This attribute expects an integer
5088 argument specifying the trap number to be used.
5091 @cindex @code{trapa_handler} function attribute, SH
5092 On SH targets this function attribute is similar to @code{interrupt_handler}
5093 but it does not save and restore all registers.
5096 @node SPU Function Attributes
5097 @subsection SPU Function Attributes
5099 These function attributes are supported by the SPU back end:
5103 @cindex @code{naked} function attribute, SPU
5104 This attribute allows the compiler to construct the
5105 requisite function declaration, while allowing the body of the
5106 function to be assembly code. The specified function will not have
5107 prologue/epilogue sequences generated by the compiler. Only basic
5108 @code{asm} statements can safely be included in naked functions
5109 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5110 basic @code{asm} and C code may appear to work, they cannot be
5111 depended upon to work reliably and are not supported.
5114 @node Symbian OS Function Attributes
5115 @subsection Symbian OS Function Attributes
5117 @xref{Microsoft Windows Function Attributes}, for discussion of the
5118 @code{dllexport} and @code{dllimport} attributes.
5120 @node V850 Function Attributes
5121 @subsection V850 Function Attributes
5123 The V850 back end supports these function attributes:
5127 @itemx interrupt_handler
5128 @cindex @code{interrupt} function attribute, V850
5129 @cindex @code{interrupt_handler} function attribute, V850
5130 Use these attributes to indicate
5131 that the specified function is an interrupt handler. The compiler generates
5132 function entry and exit sequences suitable for use in an interrupt handler
5133 when either attribute is present.
5136 @node Visium Function Attributes
5137 @subsection Visium Function Attributes
5139 These function attributes are supported by the Visium back end:
5143 @cindex @code{interrupt} function attribute, Visium
5144 Use this attribute to indicate
5145 that the specified function is an interrupt handler. The compiler generates
5146 function entry and exit sequences suitable for use in an interrupt handler
5147 when this attribute is present.
5150 @node x86 Function Attributes
5151 @subsection x86 Function Attributes
5153 These function attributes are supported by the x86 back end:
5157 @cindex @code{cdecl} function attribute, x86-32
5158 @cindex functions that pop the argument stack on x86-32
5160 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5161 assume that the calling function pops off the stack space used to
5162 pass arguments. This is
5163 useful to override the effects of the @option{-mrtd} switch.
5166 @cindex @code{fastcall} function attribute, x86-32
5167 @cindex functions that pop the argument stack on x86-32
5168 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5169 pass the first argument (if of integral type) in the register ECX and
5170 the second argument (if of integral type) in the register EDX@. Subsequent
5171 and other typed arguments are passed on the stack. The called function
5172 pops the arguments off the stack. If the number of arguments is variable all
5173 arguments are pushed on the stack.
5176 @cindex @code{thiscall} function attribute, x86-32
5177 @cindex functions that pop the argument stack on x86-32
5178 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5179 pass the first argument (if of integral type) in the register ECX.
5180 Subsequent and other typed arguments are passed on the stack. The called
5181 function pops the arguments off the stack.
5182 If the number of arguments is variable all arguments are pushed on the
5184 The @code{thiscall} attribute is intended for C++ non-static member functions.
5185 As a GCC extension, this calling convention can be used for C functions
5186 and for static member methods.
5190 @cindex @code{ms_abi} function attribute, x86
5191 @cindex @code{sysv_abi} function attribute, x86
5193 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5194 to indicate which calling convention should be used for a function. The
5195 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5196 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5197 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5198 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5200 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5201 requires the @option{-maccumulate-outgoing-args} option.
5203 @item callee_pop_aggregate_return (@var{number})
5204 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5206 On x86-32 targets, you can use this attribute to control how
5207 aggregates are returned in memory. If the caller is responsible for
5208 popping the hidden pointer together with the rest of the arguments, specify
5209 @var{number} equal to zero. If callee is responsible for popping the
5210 hidden pointer, specify @var{number} equal to one.
5212 The default x86-32 ABI assumes that the callee pops the
5213 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5214 the compiler assumes that the
5215 caller pops the stack for hidden pointer.
5217 @item ms_hook_prologue
5218 @cindex @code{ms_hook_prologue} function attribute, x86
5220 On 32-bit and 64-bit x86 targets, you can use
5221 this function attribute to make GCC generate the ``hot-patching'' function
5222 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5225 @item regparm (@var{number})
5226 @cindex @code{regparm} function attribute, x86
5227 @cindex functions that are passed arguments in registers on x86-32
5228 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5229 pass arguments number one to @var{number} if they are of integral type
5230 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5231 take a variable number of arguments continue to be passed all of their
5232 arguments on the stack.
5234 Beware that on some ELF systems this attribute is unsuitable for
5235 global functions in shared libraries with lazy binding (which is the
5236 default). Lazy binding sends the first call via resolving code in
5237 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5238 per the standard calling conventions. Solaris 8 is affected by this.
5239 Systems with the GNU C Library version 2.1 or higher
5240 and FreeBSD are believed to be
5241 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5242 disabled with the linker or the loader if desired, to avoid the
5246 @cindex @code{sseregparm} function attribute, x86
5247 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5248 causes the compiler to pass up to 3 floating-point arguments in
5249 SSE registers instead of on the stack. Functions that take a
5250 variable number of arguments continue to pass all of their
5251 floating-point arguments on the stack.
5253 @item force_align_arg_pointer
5254 @cindex @code{force_align_arg_pointer} function attribute, x86
5255 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5256 applied to individual function definitions, generating an alternate
5257 prologue and epilogue that realigns the run-time stack if necessary.
5258 This supports mixing legacy codes that run with a 4-byte aligned stack
5259 with modern codes that keep a 16-byte stack for SSE compatibility.
5262 @cindex @code{stdcall} function attribute, x86-32
5263 @cindex functions that pop the argument stack on x86-32
5264 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5265 assume that the called function pops off the stack space used to
5266 pass arguments, unless it takes a variable number of arguments.
5268 @item target (@var{options})
5269 @cindex @code{target} function attribute
5270 As discussed in @ref{Common Function Attributes}, this attribute
5271 allows specification of target-specific compilation options.
5273 On the x86, the following options are allowed:
5277 @cindex @code{target("abm")} function attribute, x86
5278 Enable/disable the generation of the advanced bit instructions.
5282 @cindex @code{target("aes")} function attribute, x86
5283 Enable/disable the generation of the AES instructions.
5286 @cindex @code{target("default")} function attribute, x86
5287 @xref{Function Multiversioning}, where it is used to specify the
5288 default function version.
5292 @cindex @code{target("mmx")} function attribute, x86
5293 Enable/disable the generation of the MMX instructions.
5297 @cindex @code{target("pclmul")} function attribute, x86
5298 Enable/disable the generation of the PCLMUL instructions.
5302 @cindex @code{target("popcnt")} function attribute, x86
5303 Enable/disable the generation of the POPCNT instruction.
5307 @cindex @code{target("sse")} function attribute, x86
5308 Enable/disable the generation of the SSE instructions.
5312 @cindex @code{target("sse2")} function attribute, x86
5313 Enable/disable the generation of the SSE2 instructions.
5317 @cindex @code{target("sse3")} function attribute, x86
5318 Enable/disable the generation of the SSE3 instructions.
5322 @cindex @code{target("sse4")} function attribute, x86
5323 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5328 @cindex @code{target("sse4.1")} function attribute, x86
5329 Enable/disable the generation of the sse4.1 instructions.
5333 @cindex @code{target("sse4.2")} function attribute, x86
5334 Enable/disable the generation of the sse4.2 instructions.
5338 @cindex @code{target("sse4a")} function attribute, x86
5339 Enable/disable the generation of the SSE4A instructions.
5343 @cindex @code{target("fma4")} function attribute, x86
5344 Enable/disable the generation of the FMA4 instructions.
5348 @cindex @code{target("xop")} function attribute, x86
5349 Enable/disable the generation of the XOP instructions.
5353 @cindex @code{target("lwp")} function attribute, x86
5354 Enable/disable the generation of the LWP instructions.
5358 @cindex @code{target("ssse3")} function attribute, x86
5359 Enable/disable the generation of the SSSE3 instructions.
5363 @cindex @code{target("cld")} function attribute, x86
5364 Enable/disable the generation of the CLD before string moves.
5366 @item fancy-math-387
5367 @itemx no-fancy-math-387
5368 @cindex @code{target("fancy-math-387")} function attribute, x86
5369 Enable/disable the generation of the @code{sin}, @code{cos}, and
5370 @code{sqrt} instructions on the 387 floating-point unit.
5373 @itemx no-fused-madd
5374 @cindex @code{target("fused-madd")} function attribute, x86
5375 Enable/disable the generation of the fused multiply/add instructions.
5379 @cindex @code{target("ieee-fp")} function attribute, x86
5380 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5382 @item inline-all-stringops
5383 @itemx no-inline-all-stringops
5384 @cindex @code{target("inline-all-stringops")} function attribute, x86
5385 Enable/disable inlining of string operations.
5387 @item inline-stringops-dynamically
5388 @itemx no-inline-stringops-dynamically
5389 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5390 Enable/disable the generation of the inline code to do small string
5391 operations and calling the library routines for large operations.
5393 @item align-stringops
5394 @itemx no-align-stringops
5395 @cindex @code{target("align-stringops")} function attribute, x86
5396 Do/do not align destination of inlined string operations.
5400 @cindex @code{target("recip")} function attribute, x86
5401 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5402 instructions followed an additional Newton-Raphson step instead of
5403 doing a floating-point division.
5405 @item arch=@var{ARCH}
5406 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5407 Specify the architecture to generate code for in compiling the function.
5409 @item tune=@var{TUNE}
5410 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5411 Specify the architecture to tune for in compiling the function.
5413 @item fpmath=@var{FPMATH}
5414 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5415 Specify which floating-point unit to use. You must specify the
5416 @code{target("fpmath=sse,387")} option as
5417 @code{target("fpmath=sse+387")} because the comma would separate
5421 On the x86, the inliner does not inline a
5422 function that has different target options than the caller, unless the
5423 callee has a subset of the target options of the caller. For example
5424 a function declared with @code{target("sse3")} can inline a function
5425 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5428 @node Xstormy16 Function Attributes
5429 @subsection Xstormy16 Function Attributes
5431 These function attributes are supported by the Xstormy16 back end:
5435 @cindex @code{interrupt} function attribute, Xstormy16
5436 Use this attribute to indicate
5437 that the specified function is an interrupt handler. The compiler generates
5438 function entry and exit sequences suitable for use in an interrupt handler
5439 when this attribute is present.
5442 @node Variable Attributes
5443 @section Specifying Attributes of Variables
5444 @cindex attribute of variables
5445 @cindex variable attributes
5447 The keyword @code{__attribute__} allows you to specify special
5448 attributes of variables or structure fields. This keyword is followed
5449 by an attribute specification inside double parentheses. Some
5450 attributes are currently defined generically for variables.
5451 Other attributes are defined for variables on particular target
5452 systems. Other attributes are available for functions
5453 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5454 enumerators (@pxref{Enumerator Attributes}), and for types
5455 (@pxref{Type Attributes}).
5456 Other front ends might define more attributes
5457 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5459 @xref{Attribute Syntax}, for details of the exact syntax for using
5463 * Common Variable Attributes::
5464 * AVR Variable Attributes::
5465 * Blackfin Variable Attributes::
5466 * H8/300 Variable Attributes::
5467 * IA-64 Variable Attributes::
5468 * M32R/D Variable Attributes::
5469 * MeP Variable Attributes::
5470 * Microsoft Windows Variable Attributes::
5471 * MSP430 Variable Attributes::
5472 * PowerPC Variable Attributes::
5473 * RL78 Variable Attributes::
5474 * SPU Variable Attributes::
5475 * V850 Variable Attributes::
5476 * x86 Variable Attributes::
5477 * Xstormy16 Variable Attributes::
5480 @node Common Variable Attributes
5481 @subsection Common Variable Attributes
5483 The following attributes are supported on most targets.
5486 @cindex @code{aligned} variable attribute
5487 @item aligned (@var{alignment})
5488 This attribute specifies a minimum alignment for the variable or
5489 structure field, measured in bytes. For example, the declaration:
5492 int x __attribute__ ((aligned (16))) = 0;
5496 causes the compiler to allocate the global variable @code{x} on a
5497 16-byte boundary. On a 68040, this could be used in conjunction with
5498 an @code{asm} expression to access the @code{move16} instruction which
5499 requires 16-byte aligned operands.
5501 You can also specify the alignment of structure fields. For example, to
5502 create a double-word aligned @code{int} pair, you could write:
5505 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5509 This is an alternative to creating a union with a @code{double} member,
5510 which forces the union to be double-word aligned.
5512 As in the preceding examples, you can explicitly specify the alignment
5513 (in bytes) that you wish the compiler to use for a given variable or
5514 structure field. Alternatively, you can leave out the alignment factor
5515 and just ask the compiler to align a variable or field to the
5516 default alignment for the target architecture you are compiling for.
5517 The default alignment is sufficient for all scalar types, but may not be
5518 enough for all vector types on a target that supports vector operations.
5519 The default alignment is fixed for a particular target ABI.
5521 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5522 which is the largest alignment ever used for any data type on the
5523 target machine you are compiling for. For example, you could write:
5526 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5529 The compiler automatically sets the alignment for the declared
5530 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5531 often make copy operations more efficient, because the compiler can
5532 use whatever instructions copy the biggest chunks of memory when
5533 performing copies to or from the variables or fields that you have
5534 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5535 may change depending on command-line options.
5537 When used on a struct, or struct member, the @code{aligned} attribute can
5538 only increase the alignment; in order to decrease it, the @code{packed}
5539 attribute must be specified as well. When used as part of a typedef, the
5540 @code{aligned} attribute can both increase and decrease alignment, and
5541 specifying the @code{packed} attribute generates a warning.
5543 Note that the effectiveness of @code{aligned} attributes may be limited
5544 by inherent limitations in your linker. On many systems, the linker is
5545 only able to arrange for variables to be aligned up to a certain maximum
5546 alignment. (For some linkers, the maximum supported alignment may
5547 be very very small.) If your linker is only able to align variables
5548 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5549 in an @code{__attribute__} still only provides you with 8-byte
5550 alignment. See your linker documentation for further information.
5552 The @code{aligned} attribute can also be used for functions
5553 (@pxref{Common Function Attributes}.)
5555 @item cleanup (@var{cleanup_function})
5556 @cindex @code{cleanup} variable attribute
5557 The @code{cleanup} attribute runs a function when the variable goes
5558 out of scope. This attribute can only be applied to auto function
5559 scope variables; it may not be applied to parameters or variables
5560 with static storage duration. The function must take one parameter,
5561 a pointer to a type compatible with the variable. The return value
5562 of the function (if any) is ignored.
5564 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5565 is run during the stack unwinding that happens during the
5566 processing of the exception. Note that the @code{cleanup} attribute
5567 does not allow the exception to be caught, only to perform an action.
5568 It is undefined what happens if @var{cleanup_function} does not
5573 @cindex @code{common} variable attribute
5574 @cindex @code{nocommon} variable attribute
5577 The @code{common} attribute requests GCC to place a variable in
5578 ``common'' storage. The @code{nocommon} attribute requests the
5579 opposite---to allocate space for it directly.
5581 These attributes override the default chosen by the
5582 @option{-fno-common} and @option{-fcommon} flags respectively.
5585 @itemx deprecated (@var{msg})
5586 @cindex @code{deprecated} variable attribute
5587 The @code{deprecated} attribute results in a warning if the variable
5588 is used anywhere in the source file. This is useful when identifying
5589 variables that are expected to be removed in a future version of a
5590 program. The warning also includes the location of the declaration
5591 of the deprecated variable, to enable users to easily find further
5592 information about why the variable is deprecated, or what they should
5593 do instead. Note that the warning only occurs for uses:
5596 extern int old_var __attribute__ ((deprecated));
5598 int new_fn () @{ return old_var; @}
5602 results in a warning on line 3 but not line 2. The optional @var{msg}
5603 argument, which must be a string, is printed in the warning if
5606 The @code{deprecated} attribute can also be used for functions and
5607 types (@pxref{Common Function Attributes},
5608 @pxref{Common Type Attributes}).
5610 @item mode (@var{mode})
5611 @cindex @code{mode} variable attribute
5612 This attribute specifies the data type for the declaration---whichever
5613 type corresponds to the mode @var{mode}. This in effect lets you
5614 request an integer or floating-point type according to its width.
5616 You may also specify a mode of @code{byte} or @code{__byte__} to
5617 indicate the mode corresponding to a one-byte integer, @code{word} or
5618 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5619 or @code{__pointer__} for the mode used to represent pointers.
5622 @cindex @code{packed} variable attribute
5623 The @code{packed} attribute specifies that a variable or structure field
5624 should have the smallest possible alignment---one byte for a variable,
5625 and one bit for a field, unless you specify a larger value with the
5626 @code{aligned} attribute.
5628 Here is a structure in which the field @code{x} is packed, so that it
5629 immediately follows @code{a}:
5635 int x[2] __attribute__ ((packed));
5639 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5640 @code{packed} attribute on bit-fields of type @code{char}. This has
5641 been fixed in GCC 4.4 but the change can lead to differences in the
5642 structure layout. See the documentation of
5643 @option{-Wpacked-bitfield-compat} for more information.
5645 @item section ("@var{section-name}")
5646 @cindex @code{section} variable attribute
5647 Normally, the compiler places the objects it generates in sections like
5648 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5649 or you need certain particular variables to appear in special sections,
5650 for example to map to special hardware. The @code{section}
5651 attribute specifies that a variable (or function) lives in a particular
5652 section. For example, this small program uses several specific section names:
5655 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5656 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5657 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5658 int init_data __attribute__ ((section ("INITDATA")));
5662 /* @r{Initialize stack pointer} */
5663 init_sp (stack + sizeof (stack));
5665 /* @r{Initialize initialized data} */
5666 memcpy (&init_data, &data, &edata - &data);
5668 /* @r{Turn on the serial ports} */
5675 Use the @code{section} attribute with
5676 @emph{global} variables and not @emph{local} variables,
5677 as shown in the example.
5679 You may use the @code{section} attribute with initialized or
5680 uninitialized global variables but the linker requires
5681 each object be defined once, with the exception that uninitialized
5682 variables tentatively go in the @code{common} (or @code{bss}) section
5683 and can be multiply ``defined''. Using the @code{section} attribute
5684 changes what section the variable goes into and may cause the
5685 linker to issue an error if an uninitialized variable has multiple
5686 definitions. You can force a variable to be initialized with the
5687 @option{-fno-common} flag or the @code{nocommon} attribute.
5689 Some file formats do not support arbitrary sections so the @code{section}
5690 attribute is not available on all platforms.
5691 If you need to map the entire contents of a module to a particular
5692 section, consider using the facilities of the linker instead.
5694 @item tls_model ("@var{tls_model}")
5695 @cindex @code{tls_model} variable attribute
5696 The @code{tls_model} attribute sets thread-local storage model
5697 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5698 overriding @option{-ftls-model=} command-line switch on a per-variable
5700 The @var{tls_model} argument should be one of @code{global-dynamic},
5701 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5703 Not all targets support this attribute.
5706 @cindex @code{unused} variable attribute
5707 This attribute, attached to a variable, means that the variable is meant
5708 to be possibly unused. GCC does not produce a warning for this
5712 @cindex @code{used} variable attribute
5713 This attribute, attached to a variable with static storage, means that
5714 the variable must be emitted even if it appears that the variable is not
5717 When applied to a static data member of a C++ class template, the
5718 attribute also means that the member is instantiated if the
5719 class itself is instantiated.
5721 @item vector_size (@var{bytes})
5722 @cindex @code{vector_size} variable attribute
5723 This attribute specifies the vector size for the variable, measured in
5724 bytes. For example, the declaration:
5727 int foo __attribute__ ((vector_size (16)));
5731 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5732 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5733 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5735 This attribute is only applicable to integral and float scalars,
5736 although arrays, pointers, and function return values are allowed in
5737 conjunction with this construct.
5739 Aggregates with this attribute are invalid, even if they are of the same
5740 size as a corresponding scalar. For example, the declaration:
5743 struct S @{ int a; @};
5744 struct S __attribute__ ((vector_size (16))) foo;
5748 is invalid even if the size of the structure is the same as the size of
5751 @item visibility ("@var{visibility_type}")
5752 @cindex @code{visibility} variable attribute
5753 This attribute affects the linkage of the declaration to which it is attached.
5754 The @code{visibility} attribute is described in
5755 @ref{Common Function Attributes}.
5758 @cindex @code{weak} variable attribute
5759 The @code{weak} attribute is described in
5760 @ref{Common Function Attributes}.
5764 @node AVR Variable Attributes
5765 @subsection AVR Variable Attributes
5769 @cindex @code{progmem} variable attribute, AVR
5770 The @code{progmem} attribute is used on the AVR to place read-only
5771 data in the non-volatile program memory (flash). The @code{progmem}
5772 attribute accomplishes this by putting respective variables into a
5773 section whose name starts with @code{.progmem}.
5775 This attribute works similar to the @code{section} attribute
5776 but adds additional checking. Notice that just like the
5777 @code{section} attribute, @code{progmem} affects the location
5778 of the data but not how this data is accessed.
5780 In order to read data located with the @code{progmem} attribute
5781 (inline) assembler must be used.
5783 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5784 #include <avr/pgmspace.h>
5786 /* Locate var in flash memory */
5787 const int var[2] PROGMEM = @{ 1, 2 @};
5789 int read_var (int i)
5791 /* Access var[] by accessor macro from avr/pgmspace.h */
5792 return (int) pgm_read_word (& var[i]);
5796 AVR is a Harvard architecture processor and data and read-only data
5797 normally resides in the data memory (RAM).
5799 See also the @ref{AVR Named Address Spaces} section for
5800 an alternate way to locate and access data in flash memory.
5803 @itemx io (@var{addr})
5804 @cindex @code{io} variable attribute, AVR
5805 Variables with the @code{io} attribute are used to address
5806 memory-mapped peripherals in the io address range.
5807 If an address is specified, the variable
5808 is assigned that address, and the value is interpreted as an
5809 address in the data address space.
5813 volatile int porta __attribute__((io (0x22)));
5816 The address specified in the address in the data address range.
5818 Otherwise, the variable it is not assigned an address, but the
5819 compiler will still use in/out instructions where applicable,
5820 assuming some other module assigns an address in the io address range.
5824 extern volatile int porta __attribute__((io));
5828 @itemx io_low (@var{addr})
5829 @cindex @code{io_low} variable attribute, AVR
5830 This is like the @code{io} attribute, but additionally it informs the
5831 compiler that the object lies in the lower half of the I/O area,
5832 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5836 @itemx address (@var{addr})
5837 @cindex @code{address} variable attribute, AVR
5838 Variables with the @code{address} attribute are used to address
5839 memory-mapped peripherals that may lie outside the io address range.
5842 volatile int porta __attribute__((address (0x600)));
5847 @node Blackfin Variable Attributes
5848 @subsection Blackfin Variable Attributes
5850 Three attributes are currently defined for the Blackfin.
5856 @cindex @code{l1_data} variable attribute, Blackfin
5857 @cindex @code{l1_data_A} variable attribute, Blackfin
5858 @cindex @code{l1_data_B} variable attribute, Blackfin
5859 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5860 Variables with @code{l1_data} attribute are put into the specific section
5861 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5862 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5863 attribute are put into the specific section named @code{.l1.data.B}.
5866 @cindex @code{l2} variable attribute, Blackfin
5867 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5868 Variables with @code{l2} attribute are put into the specific section
5869 named @code{.l2.data}.
5872 @node H8/300 Variable Attributes
5873 @subsection H8/300 Variable Attributes
5875 These variable attributes are available for H8/300 targets:
5879 @cindex @code{eightbit_data} variable attribute, H8/300
5880 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5881 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5882 variable should be placed into the eight-bit data section.
5883 The compiler generates more efficient code for certain operations
5884 on data in the eight-bit data area. Note the eight-bit data area is limited to
5887 You must use GAS and GLD from GNU binutils version 2.7 or later for
5888 this attribute to work correctly.
5891 @cindex @code{tiny_data} variable attribute, H8/300
5892 @cindex tiny data section on the H8/300H and H8S
5893 Use this attribute on the H8/300H and H8S to indicate that the specified
5894 variable should be placed into the tiny data section.
5895 The compiler generates more efficient code for loads and stores
5896 on data in the tiny data section. Note the tiny data area is limited to
5897 slightly under 32KB of data.
5901 @node IA-64 Variable Attributes
5902 @subsection IA-64 Variable Attributes
5904 The IA-64 back end supports the following variable attribute:
5907 @item model (@var{model-name})
5908 @cindex @code{model} variable attribute, IA-64
5910 On IA-64, use this attribute to set the addressability of an object.
5911 At present, the only supported identifier for @var{model-name} is
5912 @code{small}, indicating addressability via ``small'' (22-bit)
5913 addresses (so that their addresses can be loaded with the @code{addl}
5914 instruction). Caveat: such addressing is by definition not position
5915 independent and hence this attribute must not be used for objects
5916 defined by shared libraries.
5920 @node M32R/D Variable Attributes
5921 @subsection M32R/D Variable Attributes
5923 One attribute is currently defined for the M32R/D@.
5926 @item model (@var{model-name})
5927 @cindex @code{model-name} variable attribute, M32R/D
5928 @cindex variable addressability on the M32R/D
5929 Use this attribute on the M32R/D to set the addressability of an object.
5930 The identifier @var{model-name} is one of @code{small}, @code{medium},
5931 or @code{large}, representing each of the code models.
5933 Small model objects live in the lower 16MB of memory (so that their
5934 addresses can be loaded with the @code{ld24} instruction).
5936 Medium and large model objects may live anywhere in the 32-bit address space
5937 (the compiler generates @code{seth/add3} instructions to load their
5941 @node MeP Variable Attributes
5942 @subsection MeP Variable Attributes
5944 The MeP target has a number of addressing modes and busses. The
5945 @code{near} space spans the standard memory space's first 16 megabytes
5946 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5947 The @code{based} space is a 128-byte region in the memory space that
5948 is addressed relative to the @code{$tp} register. The @code{tiny}
5949 space is a 65536-byte region relative to the @code{$gp} register. In
5950 addition to these memory regions, the MeP target has a separate 16-bit
5951 control bus which is specified with @code{cb} attributes.
5956 @cindex @code{based} variable attribute, MeP
5957 Any variable with the @code{based} attribute is assigned to the
5958 @code{.based} section, and is accessed with relative to the
5959 @code{$tp} register.
5962 @cindex @code{tiny} variable attribute, MeP
5963 Likewise, the @code{tiny} attribute assigned variables to the
5964 @code{.tiny} section, relative to the @code{$gp} register.
5967 @cindex @code{near} variable attribute, MeP
5968 Variables with the @code{near} attribute are assumed to have addresses
5969 that fit in a 24-bit addressing mode. This is the default for large
5970 variables (@code{-mtiny=4} is the default) but this attribute can
5971 override @code{-mtiny=} for small variables, or override @code{-ml}.
5974 @cindex @code{far} variable attribute, MeP
5975 Variables with the @code{far} attribute are addressed using a full
5976 32-bit address. Since this covers the entire memory space, this
5977 allows modules to make no assumptions about where variables might be
5981 @cindex @code{io} variable attribute, MeP
5982 @itemx io (@var{addr})
5983 Variables with the @code{io} attribute are used to address
5984 memory-mapped peripherals. If an address is specified, the variable
5985 is assigned that address, else it is not assigned an address (it is
5986 assumed some other module assigns an address). Example:
5989 int timer_count __attribute__((io(0x123)));
5993 @itemx cb (@var{addr})
5994 @cindex @code{cb} variable attribute, MeP
5995 Variables with the @code{cb} attribute are used to access the control
5996 bus, using special instructions. @code{addr} indicates the control bus
6000 int cpu_clock __attribute__((cb(0x123)));
6005 @node Microsoft Windows Variable Attributes
6006 @subsection Microsoft Windows Variable Attributes
6008 You can use these attributes on Microsoft Windows targets.
6009 @ref{x86 Variable Attributes} for additional Windows compatibility
6010 attributes available on all x86 targets.
6015 @cindex @code{dllimport} variable attribute
6016 @cindex @code{dllexport} variable attribute
6017 The @code{dllimport} and @code{dllexport} attributes are described in
6018 @ref{Microsoft Windows Function Attributes}.
6021 @cindex @code{selectany} variable attribute
6022 The @code{selectany} attribute causes an initialized global variable to
6023 have link-once semantics. When multiple definitions of the variable are
6024 encountered by the linker, the first is selected and the remainder are
6025 discarded. Following usage by the Microsoft compiler, the linker is told
6026 @emph{not} to warn about size or content differences of the multiple
6029 Although the primary usage of this attribute is for POD types, the
6030 attribute can also be applied to global C++ objects that are initialized
6031 by a constructor. In this case, the static initialization and destruction
6032 code for the object is emitted in each translation defining the object,
6033 but the calls to the constructor and destructor are protected by a
6034 link-once guard variable.
6036 The @code{selectany} attribute is only available on Microsoft Windows
6037 targets. You can use @code{__declspec (selectany)} as a synonym for
6038 @code{__attribute__ ((selectany))} for compatibility with other
6042 @cindex @code{shared} variable attribute
6043 On Microsoft Windows, in addition to putting variable definitions in a named
6044 section, the section can also be shared among all running copies of an
6045 executable or DLL@. For example, this small program defines shared data
6046 by putting it in a named section @code{shared} and marking the section
6050 int foo __attribute__((section ("shared"), shared)) = 0;
6055 /* @r{Read and write foo. All running
6056 copies see the same value.} */
6062 You may only use the @code{shared} attribute along with @code{section}
6063 attribute with a fully-initialized global definition because of the way
6064 linkers work. See @code{section} attribute for more information.
6066 The @code{shared} attribute is only available on Microsoft Windows@.
6070 @node MSP430 Variable Attributes
6071 @subsection MSP430 Variable Attributes
6075 @cindex @code{noinit} variable attribute, MSP430
6076 Any data with the @code{noinit} attribute will not be initialised by
6077 the C runtime startup code, or the program loader. Not initialising
6078 data in this way can reduce program startup times.
6081 @cindex @code{persistent} variable attribute, MSP430
6082 Any variable with the @code{persistent} attribute will not be
6083 initialised by the C runtime startup code. Instead its value will be
6084 set once, when the application is loaded, and then never initialised
6085 again, even if the processor is reset or the program restarts.
6086 Persistent data is intended to be placed into FLASH RAM, where its
6087 value will be retained across resets. The linker script being used to
6088 create the application should ensure that persistent data is correctly
6094 @cindex @code{lower} variable attribute, MSP430
6095 @cindex @code{upper} variable attribute, MSP430
6096 @cindex @code{either} variable attribute, MSP430
6097 These attributes are the same as the MSP430 function attributes of the
6098 same name (@pxref{MSP430 Function Attributes}).
6099 These attributes can be applied to both functions and variables.
6102 @node PowerPC Variable Attributes
6103 @subsection PowerPC Variable Attributes
6105 Three attributes currently are defined for PowerPC configurations:
6106 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6108 @cindex @code{ms_struct} variable attribute, PowerPC
6109 @cindex @code{gcc_struct} variable attribute, PowerPC
6110 For full documentation of the struct attributes please see the
6111 documentation in @ref{x86 Variable Attributes}.
6113 @cindex @code{altivec} variable attribute, PowerPC
6114 For documentation of @code{altivec} attribute please see the
6115 documentation in @ref{PowerPC Type Attributes}.
6117 @node RL78 Variable Attributes
6118 @subsection RL78 Variable Attributes
6120 @cindex @code{saddr} variable attribute, RL78
6121 The RL78 back end supports the @code{saddr} variable attribute. This
6122 specifies placement of the corresponding variable in the SADDR area,
6123 which can be accessed more efficiently than the default memory region.
6125 @node SPU Variable Attributes
6126 @subsection SPU Variable Attributes
6128 @cindex @code{spu_vector} variable attribute, SPU
6129 The SPU supports the @code{spu_vector} attribute for variables. For
6130 documentation of this attribute please see the documentation in
6131 @ref{SPU Type Attributes}.
6133 @node V850 Variable Attributes
6134 @subsection V850 Variable Attributes
6136 These variable attributes are supported by the V850 back end:
6141 @cindex @code{sda} variable attribute, V850
6142 Use this attribute to explicitly place a variable in the small data area,
6143 which can hold up to 64 kilobytes.
6146 @cindex @code{tda} variable attribute, V850
6147 Use this attribute to explicitly place a variable in the tiny data area,
6148 which can hold up to 256 bytes in total.
6151 @cindex @code{zda} variable attribute, V850
6152 Use this attribute to explicitly place a variable in the first 32 kilobytes
6156 @node x86 Variable Attributes
6157 @subsection x86 Variable Attributes
6159 Two attributes are currently defined for x86 configurations:
6160 @code{ms_struct} and @code{gcc_struct}.
6165 @cindex @code{ms_struct} variable attribute, x86
6166 @cindex @code{gcc_struct} variable attribute, x86
6168 If @code{packed} is used on a structure, or if bit-fields are used,
6169 it may be that the Microsoft ABI lays out the structure differently
6170 than the way GCC normally does. Particularly when moving packed
6171 data between functions compiled with GCC and the native Microsoft compiler
6172 (either via function call or as data in a file), it may be necessary to access
6175 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6176 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6177 command-line options, respectively;
6178 see @ref{x86 Options}, for details of how structure layout is affected.
6179 @xref{x86 Type Attributes}, for information about the corresponding
6180 attributes on types.
6184 @node Xstormy16 Variable Attributes
6185 @subsection Xstormy16 Variable Attributes
6187 One attribute is currently defined for xstormy16 configurations:
6192 @cindex @code{below100} variable attribute, Xstormy16
6194 If a variable has the @code{below100} attribute (@code{BELOW100} is
6195 allowed also), GCC places the variable in the first 0x100 bytes of
6196 memory and use special opcodes to access it. Such variables are
6197 placed in either the @code{.bss_below100} section or the
6198 @code{.data_below100} section.
6202 @node Type Attributes
6203 @section Specifying Attributes of Types
6204 @cindex attribute of types
6205 @cindex type attributes
6207 The keyword @code{__attribute__} allows you to specify special
6208 attributes of types. Some type attributes apply only to @code{struct}
6209 and @code{union} types, while others can apply to any type defined
6210 via a @code{typedef} declaration. Other attributes are defined for
6211 functions (@pxref{Function Attributes}), labels (@pxref{Label
6212 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6213 variables (@pxref{Variable Attributes}).
6215 The @code{__attribute__} keyword is followed by an attribute specification
6216 inside double parentheses.
6218 You may specify type attributes in an enum, struct or union type
6219 declaration or definition by placing them immediately after the
6220 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6221 syntax is to place them just past the closing curly brace of the
6224 You can also include type attributes in a @code{typedef} declaration.
6225 @xref{Attribute Syntax}, for details of the exact syntax for using
6229 * Common Type Attributes::
6230 * ARM Type Attributes::
6231 * MeP Type Attributes::
6232 * PowerPC Type Attributes::
6233 * SPU Type Attributes::
6234 * x86 Type Attributes::
6237 @node Common Type Attributes
6238 @subsection Common Type Attributes
6240 The following type attributes are supported on most targets.
6243 @cindex @code{aligned} type attribute
6244 @item aligned (@var{alignment})
6245 This attribute specifies a minimum alignment (in bytes) for variables
6246 of the specified type. For example, the declarations:
6249 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6250 typedef int more_aligned_int __attribute__ ((aligned (8)));
6254 force the compiler to ensure (as far as it can) that each variable whose
6255 type is @code{struct S} or @code{more_aligned_int} is allocated and
6256 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6257 variables of type @code{struct S} aligned to 8-byte boundaries allows
6258 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6259 store) instructions when copying one variable of type @code{struct S} to
6260 another, thus improving run-time efficiency.
6262 Note that the alignment of any given @code{struct} or @code{union} type
6263 is required by the ISO C standard to be at least a perfect multiple of
6264 the lowest common multiple of the alignments of all of the members of
6265 the @code{struct} or @code{union} in question. This means that you @emph{can}
6266 effectively adjust the alignment of a @code{struct} or @code{union}
6267 type by attaching an @code{aligned} attribute to any one of the members
6268 of such a type, but the notation illustrated in the example above is a
6269 more obvious, intuitive, and readable way to request the compiler to
6270 adjust the alignment of an entire @code{struct} or @code{union} type.
6272 As in the preceding example, you can explicitly specify the alignment
6273 (in bytes) that you wish the compiler to use for a given @code{struct}
6274 or @code{union} type. Alternatively, you can leave out the alignment factor
6275 and just ask the compiler to align a type to the maximum
6276 useful alignment for the target machine you are compiling for. For
6277 example, you could write:
6280 struct S @{ short f[3]; @} __attribute__ ((aligned));
6283 Whenever you leave out the alignment factor in an @code{aligned}
6284 attribute specification, the compiler automatically sets the alignment
6285 for the type to the largest alignment that is ever used for any data
6286 type on the target machine you are compiling for. Doing this can often
6287 make copy operations more efficient, because the compiler can use
6288 whatever instructions copy the biggest chunks of memory when performing
6289 copies to or from the variables that have types that you have aligned
6292 In the example above, if the size of each @code{short} is 2 bytes, then
6293 the size of the entire @code{struct S} type is 6 bytes. The smallest
6294 power of two that is greater than or equal to that is 8, so the
6295 compiler sets the alignment for the entire @code{struct S} type to 8
6298 Note that although you can ask the compiler to select a time-efficient
6299 alignment for a given type and then declare only individual stand-alone
6300 objects of that type, the compiler's ability to select a time-efficient
6301 alignment is primarily useful only when you plan to create arrays of
6302 variables having the relevant (efficiently aligned) type. If you
6303 declare or use arrays of variables of an efficiently-aligned type, then
6304 it is likely that your program also does pointer arithmetic (or
6305 subscripting, which amounts to the same thing) on pointers to the
6306 relevant type, and the code that the compiler generates for these
6307 pointer arithmetic operations is often more efficient for
6308 efficiently-aligned types than for other types.
6310 Note that the effectiveness of @code{aligned} attributes may be limited
6311 by inherent limitations in your linker. On many systems, the linker is
6312 only able to arrange for variables to be aligned up to a certain maximum
6313 alignment. (For some linkers, the maximum supported alignment may
6314 be very very small.) If your linker is only able to align variables
6315 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6316 in an @code{__attribute__} still only provides you with 8-byte
6317 alignment. See your linker documentation for further information.
6319 The @code{aligned} attribute can only increase alignment. Alignment
6320 can be decreased by specifying the @code{packed} attribute. See below.
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 @opindex fshort-enums
6449 Specifying the @code{packed} attribute for @code{struct} and @code{union}
6450 types is equivalent to specifying the @code{packed} attribute on each
6451 of the structure or union members. Specifying the @option{-fshort-enums}
6452 flag on the command line is equivalent to specifying the @code{packed}
6453 attribute on all @code{enum} definitions.
6455 In the following example @code{struct my_packed_struct}'s members are
6456 packed closely together, but the internal layout of its @code{s} member
6457 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6461 struct my_unpacked_struct
6467 struct __attribute__ ((__packed__)) my_packed_struct
6471 struct my_unpacked_struct s;
6475 You may only specify the @code{packed} attribute attribute on the definition
6476 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
6477 that does not also define the enumerated type, structure or union.
6479 @item scalar_storage_order ("@var{endianness}")
6480 @cindex @code{scalar_storage_order} type attribute
6481 When attached to a @code{union} or a @code{struct}, this attribute sets
6482 the storage order, aka endianness, of the scalar fields of the type, as
6483 well as the array fields whose component is scalar. The supported
6484 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
6485 has no effects on fields which are themselves a @code{union}, a @code{struct}
6486 or an array whose component is a @code{union} or a @code{struct}, and it is
6487 possible for these fields to have a different scalar storage order than the
6490 This attribute is supported only for targets that use a uniform default
6491 scalar storage order (fortunately, most of them), i.e. targets that store
6492 the scalars either all in big-endian or all in little-endian.
6494 Additional restrictions are enforced for types with the reverse scalar
6495 storage order with regard to the scalar storage order of the target:
6498 @item Taking the address of a scalar field of a @code{union} or a
6499 @code{struct} with reverse scalar storage order is not permitted and yields
6501 @item Taking the address of an array field, whose component is scalar, of
6502 a @code{union} or a @code{struct} with reverse scalar storage order is
6503 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
6505 @item Taking the address of a @code{union} or a @code{struct} with reverse
6506 scalar storage order is permitted.
6509 These restrictions exist because the storage order attribute is lost when
6510 the address of a scalar or the address of an array with scalar component is
6511 taken, so storing indirectly through this address generally does not work.
6512 The second case is nevertheless allowed to be able to perform a block copy
6513 from or to the array.
6515 Moreover, the use of type punning or aliasing to toggle the storage order
6516 is not supported; that is to say, a given scalar object cannot be accessed
6517 through distinct types that assign a different storage order to it.
6519 @item transparent_union
6520 @cindex @code{transparent_union} type attribute
6522 This attribute, attached to a @code{union} type definition, indicates
6523 that any function parameter having that union type causes calls to that
6524 function to be treated in a special way.
6526 First, the argument corresponding to a transparent union type can be of
6527 any type in the union; no cast is required. Also, if the union contains
6528 a pointer type, the corresponding argument can be a null pointer
6529 constant or a void pointer expression; and if the union contains a void
6530 pointer type, the corresponding argument can be any pointer expression.
6531 If the union member type is a pointer, qualifiers like @code{const} on
6532 the referenced type must be respected, just as with normal pointer
6535 Second, the argument is passed to the function using the calling
6536 conventions of the first member of the transparent union, not the calling
6537 conventions of the union itself. All members of the union must have the
6538 same machine representation; this is necessary for this argument passing
6541 Transparent unions are designed for library functions that have multiple
6542 interfaces for compatibility reasons. For example, suppose the
6543 @code{wait} function must accept either a value of type @code{int *} to
6544 comply with POSIX, or a value of type @code{union wait *} to comply with
6545 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6546 @code{wait} would accept both kinds of arguments, but it would also
6547 accept any other pointer type and this would make argument type checking
6548 less useful. Instead, @code{<sys/wait.h>} might define the interface
6552 typedef union __attribute__ ((__transparent_union__))
6556 @} wait_status_ptr_t;
6558 pid_t wait (wait_status_ptr_t);
6562 This interface allows either @code{int *} or @code{union wait *}
6563 arguments to be passed, using the @code{int *} calling convention.
6564 The program can call @code{wait} with arguments of either type:
6567 int w1 () @{ int w; return wait (&w); @}
6568 int w2 () @{ union wait w; return wait (&w); @}
6572 With this interface, @code{wait}'s implementation might look like this:
6575 pid_t wait (wait_status_ptr_t p)
6577 return waitpid (-1, p.__ip, 0);
6582 @cindex @code{unused} type attribute
6583 When attached to a type (including a @code{union} or a @code{struct}),
6584 this attribute means that variables of that type are meant to appear
6585 possibly unused. GCC does not produce a warning for any variables of
6586 that type, even if the variable appears to do nothing. This is often
6587 the case with lock or thread classes, which are usually defined and then
6588 not referenced, but contain constructors and destructors that have
6589 nontrivial bookkeeping functions.
6592 @cindex @code{visibility} type attribute
6593 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6594 applied to class, struct, union and enum types. Unlike other type
6595 attributes, the attribute must appear between the initial keyword and
6596 the name of the type; it cannot appear after the body of the type.
6598 Note that the type visibility is applied to vague linkage entities
6599 associated with the class (vtable, typeinfo node, etc.). In
6600 particular, if a class is thrown as an exception in one shared object
6601 and caught in another, the class must have default visibility.
6602 Otherwise the two shared objects are unable to use the same
6603 typeinfo node and exception handling will break.
6607 To specify multiple attributes, separate them by commas within the
6608 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6611 @node ARM Type Attributes
6612 @subsection ARM Type Attributes
6614 @cindex @code{notshared} type attribute, ARM
6615 On those ARM targets that support @code{dllimport} (such as Symbian
6616 OS), you can use the @code{notshared} attribute to indicate that the
6617 virtual table and other similar data for a class should not be
6618 exported from a DLL@. For example:
6621 class __declspec(notshared) C @{
6623 __declspec(dllimport) C();
6627 __declspec(dllexport)
6632 In this code, @code{C::C} is exported from the current DLL, but the
6633 virtual table for @code{C} is not exported. (You can use
6634 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6635 most Symbian OS code uses @code{__declspec}.)
6637 @node MeP Type Attributes
6638 @subsection MeP Type Attributes
6640 @cindex @code{based} type attribute, MeP
6641 @cindex @code{tiny} type attribute, MeP
6642 @cindex @code{near} type attribute, MeP
6643 @cindex @code{far} type attribute, MeP
6644 Many of the MeP variable attributes may be applied to types as well.
6645 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6646 @code{far} attributes may be applied to either. The @code{io} and
6647 @code{cb} attributes may not be applied to types.
6649 @node PowerPC Type Attributes
6650 @subsection PowerPC Type Attributes
6652 Three attributes currently are defined for PowerPC configurations:
6653 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6655 @cindex @code{ms_struct} type attribute, PowerPC
6656 @cindex @code{gcc_struct} type attribute, PowerPC
6657 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6658 attributes please see the documentation in @ref{x86 Type Attributes}.
6660 @cindex @code{altivec} type attribute, PowerPC
6661 The @code{altivec} attribute allows one to declare AltiVec vector data
6662 types supported by the AltiVec Programming Interface Manual. The
6663 attribute requires an argument to specify one of three vector types:
6664 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6665 and @code{bool__} (always followed by unsigned).
6668 __attribute__((altivec(vector__)))
6669 __attribute__((altivec(pixel__))) unsigned short
6670 __attribute__((altivec(bool__))) unsigned
6673 These attributes mainly are intended to support the @code{__vector},
6674 @code{__pixel}, and @code{__bool} AltiVec keywords.
6676 @node SPU Type Attributes
6677 @subsection SPU Type Attributes
6679 @cindex @code{spu_vector} type attribute, SPU
6680 The SPU supports the @code{spu_vector} attribute for types. This attribute
6681 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6682 Language Extensions Specification. It is intended to support the
6683 @code{__vector} keyword.
6685 @node x86 Type Attributes
6686 @subsection x86 Type Attributes
6688 Two attributes are currently defined for x86 configurations:
6689 @code{ms_struct} and @code{gcc_struct}.
6695 @cindex @code{ms_struct} type attribute, x86
6696 @cindex @code{gcc_struct} type attribute, x86
6698 If @code{packed} is used on a structure, or if bit-fields are used
6699 it may be that the Microsoft ABI packs them differently
6700 than GCC normally packs them. Particularly when moving packed
6701 data between functions compiled with GCC and the native Microsoft compiler
6702 (either via function call or as data in a file), it may be necessary to access
6705 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6706 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6707 command-line options, respectively;
6708 see @ref{x86 Options}, for details of how structure layout is affected.
6709 @xref{x86 Variable Attributes}, for information about the corresponding
6710 attributes on variables.
6714 @node Label Attributes
6715 @section Label Attributes
6716 @cindex Label Attributes
6718 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6719 details of the exact syntax for using attributes. Other attributes are
6720 available for functions (@pxref{Function Attributes}), variables
6721 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6722 and for types (@pxref{Type Attributes}).
6724 This example uses the @code{cold} label attribute to indicate the
6725 @code{ErrorHandling} branch is unlikely to be taken and that the
6726 @code{ErrorHandling} label is unused:
6730 asm goto ("some asm" : : : : NoError);
6732 /* This branch (the fall-through from the asm) is less commonly used */
6734 __attribute__((cold, unused)); /* Semi-colon is required here */
6739 printf("no error\n");
6745 @cindex @code{unused} label attribute
6746 This feature is intended for program-generated code that may contain
6747 unused labels, but which is compiled with @option{-Wall}. It is
6748 not normally appropriate to use in it human-written code, though it
6749 could be useful in cases where the code that jumps to the label is
6750 contained within an @code{#ifdef} conditional.
6753 @cindex @code{hot} label attribute
6754 The @code{hot} attribute on a label is used to inform the compiler that
6755 the path following the label is more likely than paths that are not so
6756 annotated. This attribute is used in cases where @code{__builtin_expect}
6757 cannot be used, for instance with computed goto or @code{asm goto}.
6760 @cindex @code{cold} label attribute
6761 The @code{cold} attribute on labels is used to inform the compiler that
6762 the path following the label is unlikely to be executed. This attribute
6763 is used in cases where @code{__builtin_expect} cannot be used, for instance
6764 with computed goto or @code{asm goto}.
6768 @node Enumerator Attributes
6769 @section Enumerator Attributes
6770 @cindex Enumerator Attributes
6772 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6773 details of the exact syntax for using attributes. Other attributes are
6774 available for functions (@pxref{Function Attributes}), variables
6775 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6776 and for types (@pxref{Type Attributes}).
6778 This example uses the @code{deprecated} enumerator attribute to indicate the
6779 @code{oldval} enumerator is deprecated:
6783 oldval __attribute__((deprecated)),
6796 @cindex @code{deprecated} enumerator attribute
6797 The @code{deprecated} attribute results in a warning if the enumerator
6798 is used anywhere in the source file. This is useful when identifying
6799 enumerators that are expected to be removed in a future version of a
6800 program. The warning also includes the location of the declaration
6801 of the deprecated enumerator, to enable users to easily find further
6802 information about why the enumerator is deprecated, or what they should
6803 do instead. Note that the warnings only occurs for uses.
6807 @node Attribute Syntax
6808 @section Attribute Syntax
6809 @cindex attribute syntax
6811 This section describes the syntax with which @code{__attribute__} may be
6812 used, and the constructs to which attribute specifiers bind, for the C
6813 language. Some details may vary for C++ and Objective-C@. Because of
6814 infelicities in the grammar for attributes, some forms described here
6815 may not be successfully parsed in all cases.
6817 There are some problems with the semantics of attributes in C++. For
6818 example, there are no manglings for attributes, although they may affect
6819 code generation, so problems may arise when attributed types are used in
6820 conjunction with templates or overloading. Similarly, @code{typeid}
6821 does not distinguish between types with different attributes. Support
6822 for attributes in C++ may be restricted in future to attributes on
6823 declarations only, but not on nested declarators.
6825 @xref{Function Attributes}, for details of the semantics of attributes
6826 applying to functions. @xref{Variable Attributes}, for details of the
6827 semantics of attributes applying to variables. @xref{Type Attributes},
6828 for details of the semantics of attributes applying to structure, union
6829 and enumerated types.
6830 @xref{Label Attributes}, for details of the semantics of attributes
6832 @xref{Enumerator Attributes}, for details of the semantics of attributes
6833 applying to enumerators.
6835 An @dfn{attribute specifier} is of the form
6836 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6837 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6838 each attribute is one of the following:
6842 Empty. Empty attributes are ignored.
6846 (which may be an identifier such as @code{unused}, or a reserved
6847 word such as @code{const}).
6850 An attribute name followed by a parenthesized list of
6851 parameters for the attribute.
6852 These parameters take one of the following forms:
6856 An identifier. For example, @code{mode} attributes use this form.
6859 An identifier followed by a comma and a non-empty comma-separated list
6860 of expressions. For example, @code{format} attributes use this form.
6863 A possibly empty comma-separated list of expressions. For example,
6864 @code{format_arg} attributes use this form with the list being a single
6865 integer constant expression, and @code{alias} attributes use this form
6866 with the list being a single string constant.
6870 An @dfn{attribute specifier list} is a sequence of one or more attribute
6871 specifiers, not separated by any other tokens.
6873 You may optionally specify attribute names with @samp{__}
6874 preceding and following the name.
6875 This allows you to use them in header files without
6876 being concerned about a possible macro of the same name. For example,
6877 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6880 @subsubheading Label Attributes
6882 In GNU C, an attribute specifier list may appear after the colon following a
6883 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6884 attributes on labels if the attribute specifier is immediately
6885 followed by a semicolon (i.e., the label applies to an empty
6886 statement). If the semicolon is missing, C++ label attributes are
6887 ambiguous, as it is permissible for a declaration, which could begin
6888 with an attribute list, to be labelled in C++. Declarations cannot be
6889 labelled in C90 or C99, so the ambiguity does not arise there.
6891 @subsubheading Enumerator Attributes
6893 In GNU C, an attribute specifier list may appear as part of an enumerator.
6894 The attribute goes after the enumeration constant, before @code{=}, if
6895 present. The optional attribute in the enumerator appertains to the
6896 enumeration constant. It is not possible to place the attribute after
6897 the constant expression, if present.
6899 @subsubheading Type Attributes
6901 An attribute specifier list may appear as part of a @code{struct},
6902 @code{union} or @code{enum} specifier. It may go either immediately
6903 after the @code{struct}, @code{union} or @code{enum} keyword, or after
6904 the closing brace. The former syntax is preferred.
6905 Where attribute specifiers follow the closing brace, they are considered
6906 to relate to the structure, union or enumerated type defined, not to any
6907 enclosing declaration the type specifier appears in, and the type
6908 defined is not complete until after the attribute specifiers.
6909 @c Otherwise, there would be the following problems: a shift/reduce
6910 @c conflict between attributes binding the struct/union/enum and
6911 @c binding to the list of specifiers/qualifiers; and "aligned"
6912 @c attributes could use sizeof for the structure, but the size could be
6913 @c changed later by "packed" attributes.
6916 @subsubheading All other attributes
6918 Otherwise, an attribute specifier appears as part of a declaration,
6919 counting declarations of unnamed parameters and type names, and relates
6920 to that declaration (which may be nested in another declaration, for
6921 example in the case of a parameter declaration), or to a particular declarator
6922 within a declaration. Where an
6923 attribute specifier is applied to a parameter declared as a function or
6924 an array, it should apply to the function or array rather than the
6925 pointer to which the parameter is implicitly converted, but this is not
6926 yet correctly implemented.
6928 Any list of specifiers and qualifiers at the start of a declaration may
6929 contain attribute specifiers, whether or not such a list may in that
6930 context contain storage class specifiers. (Some attributes, however,
6931 are essentially in the nature of storage class specifiers, and only make
6932 sense where storage class specifiers may be used; for example,
6933 @code{section}.) There is one necessary limitation to this syntax: the
6934 first old-style parameter declaration in a function definition cannot
6935 begin with an attribute specifier, because such an attribute applies to
6936 the function instead by syntax described below (which, however, is not
6937 yet implemented in this case). In some other cases, attribute
6938 specifiers are permitted by this grammar but not yet supported by the
6939 compiler. All attribute specifiers in this place relate to the
6940 declaration as a whole. In the obsolescent usage where a type of
6941 @code{int} is implied by the absence of type specifiers, such a list of
6942 specifiers and qualifiers may be an attribute specifier list with no
6943 other specifiers or qualifiers.
6945 At present, the first parameter in a function prototype must have some
6946 type specifier that is not an attribute specifier; this resolves an
6947 ambiguity in the interpretation of @code{void f(int
6948 (__attribute__((foo)) x))}, but is subject to change. At present, if
6949 the parentheses of a function declarator contain only attributes then
6950 those attributes are ignored, rather than yielding an error or warning
6951 or implying a single parameter of type int, but this is subject to
6954 An attribute specifier list may appear immediately before a declarator
6955 (other than the first) in a comma-separated list of declarators in a
6956 declaration of more than one identifier using a single list of
6957 specifiers and qualifiers. Such attribute specifiers apply
6958 only to the identifier before whose declarator they appear. For
6962 __attribute__((noreturn)) void d0 (void),
6963 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
6968 the @code{noreturn} attribute applies to all the functions
6969 declared; the @code{format} attribute only applies to @code{d1}.
6971 An attribute specifier list may appear immediately before the comma,
6972 @code{=} or semicolon terminating the declaration of an identifier other
6973 than a function definition. Such attribute specifiers apply
6974 to the declared object or function. Where an
6975 assembler name for an object or function is specified (@pxref{Asm
6976 Labels}), the attribute must follow the @code{asm}
6979 An attribute specifier list may, in future, be permitted to appear after
6980 the declarator in a function definition (before any old-style parameter
6981 declarations or the function body).
6983 Attribute specifiers may be mixed with type qualifiers appearing inside
6984 the @code{[]} of a parameter array declarator, in the C99 construct by
6985 which such qualifiers are applied to the pointer to which the array is
6986 implicitly converted. Such attribute specifiers apply to the pointer,
6987 not to the array, but at present this is not implemented and they are
6990 An attribute specifier list may appear at the start of a nested
6991 declarator. At present, there are some limitations in this usage: the
6992 attributes correctly apply to the declarator, but for most individual
6993 attributes the semantics this implies are not implemented.
6994 When attribute specifiers follow the @code{*} of a pointer
6995 declarator, they may be mixed with any type qualifiers present.
6996 The following describes the formal semantics of this syntax. It makes the
6997 most sense if you are familiar with the formal specification of
6998 declarators in the ISO C standard.
7000 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
7001 D1}, where @code{T} contains declaration specifiers that specify a type
7002 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
7003 contains an identifier @var{ident}. The type specified for @var{ident}
7004 for derived declarators whose type does not include an attribute
7005 specifier is as in the ISO C standard.
7007 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7008 and the declaration @code{T D} specifies the type
7009 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7010 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7011 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7013 If @code{D1} has the form @code{*
7014 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7015 declaration @code{T D} specifies the type
7016 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7017 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7018 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7024 void (__attribute__((noreturn)) ****f) (void);
7028 specifies the type ``pointer to pointer to pointer to pointer to
7029 non-returning function returning @code{void}''. As another example,
7032 char *__attribute__((aligned(8))) *f;
7036 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7037 Note again that this does not work with most attributes; for example,
7038 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7039 is not yet supported.
7041 For compatibility with existing code written for compiler versions that
7042 did not implement attributes on nested declarators, some laxity is
7043 allowed in the placing of attributes. If an attribute that only applies
7044 to types is applied to a declaration, it is treated as applying to
7045 the type of that declaration. If an attribute that only applies to
7046 declarations is applied to the type of a declaration, it is treated
7047 as applying to that declaration; and, for compatibility with code
7048 placing the attributes immediately before the identifier declared, such
7049 an attribute applied to a function return type is treated as
7050 applying to the function type, and such an attribute applied to an array
7051 element type is treated as applying to the array type. If an
7052 attribute that only applies to function types is applied to a
7053 pointer-to-function type, it is treated as applying to the pointer
7054 target type; if such an attribute is applied to a function return type
7055 that is not a pointer-to-function type, it is treated as applying
7056 to the function type.
7058 @node Function Prototypes
7059 @section Prototypes and Old-Style Function Definitions
7060 @cindex function prototype declarations
7061 @cindex old-style function definitions
7062 @cindex promotion of formal parameters
7064 GNU C extends ISO C to allow a function prototype to override a later
7065 old-style non-prototype definition. Consider the following example:
7068 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7075 /* @r{Prototype function declaration.} */
7076 int isroot P((uid_t));
7078 /* @r{Old-style function definition.} */
7080 isroot (x) /* @r{??? lossage here ???} */
7087 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7088 not allow this example, because subword arguments in old-style
7089 non-prototype definitions are promoted. Therefore in this example the
7090 function definition's argument is really an @code{int}, which does not
7091 match the prototype argument type of @code{short}.
7093 This restriction of ISO C makes it hard to write code that is portable
7094 to traditional C compilers, because the programmer does not know
7095 whether the @code{uid_t} type is @code{short}, @code{int}, or
7096 @code{long}. Therefore, in cases like these GNU C allows a prototype
7097 to override a later old-style definition. More precisely, in GNU C, a
7098 function prototype argument type overrides the argument type specified
7099 by a later old-style definition if the former type is the same as the
7100 latter type before promotion. Thus in GNU C the above example is
7101 equivalent to the following:
7114 GNU C++ does not support old-style function definitions, so this
7115 extension is irrelevant.
7118 @section C++ Style Comments
7120 @cindex C++ comments
7121 @cindex comments, C++ style
7123 In GNU C, you may use C++ style comments, which start with @samp{//} and
7124 continue until the end of the line. Many other C implementations allow
7125 such comments, and they are included in the 1999 C standard. However,
7126 C++ style comments are not recognized if you specify an @option{-std}
7127 option specifying a version of ISO C before C99, or @option{-ansi}
7128 (equivalent to @option{-std=c90}).
7131 @section Dollar Signs in Identifier Names
7133 @cindex dollar signs in identifier names
7134 @cindex identifier names, dollar signs in
7136 In GNU C, you may normally use dollar signs in identifier names.
7137 This is because many traditional C implementations allow such identifiers.
7138 However, dollar signs in identifiers are not supported on a few target
7139 machines, typically because the target assembler does not allow them.
7141 @node Character Escapes
7142 @section The Character @key{ESC} in Constants
7144 You can use the sequence @samp{\e} in a string or character constant to
7145 stand for the ASCII character @key{ESC}.
7148 @section Inquiring on Alignment of Types or Variables
7150 @cindex type alignment
7151 @cindex variable alignment
7153 The keyword @code{__alignof__} allows you to inquire about how an object
7154 is aligned, or the minimum alignment usually required by a type. Its
7155 syntax is just like @code{sizeof}.
7157 For example, if the target machine requires a @code{double} value to be
7158 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7159 This is true on many RISC machines. On more traditional machine
7160 designs, @code{__alignof__ (double)} is 4 or even 2.
7162 Some machines never actually require alignment; they allow reference to any
7163 data type even at an odd address. For these machines, @code{__alignof__}
7164 reports the smallest alignment that GCC gives the data type, usually as
7165 mandated by the target ABI.
7167 If the operand of @code{__alignof__} is an lvalue rather than a type,
7168 its value is the required alignment for its type, taking into account
7169 any minimum alignment specified with GCC's @code{__attribute__}
7170 extension (@pxref{Variable Attributes}). For example, after this
7174 struct foo @{ int x; char y; @} foo1;
7178 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7179 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7181 It is an error to ask for the alignment of an incomplete type.
7185 @section An Inline Function is As Fast As a Macro
7186 @cindex inline functions
7187 @cindex integrating function code
7189 @cindex macros, inline alternative
7191 By declaring a function inline, you can direct GCC to make
7192 calls to that function faster. One way GCC can achieve this is to
7193 integrate that function's code into the code for its callers. This
7194 makes execution faster by eliminating the function-call overhead; in
7195 addition, if any of the actual argument values are constant, their
7196 known values may permit simplifications at compile time so that not
7197 all of the inline function's code needs to be included. The effect on
7198 code size is less predictable; object code may be larger or smaller
7199 with function inlining, depending on the particular case. You can
7200 also direct GCC to try to integrate all ``simple enough'' functions
7201 into their callers with the option @option{-finline-functions}.
7203 GCC implements three different semantics of declaring a function
7204 inline. One is available with @option{-std=gnu89} or
7205 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7206 on all inline declarations, another when
7207 @option{-std=c99}, @option{-std=c11},
7208 @option{-std=gnu99} or @option{-std=gnu11}
7209 (without @option{-fgnu89-inline}), and the third
7210 is used when compiling C++.
7212 To declare a function inline, use the @code{inline} keyword in its
7213 declaration, like this:
7223 If you are writing a header file to be included in ISO C90 programs, write
7224 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7226 The three types of inlining behave similarly in two important cases:
7227 when the @code{inline} keyword is used on a @code{static} function,
7228 like the example above, and when a function is first declared without
7229 using the @code{inline} keyword and then is defined with
7230 @code{inline}, like this:
7233 extern int inc (int *a);
7241 In both of these common cases, the program behaves the same as if you
7242 had not used the @code{inline} keyword, except for its speed.
7244 @cindex inline functions, omission of
7245 @opindex fkeep-inline-functions
7246 When a function is both inline and @code{static}, if all calls to the
7247 function are integrated into the caller, and the function's address is
7248 never used, then the function's own assembler code is never referenced.
7249 In this case, GCC does not actually output assembler code for the
7250 function, unless you specify the option @option{-fkeep-inline-functions}.
7251 If there is a nonintegrated call, then the function is compiled to
7252 assembler code as usual. The function must also be compiled as usual if
7253 the program refers to its address, because that can't be inlined.
7256 Note that certain usages in a function definition can make it unsuitable
7257 for inline substitution. Among these usages are: variadic functions,
7258 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7259 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7260 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7261 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7262 function marked @code{inline} could not be substituted, and gives the
7263 reason for the failure.
7265 @cindex automatic @code{inline} for C++ member fns
7266 @cindex @code{inline} automatic for C++ member fns
7267 @cindex member fns, automatically @code{inline}
7268 @cindex C++ member fns, automatically @code{inline}
7269 @opindex fno-default-inline
7270 As required by ISO C++, GCC considers member functions defined within
7271 the body of a class to be marked inline even if they are
7272 not explicitly declared with the @code{inline} keyword. You can
7273 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7274 Options,,Options Controlling C++ Dialect}.
7276 GCC does not inline any functions when not optimizing unless you specify
7277 the @samp{always_inline} attribute for the function, like this:
7280 /* @r{Prototype.} */
7281 inline void foo (const char) __attribute__((always_inline));
7284 The remainder of this section is specific to GNU C90 inlining.
7286 @cindex non-static inline function
7287 When an inline function is not @code{static}, then the compiler must assume
7288 that there may be calls from other source files; since a global symbol can
7289 be defined only once in any program, the function must not be defined in
7290 the other source files, so the calls therein cannot be integrated.
7291 Therefore, a non-@code{static} inline function is always compiled on its
7292 own in the usual fashion.
7294 If you specify both @code{inline} and @code{extern} in the function
7295 definition, then the definition is used only for inlining. In no case
7296 is the function compiled on its own, not even if you refer to its
7297 address explicitly. Such an address becomes an external reference, as
7298 if you had only declared the function, and had not defined it.
7300 This combination of @code{inline} and @code{extern} has almost the
7301 effect of a macro. The way to use it is to put a function definition in
7302 a header file with these keywords, and put another copy of the
7303 definition (lacking @code{inline} and @code{extern}) in a library file.
7304 The definition in the header file causes most calls to the function
7305 to be inlined. If any uses of the function remain, they refer to
7306 the single copy in the library.
7309 @section When is a Volatile Object Accessed?
7310 @cindex accessing volatiles
7311 @cindex volatile read
7312 @cindex volatile write
7313 @cindex volatile access
7315 C has the concept of volatile objects. These are normally accessed by
7316 pointers and used for accessing hardware or inter-thread
7317 communication. The standard encourages compilers to refrain from
7318 optimizations concerning accesses to volatile objects, but leaves it
7319 implementation defined as to what constitutes a volatile access. The
7320 minimum requirement is that at a sequence point all previous accesses
7321 to volatile objects have stabilized and no subsequent accesses have
7322 occurred. Thus an implementation is free to reorder and combine
7323 volatile accesses that occur between sequence points, but cannot do
7324 so for accesses across a sequence point. The use of volatile does
7325 not allow you to violate the restriction on updating objects multiple
7326 times between two sequence points.
7328 Accesses to non-volatile objects are not ordered with respect to
7329 volatile accesses. You cannot use a volatile object as a memory
7330 barrier to order a sequence of writes to non-volatile memory. For
7334 int *ptr = @var{something};
7336 *ptr = @var{something};
7341 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7342 that the write to @var{*ptr} occurs by the time the update
7343 of @var{vobj} happens. If you need this guarantee, you must use
7344 a stronger memory barrier such as:
7347 int *ptr = @var{something};
7349 *ptr = @var{something};
7350 asm volatile ("" : : : "memory");
7354 A scalar volatile object is read when it is accessed in a void context:
7357 volatile int *src = @var{somevalue};
7361 Such expressions are rvalues, and GCC implements this as a
7362 read of the volatile object being pointed to.
7364 Assignments are also expressions and have an rvalue. However when
7365 assigning to a scalar volatile, the volatile object is not reread,
7366 regardless of whether the assignment expression's rvalue is used or
7367 not. If the assignment's rvalue is used, the value is that assigned
7368 to the volatile object. For instance, there is no read of @var{vobj}
7369 in all the following cases:
7374 vobj = @var{something};
7375 obj = vobj = @var{something};
7376 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7377 obj = (@var{something}, vobj = @var{anotherthing});
7380 If you need to read the volatile object after an assignment has
7381 occurred, you must use a separate expression with an intervening
7384 As bit-fields are not individually addressable, volatile bit-fields may
7385 be implicitly read when written to, or when adjacent bit-fields are
7386 accessed. Bit-field operations may be optimized such that adjacent
7387 bit-fields are only partially accessed, if they straddle a storage unit
7388 boundary. For these reasons it is unwise to use volatile bit-fields to
7391 @node Using Assembly Language with C
7392 @section How to Use Inline Assembly Language in C Code
7393 @cindex @code{asm} keyword
7394 @cindex assembly language in C
7395 @cindex inline assembly language
7396 @cindex mixing assembly language and C
7398 The @code{asm} keyword allows you to embed assembler instructions
7399 within C code. GCC provides two forms of inline @code{asm}
7400 statements. A @dfn{basic @code{asm}} statement is one with no
7401 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7402 statement (@pxref{Extended Asm}) includes one or more operands.
7403 The extended form is preferred for mixing C and assembly language
7404 within a function, but to include assembly language at
7405 top level you must use basic @code{asm}.
7407 You can also use the @code{asm} keyword to override the assembler name
7408 for a C symbol, or to place a C variable in a specific register.
7411 * Basic Asm:: Inline assembler without operands.
7412 * Extended Asm:: Inline assembler with operands.
7413 * Constraints:: Constraints for @code{asm} operands
7414 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7415 * Explicit Register Variables:: Defining variables residing in specified
7417 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7421 @subsection Basic Asm --- Assembler Instructions Without Operands
7422 @cindex basic @code{asm}
7423 @cindex assembly language in C, basic
7425 A basic @code{asm} statement has the following syntax:
7428 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7431 The @code{asm} keyword is a GNU extension.
7432 When writing code that can be compiled with @option{-ansi} and the
7433 various @option{-std} options, use @code{__asm__} instead of
7434 @code{asm} (@pxref{Alternate Keywords}).
7436 @subsubheading Qualifiers
7439 The optional @code{volatile} qualifier has no effect.
7440 All basic @code{asm} blocks are implicitly volatile.
7443 @subsubheading Parameters
7446 @item AssemblerInstructions
7447 This is a literal string that specifies the assembler code. The string can
7448 contain any instructions recognized by the assembler, including directives.
7449 GCC does not parse the assembler instructions themselves and
7450 does not know what they mean or even whether they are valid assembler input.
7452 You may place multiple assembler instructions together in a single @code{asm}
7453 string, separated by the characters normally used in assembly code for the
7454 system. A combination that works in most places is a newline to break the
7455 line, plus a tab character (written as @samp{\n\t}).
7456 Some assemblers allow semicolons as a line separator. However,
7457 note that some assembler dialects use semicolons to start a comment.
7460 @subsubheading Remarks
7461 Using extended @code{asm} typically produces smaller, safer, and more
7462 efficient code, and in most cases it is a better solution than basic
7463 @code{asm}. However, there are two situations where only basic @code{asm}
7468 Extended @code{asm} statements have to be inside a C
7469 function, so to write inline assembly language at file scope (``top-level''),
7470 outside of C functions, you must use basic @code{asm}.
7471 You can use this technique to emit assembler directives,
7472 define assembly language macros that can be invoked elsewhere in the file,
7473 or write entire functions in assembly language.
7477 with the @code{naked} attribute also require basic @code{asm}
7478 (@pxref{Function Attributes}).
7481 Safely accessing C data and calling functions from basic @code{asm} is more
7482 complex than it may appear. To access C data, it is better to use extended
7485 Do not expect a sequence of @code{asm} statements to remain perfectly
7486 consecutive after compilation. If certain instructions need to remain
7487 consecutive in the output, put them in a single multi-instruction @code{asm}
7488 statement. Note that GCC's optimizers can move @code{asm} statements
7489 relative to other code, including across jumps.
7491 @code{asm} statements may not perform jumps into other @code{asm} statements.
7492 GCC does not know about these jumps, and therefore cannot take
7493 account of them when deciding how to optimize. Jumps from @code{asm} to C
7494 labels are only supported in extended @code{asm}.
7496 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7497 assembly code when optimizing. This can lead to unexpected duplicate
7498 symbol errors during compilation if your assembly code defines symbols or
7501 Since GCC does not parse the @var{AssemblerInstructions}, it has no
7502 visibility of any symbols it references. This may result in GCC discarding
7503 those symbols as unreferenced.
7505 The compiler copies the assembler instructions in a basic @code{asm}
7506 verbatim to the assembly language output file, without
7507 processing dialects or any of the @samp{%} operators that are available with
7508 extended @code{asm}. This results in minor differences between basic
7509 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7510 registers you might use @samp{%eax} in basic @code{asm} and
7511 @samp{%%eax} in extended @code{asm}.
7513 On targets such as x86 that support multiple assembler dialects,
7514 all basic @code{asm} blocks use the assembler dialect specified by the
7515 @option{-masm} command-line option (@pxref{x86 Options}).
7516 Basic @code{asm} provides no
7517 mechanism to provide different assembler strings for different dialects.
7519 Here is an example of basic @code{asm} for i386:
7522 /* Note that this code will not compile with -masm=intel */
7523 #define DebugBreak() asm("int $3")
7527 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7528 @cindex extended @code{asm}
7529 @cindex assembly language in C, extended
7531 With extended @code{asm} you can read and write C variables from
7532 assembler and perform jumps from assembler code to C labels.
7533 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7534 the operand parameters after the assembler template:
7537 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7538 : @var{OutputOperands}
7539 @r{[} : @var{InputOperands}
7540 @r{[} : @var{Clobbers} @r{]} @r{]})
7542 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7544 : @var{InputOperands}
7549 The @code{asm} keyword is a GNU extension.
7550 When writing code that can be compiled with @option{-ansi} and the
7551 various @option{-std} options, use @code{__asm__} instead of
7552 @code{asm} (@pxref{Alternate Keywords}).
7554 @subsubheading Qualifiers
7558 The typical use of extended @code{asm} statements is to manipulate input
7559 values to produce output values. However, your @code{asm} statements may
7560 also produce side effects. If so, you may need to use the @code{volatile}
7561 qualifier to disable certain optimizations. @xref{Volatile}.
7564 This qualifier informs the compiler that the @code{asm} statement may
7565 perform a jump to one of the labels listed in the @var{GotoLabels}.
7569 @subsubheading Parameters
7571 @item AssemblerTemplate
7572 This is a literal string that is the template for the assembler code. It is a
7573 combination of fixed text and tokens that refer to the input, output,
7574 and goto parameters. @xref{AssemblerTemplate}.
7576 @item OutputOperands
7577 A comma-separated list of the C variables modified by the instructions in the
7578 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7581 A comma-separated list of C expressions read by the instructions in the
7582 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7585 A comma-separated list of registers or other values changed by the
7586 @var{AssemblerTemplate}, beyond those listed as outputs.
7587 An empty list is permitted. @xref{Clobbers}.
7590 When you are using the @code{goto} form of @code{asm}, this section contains
7591 the list of all C labels to which the code in the
7592 @var{AssemblerTemplate} may jump.
7595 @code{asm} statements may not perform jumps into other @code{asm} statements,
7596 only to the listed @var{GotoLabels}.
7597 GCC's optimizers do not know about other jumps; therefore they cannot take
7598 account of them when deciding how to optimize.
7601 The total number of input + output + goto operands is limited to 30.
7603 @subsubheading Remarks
7604 The @code{asm} statement allows you to include assembly instructions directly
7605 within C code. This may help you to maximize performance in time-sensitive
7606 code or to access assembly instructions that are not readily available to C
7609 Note that extended @code{asm} statements must be inside a function. Only
7610 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7611 Functions declared with the @code{naked} attribute also require basic
7612 @code{asm} (@pxref{Function Attributes}).
7614 While the uses of @code{asm} are many and varied, it may help to think of an
7615 @code{asm} statement as a series of low-level instructions that convert input
7616 parameters to output parameters. So a simple (if not particularly useful)
7617 example for i386 using @code{asm} might look like this:
7623 asm ("mov %1, %0\n\t"
7628 printf("%d\n", dst);
7631 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7634 @subsubsection Volatile
7635 @cindex volatile @code{asm}
7636 @cindex @code{asm} volatile
7638 GCC's optimizers sometimes discard @code{asm} statements if they determine
7639 there is no need for the output variables. Also, the optimizers may move
7640 code out of loops if they believe that the code will always return the same
7641 result (i.e. none of its input values change between calls). Using the
7642 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7643 that have no output operands, including @code{asm goto} statements,
7644 are implicitly volatile.
7646 This i386 code demonstrates a case that does not use (or require) the
7647 @code{volatile} qualifier. If it is performing assertion checking, this code
7648 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7649 unreferenced by any code. As a result, the optimizers can discard the
7650 @code{asm} statement, which in turn removes the need for the entire
7651 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7652 isn't needed you allow the optimizers to produce the most efficient code
7656 void DoCheck(uint32_t dwSomeValue)
7660 // Assumes dwSomeValue is not zero.
7670 The next example shows a case where the optimizers can recognize that the input
7671 (@code{dwSomeValue}) never changes during the execution of the function and can
7672 therefore move the @code{asm} outside the loop to produce more efficient code.
7673 Again, using @code{volatile} disables this type of optimization.
7676 void do_print(uint32_t dwSomeValue)
7680 for (uint32_t x=0; x < 5; x++)
7682 // Assumes dwSomeValue is not zero.
7688 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7693 The following example demonstrates a case where you need to use the
7694 @code{volatile} qualifier.
7695 It uses the x86 @code{rdtsc} instruction, which reads
7696 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7697 the optimizers might assume that the @code{asm} block will always return the
7698 same value and therefore optimize away the second call.
7703 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7704 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7705 "or %%rdx, %0" // 'Or' in the lower bits.
7710 printf("msr: %llx\n", msr);
7714 // Reprint the timestamp
7715 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7716 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7717 "or %%rdx, %0" // 'Or' in the lower bits.
7722 printf("msr: %llx\n", msr);
7725 GCC's optimizers do not treat this code like the non-volatile code in the
7726 earlier examples. They do not move it out of loops or omit it on the
7727 assumption that the result from a previous call is still valid.
7729 Note that the compiler can move even volatile @code{asm} instructions relative
7730 to other code, including across jump instructions. For example, on many
7731 targets there is a system register that controls the rounding mode of
7732 floating-point operations. Setting it with a volatile @code{asm}, as in the
7733 following PowerPC example, does not work reliably.
7736 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7740 The compiler may move the addition back before the volatile @code{asm}. To
7741 make it work as expected, add an artificial dependency to the @code{asm} by
7742 referencing a variable in the subsequent code, for example:
7745 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7749 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7750 assembly code when optimizing. This can lead to unexpected duplicate symbol
7751 errors during compilation if your asm code defines symbols or labels.
7753 (@pxref{AssemblerTemplate}) may help resolve this problem.
7755 @anchor{AssemblerTemplate}
7756 @subsubsection Assembler Template
7757 @cindex @code{asm} assembler template
7759 An assembler template is a literal string containing assembler instructions.
7760 The compiler replaces tokens in the template that refer
7761 to inputs, outputs, and goto labels,
7762 and then outputs the resulting string to the assembler. The
7763 string can contain any instructions recognized by the assembler, including
7764 directives. GCC does not parse the assembler instructions
7765 themselves and does not know what they mean or even whether they are valid
7766 assembler input. However, it does count the statements
7767 (@pxref{Size of an asm}).
7769 You may place multiple assembler instructions together in a single @code{asm}
7770 string, separated by the characters normally used in assembly code for the
7771 system. A combination that works in most places is a newline to break the
7772 line, plus a tab character to move to the instruction field (written as
7774 Some assemblers allow semicolons as a line separator. However, note
7775 that some assembler dialects use semicolons to start a comment.
7777 Do not expect a sequence of @code{asm} statements to remain perfectly
7778 consecutive after compilation, even when you are using the @code{volatile}
7779 qualifier. If certain instructions need to remain consecutive in the output,
7780 put them in a single multi-instruction asm statement.
7782 Accessing data from C programs without using input/output operands (such as
7783 by using global symbols directly from the assembler template) may not work as
7784 expected. Similarly, calling functions directly from an assembler template
7785 requires a detailed understanding of the target assembler and ABI.
7787 Since GCC does not parse the assembler template,
7788 it has no visibility of any
7789 symbols it references. This may result in GCC discarding those symbols as
7790 unreferenced unless they are also listed as input, output, or goto operands.
7792 @subsubheading Special format strings
7794 In addition to the tokens described by the input, output, and goto operands,
7795 these tokens have special meanings in the assembler template:
7799 Outputs a single @samp{%} into the assembler code.
7802 Outputs a number that is unique to each instance of the @code{asm}
7803 statement in the entire compilation. This option is useful when creating local
7804 labels and referring to them multiple times in a single template that
7805 generates multiple assembler instructions.
7810 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7811 into the assembler code. When unescaped, these characters have special
7812 meaning to indicate multiple assembler dialects, as described below.
7815 @subsubheading Multiple assembler dialects in @code{asm} templates
7817 On targets such as x86, GCC supports multiple assembler dialects.
7818 The @option{-masm} option controls which dialect GCC uses as its
7819 default for inline assembler. The target-specific documentation for the
7820 @option{-masm} option contains the list of supported dialects, as well as the
7821 default dialect if the option is not specified. This information may be
7822 important to understand, since assembler code that works correctly when
7823 compiled using one dialect will likely fail if compiled using another.
7826 If your code needs to support multiple assembler dialects (for example, if
7827 you are writing public headers that need to support a variety of compilation
7828 options), use constructs of this form:
7831 @{ dialect0 | dialect1 | dialect2... @}
7834 This construct outputs @code{dialect0}
7835 when using dialect #0 to compile the code,
7836 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7837 braces than the number of dialects the compiler supports, the construct
7840 For example, if an x86 compiler supports two dialects
7841 (@samp{att}, @samp{intel}), an
7842 assembler template such as this:
7845 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7849 is equivalent to one of
7852 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7853 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7856 Using that same compiler, this code:
7859 "xchg@{l@}\t@{%%@}ebx, %1"
7863 corresponds to either
7866 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7867 "xchg\tebx, %1" @r{/* intel dialect */}
7870 There is no support for nesting dialect alternatives.
7872 @anchor{OutputOperands}
7873 @subsubsection Output Operands
7874 @cindex @code{asm} output operands
7876 An @code{asm} statement has zero or more output operands indicating the names
7877 of C variables modified by the assembler code.
7879 In this i386 example, @code{old} (referred to in the template string as
7880 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
7881 (@code{%2}) is an input:
7886 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
7887 "sbb %0,%0" // Use the CF to calculate old.
7888 : "=r" (old), "+rm" (*Base)
7895 Operands are separated by commas. Each operand has this format:
7898 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
7902 @item asmSymbolicName
7903 Specifies a symbolic name for the operand.
7904 Reference the name in the assembler template
7905 by enclosing it in square brackets
7906 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
7907 that contains the definition. Any valid C variable name is acceptable,
7908 including names already defined in the surrounding code. No two operands
7909 within the same @code{asm} statement can use the same symbolic name.
7911 When not using an @var{asmSymbolicName}, use the (zero-based) position
7913 in the list of operands in the assembler template. For example if there are
7914 three output operands, use @samp{%0} in the template to refer to the first,
7915 @samp{%1} for the second, and @samp{%2} for the third.
7918 A string constant specifying constraints on the placement of the operand;
7919 @xref{Constraints}, for details.
7921 Output constraints must begin with either @samp{=} (a variable overwriting an
7922 existing value) or @samp{+} (when reading and writing). When using
7923 @samp{=}, do not assume the location contains the existing value
7924 on entry to the @code{asm}, except
7925 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
7927 After the prefix, there must be one or more additional constraints
7928 (@pxref{Constraints}) that describe where the value resides. Common
7929 constraints include @samp{r} for register and @samp{m} for memory.
7930 When you list more than one possible location (for example, @code{"=rm"}),
7931 the compiler chooses the most efficient one based on the current context.
7932 If you list as many alternates as the @code{asm} statement allows, you permit
7933 the optimizers to produce the best possible code.
7934 If you must use a specific register, but your Machine Constraints do not
7935 provide sufficient control to select the specific register you want,
7936 local register variables may provide a solution (@pxref{Local Register
7940 Specifies a C lvalue expression to hold the output, typically a variable name.
7941 The enclosing parentheses are a required part of the syntax.
7945 When the compiler selects the registers to use to
7946 represent the output operands, it does not use any of the clobbered registers
7949 Output operand expressions must be lvalues. The compiler cannot check whether
7950 the operands have data types that are reasonable for the instruction being
7951 executed. For output expressions that are not directly addressable (for
7952 example a bit-field), the constraint must allow a register. In that case, GCC
7953 uses the register as the output of the @code{asm}, and then stores that
7954 register into the output.
7956 Operands using the @samp{+} constraint modifier count as two operands
7957 (that is, both as input and output) towards the total maximum of 30 operands
7958 per @code{asm} statement.
7960 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
7961 operands that must not overlap an input. Otherwise,
7962 GCC may allocate the output operand in the same register as an unrelated
7963 input operand, on the assumption that the assembler code consumes its
7964 inputs before producing outputs. This assumption may be false if the assembler
7965 code actually consists of more than one instruction.
7967 The same problem can occur if one output parameter (@var{a}) allows a register
7968 constraint and another output parameter (@var{b}) allows a memory constraint.
7969 The code generated by GCC to access the memory address in @var{b} can contain
7970 registers which @emph{might} be shared by @var{a}, and GCC considers those
7971 registers to be inputs to the asm. As above, GCC assumes that such input
7972 registers are consumed before any outputs are written. This assumption may
7973 result in incorrect behavior if the asm writes to @var{a} before using
7974 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
7975 ensures that modifying @var{a} does not affect the address referenced by
7976 @var{b}. Otherwise, the location of @var{b}
7977 is undefined if @var{a} is modified before using @var{b}.
7979 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
7980 instead of simply @samp{%2}). Typically these qualifiers are hardware
7981 dependent. The list of supported modifiers for x86 is found at
7982 @ref{x86Operandmodifiers,x86 Operand modifiers}.
7984 If the C code that follows the @code{asm} makes no use of any of the output
7985 operands, use @code{volatile} for the @code{asm} statement to prevent the
7986 optimizers from discarding the @code{asm} statement as unneeded
7987 (see @ref{Volatile}).
7989 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
7990 references the first output operand as @code{%0} (were there a second, it
7991 would be @code{%1}, etc). The number of the first input operand is one greater
7992 than that of the last output operand. In this i386 example, that makes
7993 @code{Mask} referenced as @code{%1}:
7996 uint32_t Mask = 1234;
8005 That code overwrites the variable @code{Index} (@samp{=}),
8006 placing the value in a register (@samp{r}).
8007 Using the generic @samp{r} constraint instead of a constraint for a specific
8008 register allows the compiler to pick the register to use, which can result
8009 in more efficient code. This may not be possible if an assembler instruction
8010 requires a specific register.
8012 The following i386 example uses the @var{asmSymbolicName} syntax.
8014 same result as the code above, but some may consider it more readable or more
8015 maintainable since reordering index numbers is not necessary when adding or
8016 removing operands. The names @code{aIndex} and @code{aMask}
8017 are only used in this example to emphasize which
8018 names get used where.
8019 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8022 uint32_t Mask = 1234;
8025 asm ("bsfl %[aMask], %[aIndex]"
8026 : [aIndex] "=r" (Index)
8027 : [aMask] "r" (Mask)
8031 Here are some more examples of output operands.
8038 asm ("mov %[e], %[d]"
8043 Here, @code{d} may either be in a register or in memory. Since the compiler
8044 might already have the current value of the @code{uint32_t} location
8045 pointed to by @code{e}
8046 in a register, you can enable it to choose the best location
8047 for @code{d} by specifying both constraints.
8049 @anchor{FlagOutputOperands}
8050 @subsection Flag Output Operands
8051 @cindex @code{asm} flag output operands
8053 Some targets have a special register that holds the ``flags'' for the
8054 result of an operation or comparison. Normally, the contents of that
8055 register are either unmodifed by the asm, or the asm is considered to
8056 clobber the contents.
8058 On some targets, a special form of output operand exists by which
8059 conditions in the flags register may be outputs of the asm. The set of
8060 conditions supported are target specific, but the general rule is that
8061 the output variable must be a scalar integer, and the value will be boolean.
8062 When supported, the target will define the preprocessor symbol
8063 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8065 Because of the special nature of the flag output operands, the constraint
8066 may not include alternatives.
8068 Most often, the target has only one flags register, and thus is an implied
8069 operand of many instructions. In this case, the operand should not be
8070 referenced within the assembler template via @code{%0} etc, as there's
8071 no corresponding text in the assembly language.
8075 The flag output constraints for the x86 family are of the form
8076 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8077 conditions defined in the ISA manual for @code{j@var{cc}} or
8082 ``above'' or unsigned greater than
8084 ``above or equal'' or unsigned greater than or equal
8086 ``below'' or unsigned less than
8088 ``below or equal'' or unsigned less than or equal
8093 ``equal'' or zero flag set
8097 signed greater than or equal
8101 signed less than or equal
8122 ``not'' @var{flag}, or inverted versions of those above
8127 @anchor{InputOperands}
8128 @subsubsection Input Operands
8129 @cindex @code{asm} input operands
8130 @cindex @code{asm} expressions
8132 Input operands make values from C variables and expressions available to the
8135 Operands are separated by commas. Each operand has this format:
8138 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8142 @item asmSymbolicName
8143 Specifies a symbolic name for the operand.
8144 Reference the name in the assembler template
8145 by enclosing it in square brackets
8146 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8147 that contains the definition. Any valid C variable name is acceptable,
8148 including names already defined in the surrounding code. No two operands
8149 within the same @code{asm} statement can use the same symbolic name.
8151 When not using an @var{asmSymbolicName}, use the (zero-based) position
8153 in the list of operands in the assembler template. For example if there are
8154 two output operands and three inputs,
8155 use @samp{%2} in the template to refer to the first input operand,
8156 @samp{%3} for the second, and @samp{%4} for the third.
8159 A string constant specifying constraints on the placement of the operand;
8160 @xref{Constraints}, for details.
8162 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8163 When you list more than one possible location (for example, @samp{"irm"}),
8164 the compiler chooses the most efficient one based on the current context.
8165 If you must use a specific register, but your Machine Constraints do not
8166 provide sufficient control to select the specific register you want,
8167 local register variables may provide a solution (@pxref{Local Register
8170 Input constraints can also be digits (for example, @code{"0"}). This indicates
8171 that the specified input must be in the same place as the output constraint
8172 at the (zero-based) index in the output constraint list.
8173 When using @var{asmSymbolicName} syntax for the output operands,
8174 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8177 This is the C variable or expression being passed to the @code{asm} statement
8178 as input. The enclosing parentheses are a required part of the syntax.
8182 When the compiler selects the registers to use to represent the input
8183 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8185 If there are no output operands but there are input operands, place two
8186 consecutive colons where the output operands would go:
8189 __asm__ ("some instructions"
8191 : "r" (Offset / 8));
8194 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8195 (except for inputs tied to outputs). The compiler assumes that on exit from
8196 the @code{asm} statement these operands contain the same values as they
8197 had before executing the statement.
8198 It is @emph{not} possible to use clobbers
8199 to inform the compiler that the values in these inputs are changing. One
8200 common work-around is to tie the changing input variable to an output variable
8201 that never gets used. Note, however, that if the code that follows the
8202 @code{asm} statement makes no use of any of the output operands, the GCC
8203 optimizers may discard the @code{asm} statement as unneeded
8204 (see @ref{Volatile}).
8206 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8207 instead of simply @samp{%2}). Typically these qualifiers are hardware
8208 dependent. The list of supported modifiers for x86 is found at
8209 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8211 In this example using the fictitious @code{combine} instruction, the
8212 constraint @code{"0"} for input operand 1 says that it must occupy the same
8213 location as output operand 0. Only input operands may use numbers in
8214 constraints, and they must each refer to an output operand. Only a number (or
8215 the symbolic assembler name) in the constraint can guarantee that one operand
8216 is in the same place as another. The mere fact that @code{foo} is the value of
8217 both operands is not enough to guarantee that they are in the same place in
8218 the generated assembler code.
8221 asm ("combine %2, %0"
8223 : "0" (foo), "g" (bar));
8226 Here is an example using symbolic names.
8229 asm ("cmoveq %1, %2, %[result]"
8230 : [result] "=r"(result)
8231 : "r" (test), "r" (new), "[result]" (old));
8235 @subsubsection Clobbers
8236 @cindex @code{asm} clobbers
8238 While the compiler is aware of changes to entries listed in the output
8239 operands, the inline @code{asm} code may modify more than just the outputs. For
8240 example, calculations may require additional registers, or the processor may
8241 overwrite a register as a side effect of a particular assembler instruction.
8242 In order to inform the compiler of these changes, list them in the clobber
8243 list. Clobber list items are either register names or the special clobbers
8244 (listed below). Each clobber list item is a string constant
8245 enclosed in double quotes and separated by commas.
8247 Clobber descriptions may not in any way overlap with an input or output
8248 operand. For example, you may not have an operand describing a register class
8249 with one member when listing that register in the clobber list. Variables
8250 declared to live in specific registers (@pxref{Explicit Register
8251 Variables}) and used
8252 as @code{asm} input or output operands must have no part mentioned in the
8253 clobber description. In particular, there is no way to specify that input
8254 operands get modified without also specifying them as output operands.
8256 When the compiler selects which registers to use to represent input and output
8257 operands, it does not use any of the clobbered registers. As a result,
8258 clobbered registers are available for any use in the assembler code.
8260 Here is a realistic example for the VAX showing the use of clobbered
8264 asm volatile ("movc3 %0, %1, %2"
8266 : "g" (from), "g" (to), "g" (count)
8267 : "r0", "r1", "r2", "r3", "r4", "r5");
8270 Also, there are two special clobber arguments:
8274 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8275 register. On some machines, GCC represents the condition codes as a specific
8276 hardware register; @code{"cc"} serves to name this register.
8277 On other machines, condition code handling is different,
8278 and specifying @code{"cc"} has no effect. But
8279 it is valid no matter what the target.
8282 The @code{"memory"} clobber tells the compiler that the assembly code
8284 reads or writes to items other than those listed in the input and output
8285 operands (for example, accessing the memory pointed to by one of the input
8286 parameters). To ensure memory contains correct values, GCC may need to flush
8287 specific register values to memory before executing the @code{asm}. Further,
8288 the compiler does not assume that any values read from memory before an
8289 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8291 Using the @code{"memory"} clobber effectively forms a read/write
8292 memory barrier for the compiler.
8294 Note that this clobber does not prevent the @emph{processor} from doing
8295 speculative reads past the @code{asm} statement. To prevent that, you need
8296 processor-specific fence instructions.
8298 Flushing registers to memory has performance implications and may be an issue
8299 for time-sensitive code. You can use a trick to avoid this if the size of
8300 the memory being accessed is known at compile time. For example, if accessing
8301 ten bytes of a string, use a memory input like:
8303 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8308 @subsubsection Goto Labels
8309 @cindex @code{asm} goto labels
8311 @code{asm goto} allows assembly code to jump to one or more C labels. The
8312 @var{GotoLabels} section in an @code{asm goto} statement contains
8314 list of all C labels to which the assembler code may jump. GCC assumes that
8315 @code{asm} execution falls through to the next statement (if this is not the
8316 case, consider using the @code{__builtin_unreachable} intrinsic after the
8317 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8318 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8321 An @code{asm goto} statement cannot have outputs.
8322 This is due to an internal restriction of
8323 the compiler: control transfer instructions cannot have outputs.
8324 If the assembler code does modify anything, use the @code{"memory"} clobber
8326 optimizers to flush all register values to memory and reload them if
8327 necessary after the @code{asm} statement.
8329 Also note that an @code{asm goto} statement is always implicitly
8330 considered volatile.
8332 To reference a label in the assembler template,
8333 prefix it with @samp{%l} (lowercase @samp{L}) followed
8334 by its (zero-based) position in @var{GotoLabels} plus the number of input
8335 operands. For example, if the @code{asm} has three inputs and references two
8336 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8338 Alternately, you can reference labels using the actual C label name enclosed
8339 in brackets. For example, to reference a label named @code{carry}, you can
8340 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8341 section when using this approach.
8343 Here is an example of @code{asm goto} for i386:
8350 : "r" (p1), "r" (p2)
8360 The following example shows an @code{asm goto} that uses a memory clobber.
8366 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8377 @anchor{x86Operandmodifiers}
8378 @subsubsection x86 Operand Modifiers
8380 References to input, output, and goto operands in the assembler template
8381 of extended @code{asm} statements can use
8382 modifiers to affect the way the operands are formatted in
8383 the code output to the assembler. For example, the
8384 following code uses the @samp{h} and @samp{b} modifiers for x86:
8388 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8392 These modifiers generate this assembler code:
8398 The rest of this discussion uses the following code for illustrative purposes.
8407 asm volatile goto ("some assembler instructions here"
8409 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8410 : /* No clobbers. */
8415 With no modifiers, this is what the output from the operands would be for the
8416 @samp{att} and @samp{intel} dialects of assembler:
8418 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8419 @headitem Operand @tab masm=att @tab masm=intel
8428 @tab @code{OFFSET FLAT:.L2}
8431 The table below shows the list of supported modifiers and their effects.
8433 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8434 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8436 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8441 @tab Print the QImode name of the register.
8446 @tab Print the QImode name for a ``high'' register.
8451 @tab Print the HImode name of the register.
8456 @tab Print the SImode name of the register.
8461 @tab Print the DImode name of the register.
8466 @tab Print the label name with no punctuation.
8471 @tab Require a constant operand and print the constant expression with no punctuation.
8477 @anchor{x86floatingpointasmoperands}
8478 @subsubsection x86 Floating-Point @code{asm} Operands
8480 On x86 targets, there are several rules on the usage of stack-like registers
8481 in the operands of an @code{asm}. These rules apply only to the operands
8482 that are stack-like registers:
8486 Given a set of input registers that die in an @code{asm}, it is
8487 necessary to know which are implicitly popped by the @code{asm}, and
8488 which must be explicitly popped by GCC@.
8490 An input register that is implicitly popped by the @code{asm} must be
8491 explicitly clobbered, unless it is constrained to match an
8495 For any input register that is implicitly popped by an @code{asm}, it is
8496 necessary to know how to adjust the stack to compensate for the pop.
8497 If any non-popped input is closer to the top of the reg-stack than
8498 the implicitly popped register, it would not be possible to know what the
8499 stack looked like---it's not clear how the rest of the stack ``slides
8502 All implicitly popped input registers must be closer to the top of
8503 the reg-stack than any input that is not implicitly popped.
8505 It is possible that if an input dies in an @code{asm}, the compiler might
8506 use the input register for an output reload. Consider this example:
8509 asm ("foo" : "=t" (a) : "f" (b));
8513 This code says that input @code{b} is not popped by the @code{asm}, and that
8514 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8515 deeper after the @code{asm} than it was before. But, it is possible that
8516 reload may think that it can use the same register for both the input and
8519 To prevent this from happening,
8520 if any input operand uses the @samp{f} constraint, all output register
8521 constraints must use the @samp{&} early-clobber modifier.
8523 The example above is correctly written as:
8526 asm ("foo" : "=&t" (a) : "f" (b));
8530 Some operands need to be in particular places on the stack. All
8531 output operands fall in this category---GCC has no other way to
8532 know which registers the outputs appear in unless you indicate
8533 this in the constraints.
8535 Output operands must specifically indicate which register an output
8536 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8537 constraints must select a class with a single register.
8540 Output operands may not be ``inserted'' between existing stack registers.
8541 Since no 387 opcode uses a read/write operand, all output operands
8542 are dead before the @code{asm}, and are pushed by the @code{asm}.
8543 It makes no sense to push anywhere but the top of the reg-stack.
8545 Output operands must start at the top of the reg-stack: output
8546 operands may not ``skip'' a register.
8549 Some @code{asm} statements may need extra stack space for internal
8550 calculations. This can be guaranteed by clobbering stack registers
8551 unrelated to the inputs and outputs.
8556 takes one input, which is internally popped, and produces two outputs.
8559 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8563 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8564 and replaces them with one output. The @code{st(1)} clobber is necessary
8565 for the compiler to know that @code{fyl2xp1} pops both inputs.
8568 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8576 @subsection Controlling Names Used in Assembler Code
8577 @cindex assembler names for identifiers
8578 @cindex names used in assembler code
8579 @cindex identifiers, names in assembler code
8581 You can specify the name to be used in the assembler code for a C
8582 function or variable by writing the @code{asm} (or @code{__asm__})
8583 keyword after the declarator.
8584 It is up to you to make sure that the assembler names you choose do not
8585 conflict with any other assembler symbols, or reference registers.
8587 @subsubheading Assembler names for data:
8589 This sample shows how to specify the assembler name for data:
8592 int foo asm ("myfoo") = 2;
8596 This specifies that the name to be used for the variable @code{foo} in
8597 the assembler code should be @samp{myfoo} rather than the usual
8600 On systems where an underscore is normally prepended to the name of a C
8601 variable, this feature allows you to define names for the
8602 linker that do not start with an underscore.
8604 GCC does not support using this feature with a non-static local variable
8605 since such variables do not have assembler names. If you are
8606 trying to put the variable in a particular register, see
8607 @ref{Explicit Register Variables}.
8609 @subsubheading Assembler names for functions:
8611 To specify the assembler name for functions, write a declaration for the
8612 function before its definition and put @code{asm} there, like this:
8615 int func (int x, int y) asm ("MYFUNC");
8617 int func (int x, int y)
8623 This specifies that the name to be used for the function @code{func} in
8624 the assembler code should be @code{MYFUNC}.
8626 @node Explicit Register Variables
8627 @subsection Variables in Specified Registers
8628 @anchor{Explicit Reg Vars}
8629 @cindex explicit register variables
8630 @cindex variables in specified registers
8631 @cindex specified registers
8633 GNU C allows you to associate specific hardware registers with C
8634 variables. In almost all cases, allowing the compiler to assign
8635 registers produces the best code. However under certain unusual
8636 circumstances, more precise control over the variable storage is
8639 Both global and local variables can be associated with a register. The
8640 consequences of performing this association are very different between
8641 the two, as explained in the sections below.
8644 * Global Register Variables:: Variables declared at global scope.
8645 * Local Register Variables:: Variables declared within a function.
8648 @node Global Register Variables
8649 @subsubsection Defining Global Register Variables
8650 @anchor{Global Reg Vars}
8651 @cindex global register variables
8652 @cindex registers, global variables in
8653 @cindex registers, global allocation
8655 You can define a global register variable and associate it with a specified
8659 register int *foo asm ("r12");
8663 Here @code{r12} is the name of the register that should be used. Note that
8664 this is the same syntax used for defining local register variables, but for
8665 a global variable the declaration appears outside a function. The
8666 @code{register} keyword is required, and cannot be combined with
8667 @code{static}. The register name must be a valid register name for the
8670 Registers are a scarce resource on most systems and allowing the
8671 compiler to manage their usage usually results in the best code. However,
8672 under special circumstances it can make sense to reserve some globally.
8673 For example this may be useful in programs such as programming language
8674 interpreters that have a couple of global variables that are accessed
8677 After defining a global register variable, for the current compilation
8681 @item The register is reserved entirely for this use, and will not be
8682 allocated for any other purpose.
8683 @item The register is not saved and restored by any functions.
8684 @item Stores into this register are never deleted even if they appear to be
8685 dead, but references may be deleted, moved or simplified.
8688 Note that these points @emph{only} apply to code that is compiled with the
8689 definition. The behavior of code that is merely linked in (for example
8690 code from libraries) is not affected.
8692 If you want to recompile source files that do not actually use your global
8693 register variable so they do not use the specified register for any other
8694 purpose, you need not actually add the global register declaration to
8695 their source code. It suffices to specify the compiler option
8696 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8699 @subsubheading Declaring the variable
8701 Global register variables can not have initial values, because an
8702 executable file has no means to supply initial contents for a register.
8704 When selecting a register, choose one that is normally saved and
8705 restored by function calls on your machine. This ensures that code
8706 which is unaware of this reservation (such as library routines) will
8707 restore it before returning.
8709 On machines with register windows, be sure to choose a global
8710 register that is not affected magically by the function call mechanism.
8712 @subsubheading Using the variable
8714 @cindex @code{qsort}, and global register variables
8715 When calling routines that are not aware of the reservation, be
8716 cautious if those routines call back into code which uses them. As an
8717 example, if you call the system library version of @code{qsort}, it may
8718 clobber your registers during execution, but (if you have selected
8719 appropriate registers) it will restore them before returning. However
8720 it will @emph{not} restore them before calling @code{qsort}'s comparison
8721 function. As a result, global values will not reliably be available to
8722 the comparison function unless the @code{qsort} function itself is rebuilt.
8724 Similarly, it is not safe to access the global register variables from signal
8725 handlers or from more than one thread of control. Unless you recompile
8726 them specially for the task at hand, the system library routines may
8727 temporarily use the register for other things.
8729 @cindex register variable after @code{longjmp}
8730 @cindex global register after @code{longjmp}
8731 @cindex value after @code{longjmp}
8734 On most machines, @code{longjmp} restores to each global register
8735 variable the value it had at the time of the @code{setjmp}. On some
8736 machines, however, @code{longjmp} does not change the value of global
8737 register variables. To be portable, the function that called @code{setjmp}
8738 should make other arrangements to save the values of the global register
8739 variables, and to restore them in a @code{longjmp}. This way, the same
8740 thing happens regardless of what @code{longjmp} does.
8742 Eventually there may be a way of asking the compiler to choose a register
8743 automatically, but first we need to figure out how it should choose and
8744 how to enable you to guide the choice. No solution is evident.
8746 @node Local Register Variables
8747 @subsubsection Specifying Registers for Local Variables
8748 @anchor{Local Reg Vars}
8749 @cindex local variables, specifying registers
8750 @cindex specifying registers for local variables
8751 @cindex registers for local variables
8753 You can define a local register variable and associate it with a specified
8757 register int *foo asm ("r12");
8761 Here @code{r12} is the name of the register that should be used. Note
8762 that this is the same syntax used for defining global register variables,
8763 but for a local variable the declaration appears within a function. The
8764 @code{register} keyword is required, and cannot be combined with
8765 @code{static}. The register name must be a valid register name for the
8768 As with global register variables, it is recommended that you choose
8769 a register that is normally saved and restored by function calls on your
8770 machine, so that calls to library routines will not clobber it.
8772 The only supported use for this feature is to specify registers
8773 for input and output operands when calling Extended @code{asm}
8774 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8775 particular machine don't provide sufficient control to select the desired
8776 register. To force an operand into a register, create a local variable
8777 and specify the register name after the variable's declaration. Then use
8778 the local variable for the @code{asm} operand and specify any constraint
8779 letter that matches the register:
8782 register int *p1 asm ("r0") = @dots{};
8783 register int *p2 asm ("r1") = @dots{};
8784 register int *result asm ("r0");
8785 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8788 @emph{Warning:} In the above example, be aware that a register (for example
8789 @code{r0}) can be call-clobbered by subsequent code, including function
8790 calls and library calls for arithmetic operators on other variables (for
8791 example the initialization of @code{p2}). In this case, use temporary
8792 variables for expressions between the register assignments:
8796 register int *p1 asm ("r0") = @dots{};
8797 register int *p2 asm ("r1") = t1;
8798 register int *result asm ("r0");
8799 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8802 Defining a register variable does not reserve the register. Other than
8803 when invoking the Extended @code{asm}, the contents of the specified
8804 register are not guaranteed. For this reason, the following uses
8805 are explicitly @emph{not} supported. If they appear to work, it is only
8806 happenstance, and may stop working as intended due to (seemingly)
8807 unrelated changes in surrounding code, or even minor changes in the
8808 optimization of a future version of gcc:
8811 @item Passing parameters to or from Basic @code{asm}
8812 @item Passing parameters to or from Extended @code{asm} without using input
8814 @item Passing parameters to or from routines written in assembler (or
8815 other languages) using non-standard calling conventions.
8818 Some developers use Local Register Variables in an attempt to improve
8819 gcc's allocation of registers, especially in large functions. In this
8820 case the register name is essentially a hint to the register allocator.
8821 While in some instances this can generate better code, improvements are
8822 subject to the whims of the allocator/optimizers. Since there are no
8823 guarantees that your improvements won't be lost, this usage of Local
8824 Register Variables is discouraged.
8826 On the MIPS platform, there is related use for local register variables
8827 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8828 Defining coprocessor specifics for MIPS targets, gccint,
8829 GNU Compiler Collection (GCC) Internals}).
8831 @node Size of an asm
8832 @subsection Size of an @code{asm}
8834 Some targets require that GCC track the size of each instruction used
8835 in order to generate correct code. Because the final length of the
8836 code produced by an @code{asm} statement is only known by the
8837 assembler, GCC must make an estimate as to how big it will be. It
8838 does this by counting the number of instructions in the pattern of the
8839 @code{asm} and multiplying that by the length of the longest
8840 instruction supported by that processor. (When working out the number
8841 of instructions, it assumes that any occurrence of a newline or of
8842 whatever statement separator character is supported by the assembler --
8843 typically @samp{;} --- indicates the end of an instruction.)
8845 Normally, GCC's estimate is adequate to ensure that correct
8846 code is generated, but it is possible to confuse the compiler if you use
8847 pseudo instructions or assembler macros that expand into multiple real
8848 instructions, or if you use assembler directives that expand to more
8849 space in the object file than is needed for a single instruction.
8850 If this happens then the assembler may produce a diagnostic saying that
8851 a label is unreachable.
8853 @node Alternate Keywords
8854 @section Alternate Keywords
8855 @cindex alternate keywords
8856 @cindex keywords, alternate
8858 @option{-ansi} and the various @option{-std} options disable certain
8859 keywords. This causes trouble when you want to use GNU C extensions, or
8860 a general-purpose header file that should be usable by all programs,
8861 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8862 @code{inline} are not available in programs compiled with
8863 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8864 program compiled with @option{-std=c99} or @option{-std=c11}). The
8866 @code{restrict} is only available when @option{-std=gnu99} (which will
8867 eventually be the default) or @option{-std=c99} (or the equivalent
8868 @option{-std=iso9899:1999}), or an option for a later standard
8871 The way to solve these problems is to put @samp{__} at the beginning and
8872 end of each problematical keyword. For example, use @code{__asm__}
8873 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8875 Other C compilers won't accept these alternative keywords; if you want to
8876 compile with another compiler, you can define the alternate keywords as
8877 macros to replace them with the customary keywords. It looks like this:
8885 @findex __extension__
8887 @option{-pedantic} and other options cause warnings for many GNU C extensions.
8889 prevent such warnings within one expression by writing
8890 @code{__extension__} before the expression. @code{__extension__} has no
8891 effect aside from this.
8893 @node Incomplete Enums
8894 @section Incomplete @code{enum} Types
8896 You can define an @code{enum} tag without specifying its possible values.
8897 This results in an incomplete type, much like what you get if you write
8898 @code{struct foo} without describing the elements. A later declaration
8899 that does specify the possible values completes the type.
8901 You can't allocate variables or storage using the type while it is
8902 incomplete. However, you can work with pointers to that type.
8904 This extension may not be very useful, but it makes the handling of
8905 @code{enum} more consistent with the way @code{struct} and @code{union}
8908 This extension is not supported by GNU C++.
8910 @node Function Names
8911 @section Function Names as Strings
8912 @cindex @code{__func__} identifier
8913 @cindex @code{__FUNCTION__} identifier
8914 @cindex @code{__PRETTY_FUNCTION__} identifier
8916 GCC provides three magic variables that hold the name of the current
8917 function, as a string. The first of these is @code{__func__}, which
8918 is part of the C99 standard:
8920 The identifier @code{__func__} is implicitly declared by the translator
8921 as if, immediately following the opening brace of each function
8922 definition, the declaration
8925 static const char __func__[] = "function-name";
8929 appeared, where function-name is the name of the lexically-enclosing
8930 function. This name is the unadorned name of the function.
8932 @code{__FUNCTION__} is another name for @code{__func__}, provided for
8933 backward compatibility with old versions of GCC.
8935 In C, @code{__PRETTY_FUNCTION__} is yet another name for
8936 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
8937 the type signature of the function as well as its bare name. For
8938 example, this program:
8942 extern int printf (char *, ...);
8949 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
8950 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
8968 __PRETTY_FUNCTION__ = void a::sub(int)
8971 These identifiers are variables, not preprocessor macros, and may not
8972 be used to initialize @code{char} arrays or be concatenated with other string
8975 @node Return Address
8976 @section Getting the Return or Frame Address of a Function
8978 These functions may be used to get information about the callers of a
8981 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
8982 This function returns the return address of the current function, or of
8983 one of its callers. The @var{level} argument is number of frames to
8984 scan up the call stack. A value of @code{0} yields the return address
8985 of the current function, a value of @code{1} yields the return address
8986 of the caller of the current function, and so forth. When inlining
8987 the expected behavior is that the function returns the address of
8988 the function that is returned to. To work around this behavior use
8989 the @code{noinline} function attribute.
8991 The @var{level} argument must be a constant integer.
8993 On some machines it may be impossible to determine the return address of
8994 any function other than the current one; in such cases, or when the top
8995 of the stack has been reached, this function returns @code{0} or a
8996 random value. In addition, @code{__builtin_frame_address} may be used
8997 to determine if the top of the stack has been reached.
8999 Additional post-processing of the returned value may be needed, see
9000 @code{__builtin_extract_return_addr}.
9002 Calling this function with a nonzero argument can have unpredictable
9003 effects, including crashing the calling program. As a result, calls
9004 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9005 option is in effect. Such calls should only be made in debugging
9009 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9010 The address as returned by @code{__builtin_return_address} may have to be fed
9011 through this function to get the actual encoded address. For example, on the
9012 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9013 platforms an offset has to be added for the true next instruction to be
9016 If no fixup is needed, this function simply passes through @var{addr}.
9019 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9020 This function does the reverse of @code{__builtin_extract_return_addr}.
9023 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9024 This function is similar to @code{__builtin_return_address}, but it
9025 returns the address of the function frame rather than the return address
9026 of the function. Calling @code{__builtin_frame_address} with a value of
9027 @code{0} yields the frame address of the current function, a value of
9028 @code{1} yields the frame address of the caller of the current function,
9031 The frame is the area on the stack that holds local variables and saved
9032 registers. The frame address is normally the address of the first word
9033 pushed on to the stack by the function. However, the exact definition
9034 depends upon the processor and the calling convention. If the processor
9035 has a dedicated frame pointer register, and the function has a frame,
9036 then @code{__builtin_frame_address} returns the value of the frame
9039 On some machines it may be impossible to determine the frame address of
9040 any function other than the current one; in such cases, or when the top
9041 of the stack has been reached, this function returns @code{0} if
9042 the first frame pointer is properly initialized by the startup code.
9044 Calling this function with a nonzero argument can have unpredictable
9045 effects, including crashing the calling program. As a result, calls
9046 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9047 option is in effect. Such calls should only be made in debugging
9051 @node Vector Extensions
9052 @section Using Vector Instructions through Built-in Functions
9054 On some targets, the instruction set contains SIMD vector instructions which
9055 operate on multiple values contained in one large register at the same time.
9056 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9059 The first step in using these extensions is to provide the necessary data
9060 types. This should be done using an appropriate @code{typedef}:
9063 typedef int v4si __attribute__ ((vector_size (16)));
9067 The @code{int} type specifies the base type, while the attribute specifies
9068 the vector size for the variable, measured in bytes. For example, the
9069 declaration above causes the compiler to set the mode for the @code{v4si}
9070 type to be 16 bytes wide and divided into @code{int} sized units. For
9071 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9072 corresponding mode of @code{foo} is @acronym{V4SI}.
9074 The @code{vector_size} attribute is only applicable to integral and
9075 float scalars, although arrays, pointers, and function return values
9076 are allowed in conjunction with this construct. Only sizes that are
9077 a power of two are currently allowed.
9079 All the basic integer types can be used as base types, both as signed
9080 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9081 @code{long long}. In addition, @code{float} and @code{double} can be
9082 used to build floating-point vector types.
9084 Specifying a combination that is not valid for the current architecture
9085 causes GCC to synthesize the instructions using a narrower mode.
9086 For example, if you specify a variable of type @code{V4SI} and your
9087 architecture does not allow for this specific SIMD type, GCC
9088 produces code that uses 4 @code{SIs}.
9090 The types defined in this manner can be used with a subset of normal C
9091 operations. Currently, GCC allows using the following operators
9092 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9094 The operations behave like C++ @code{valarrays}. Addition is defined as
9095 the addition of the corresponding elements of the operands. For
9096 example, in the code below, each of the 4 elements in @var{a} is
9097 added to the corresponding 4 elements in @var{b} and the resulting
9098 vector is stored in @var{c}.
9101 typedef int v4si __attribute__ ((vector_size (16)));
9108 Subtraction, multiplication, division, and the logical operations
9109 operate in a similar manner. Likewise, the result of using the unary
9110 minus or complement operators on a vector type is a vector whose
9111 elements are the negative or complemented values of the corresponding
9112 elements in the operand.
9114 It is possible to use shifting operators @code{<<}, @code{>>} on
9115 integer-type vectors. The operation is defined as following: @code{@{a0,
9116 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9117 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9120 For convenience, it is allowed to use a binary vector operation
9121 where one operand is a scalar. In that case the compiler transforms
9122 the scalar operand into a vector where each element is the scalar from
9123 the operation. The transformation happens only if the scalar could be
9124 safely converted to the vector-element type.
9125 Consider the following code.
9128 typedef int v4si __attribute__ ((vector_size (16)));
9133 a = b + 1; /* a = b + @{1,1,1,1@}; */
9134 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9136 a = l + a; /* Error, cannot convert long to int. */
9139 Vectors can be subscripted as if the vector were an array with
9140 the same number of elements and base type. Out of bound accesses
9141 invoke undefined behavior at run time. Warnings for out of bound
9142 accesses for vector subscription can be enabled with
9143 @option{-Warray-bounds}.
9145 Vector comparison is supported with standard comparison
9146 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9147 vector expressions of integer-type or real-type. Comparison between
9148 integer-type vectors and real-type vectors are not supported. The
9149 result of the comparison is a vector of the same width and number of
9150 elements as the comparison operands with a signed integral element
9153 Vectors are compared element-wise producing 0 when comparison is false
9154 and -1 (constant of the appropriate type where all bits are set)
9155 otherwise. Consider the following example.
9158 typedef int v4si __attribute__ ((vector_size (16)));
9160 v4si a = @{1,2,3,4@};
9161 v4si b = @{3,2,1,4@};
9164 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9165 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9168 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9169 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9170 integer vector with the same number of elements of the same size as @code{b}
9171 and @code{c}, computes all three arguments and creates a vector
9172 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9173 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9174 As in the case of binary operations, this syntax is also accepted when
9175 one of @code{b} or @code{c} is a scalar that is then transformed into a
9176 vector. If both @code{b} and @code{c} are scalars and the type of
9177 @code{true?b:c} has the same size as the element type of @code{a}, then
9178 @code{b} and @code{c} are converted to a vector type whose elements have
9179 this type and with the same number of elements as @code{a}.
9181 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9182 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9183 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9184 For mixed operations between a scalar @code{s} and a vector @code{v},
9185 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9186 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9188 Vector shuffling is available using functions
9189 @code{__builtin_shuffle (vec, mask)} and
9190 @code{__builtin_shuffle (vec0, vec1, mask)}.
9191 Both functions construct a permutation of elements from one or two
9192 vectors and return a vector of the same type as the input vector(s).
9193 The @var{mask} is an integral vector with the same width (@var{W})
9194 and element count (@var{N}) as the output vector.
9196 The elements of the input vectors are numbered in memory ordering of
9197 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9198 elements of @var{mask} are considered modulo @var{N} in the single-operand
9199 case and modulo @math{2*@var{N}} in the two-operand case.
9201 Consider the following example,
9204 typedef int v4si __attribute__ ((vector_size (16)));
9206 v4si a = @{1,2,3,4@};
9207 v4si b = @{5,6,7,8@};
9208 v4si mask1 = @{0,1,1,3@};
9209 v4si mask2 = @{0,4,2,5@};
9212 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9213 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9216 Note that @code{__builtin_shuffle} is intentionally semantically
9217 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9219 You can declare variables and use them in function calls and returns, as
9220 well as in assignments and some casts. You can specify a vector type as
9221 a return type for a function. Vector types can also be used as function
9222 arguments. It is possible to cast from one vector type to another,
9223 provided they are of the same size (in fact, you can also cast vectors
9224 to and from other datatypes of the same size).
9226 You cannot operate between vectors of different lengths or different
9227 signedness without a cast.
9230 @section Support for @code{offsetof}
9231 @findex __builtin_offsetof
9233 GCC implements for both C and C++ a syntactic extension to implement
9234 the @code{offsetof} macro.
9238 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9240 offsetof_member_designator:
9242 | offsetof_member_designator "." @code{identifier}
9243 | offsetof_member_designator "[" @code{expr} "]"
9246 This extension is sufficient such that
9249 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9253 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9254 may be dependent. In either case, @var{member} may consist of a single
9255 identifier, or a sequence of member accesses and array references.
9257 @node __sync Builtins
9258 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9260 The following built-in functions
9261 are intended to be compatible with those described
9262 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9263 section 7.4. As such, they depart from normal GCC practice by not using
9264 the @samp{__builtin_} prefix and also by being overloaded so that they
9265 work on multiple types.
9267 The definition given in the Intel documentation allows only for the use of
9268 the types @code{int}, @code{long}, @code{long long} or their unsigned
9269 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
9270 size other than the C type @code{_Bool} or the C++ type @code{bool}.
9271 Operations on pointer arguments are performed as if the operands were
9272 of the @code{uintptr_t} type. That is, they are not scaled by the size
9273 of the type to which the pointer points.
9275 These functions are implemented in terms of the @samp{__atomic}
9276 builtins (@pxref{__atomic Builtins}). They should not be used for new
9277 code which should use the @samp{__atomic} builtins instead.
9279 Not all operations are supported by all target processors. If a particular
9280 operation cannot be implemented on the target processor, a warning is
9281 generated and a call to an external function is generated. The external
9282 function carries the same name as the built-in version,
9283 with an additional suffix
9284 @samp{_@var{n}} where @var{n} is the size of the data type.
9286 @c ??? Should we have a mechanism to suppress this warning? This is almost
9287 @c useful for implementing the operation under the control of an external
9290 In most cases, these built-in functions are considered a @dfn{full barrier}.
9292 no memory operand is moved across the operation, either forward or
9293 backward. Further, instructions are issued as necessary to prevent the
9294 processor from speculating loads across the operation and from queuing stores
9295 after the operation.
9297 All of the routines are described in the Intel documentation to take
9298 ``an optional list of variables protected by the memory barrier''. It's
9299 not clear what is meant by that; it could mean that @emph{only} the
9300 listed variables are protected, or it could mean a list of additional
9301 variables to be protected. The list is ignored by GCC which treats it as
9302 empty. GCC interprets an empty list as meaning that all globally
9303 accessible variables should be protected.
9306 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9307 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9308 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9309 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9310 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9311 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9312 @findex __sync_fetch_and_add
9313 @findex __sync_fetch_and_sub
9314 @findex __sync_fetch_and_or
9315 @findex __sync_fetch_and_and
9316 @findex __sync_fetch_and_xor
9317 @findex __sync_fetch_and_nand
9318 These built-in functions perform the operation suggested by the name, and
9319 returns the value that had previously been in memory. That is, operations
9320 on integer operands have the following semantics. Operations on pointer
9321 arguments are performed as if the operands were of the @code{uintptr_t}
9322 type. That is, they are not scaled by the size of the type to which
9326 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9327 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9330 The object pointed to by the first argument must be of integer or pointer
9331 type. It must not be a Boolean type.
9333 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9334 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9336 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9337 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9338 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9339 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9340 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9341 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9342 @findex __sync_add_and_fetch
9343 @findex __sync_sub_and_fetch
9344 @findex __sync_or_and_fetch
9345 @findex __sync_and_and_fetch
9346 @findex __sync_xor_and_fetch
9347 @findex __sync_nand_and_fetch
9348 These built-in functions perform the operation suggested by the name, and
9349 return the new value. That is, operations on integer operands have
9350 the following semantics. Operations on pointer operands are performed as
9351 if the operand's type were @code{uintptr_t}.
9354 @{ *ptr @var{op}= value; return *ptr; @}
9355 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9358 The same constraints on arguments apply as for the corresponding
9359 @code{__sync_op_and_fetch} built-in functions.
9361 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9362 as @code{*ptr = ~(*ptr & value)} instead of
9363 @code{*ptr = ~*ptr & value}.
9365 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9366 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9367 @findex __sync_bool_compare_and_swap
9368 @findex __sync_val_compare_and_swap
9369 These built-in functions perform an atomic compare and swap.
9370 That is, if the current
9371 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9374 The ``bool'' version returns true if the comparison is successful and
9375 @var{newval} is written. The ``val'' version returns the contents
9376 of @code{*@var{ptr}} before the operation.
9378 @item __sync_synchronize (...)
9379 @findex __sync_synchronize
9380 This built-in function issues a full memory barrier.
9382 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9383 @findex __sync_lock_test_and_set
9384 This built-in function, as described by Intel, is not a traditional test-and-set
9385 operation, but rather an atomic exchange operation. It writes @var{value}
9386 into @code{*@var{ptr}}, and returns the previous contents of
9389 Many targets have only minimal support for such locks, and do not support
9390 a full exchange operation. In this case, a target may support reduced
9391 functionality here by which the @emph{only} valid value to store is the
9392 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9393 is implementation defined.
9395 This built-in function is not a full barrier,
9396 but rather an @dfn{acquire barrier}.
9397 This means that references after the operation cannot move to (or be
9398 speculated to) before the operation, but previous memory stores may not
9399 be globally visible yet, and previous memory loads may not yet be
9402 @item void __sync_lock_release (@var{type} *ptr, ...)
9403 @findex __sync_lock_release
9404 This built-in function releases the lock acquired by
9405 @code{__sync_lock_test_and_set}.
9406 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9408 This built-in function is not a full barrier,
9409 but rather a @dfn{release barrier}.
9410 This means that all previous memory stores are globally visible, and all
9411 previous memory loads have been satisfied, but following memory reads
9412 are not prevented from being speculated to before the barrier.
9415 @node __atomic Builtins
9416 @section Built-in Functions for Memory Model Aware Atomic Operations
9418 The following built-in functions approximately match the requirements
9419 for the C++11 memory model. They are all
9420 identified by being prefixed with @samp{__atomic} and most are
9421 overloaded so that they work with multiple types.
9423 These functions are intended to replace the legacy @samp{__sync}
9424 builtins. The main difference is that the memory order that is requested
9425 is a parameter to the functions. New code should always use the
9426 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9428 Note that the @samp{__atomic} builtins assume that programs will
9429 conform to the C++11 memory model. In particular, they assume
9430 that programs are free of data races. See the C++11 standard for
9431 detailed requirements.
9433 The @samp{__atomic} builtins can be used with any integral scalar or
9434 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9435 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9436 supported by the architecture.
9438 The four non-arithmetic functions (load, store, exchange, and
9439 compare_exchange) all have a generic version as well. This generic
9440 version works on any data type. It uses the lock-free built-in function
9441 if the specific data type size makes that possible; otherwise, an
9442 external call is left to be resolved at run time. This external call is
9443 the same format with the addition of a @samp{size_t} parameter inserted
9444 as the first parameter indicating the size of the object being pointed to.
9445 All objects must be the same size.
9447 There are 6 different memory orders that can be specified. These map
9448 to the C++11 memory orders with the same names, see the C++11 standard
9449 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9450 on atomic synchronization} for detailed definitions. Individual
9451 targets may also support additional memory orders for use on specific
9452 architectures. Refer to the target documentation for details of
9455 An atomic operation can both constrain code motion and
9456 be mapped to hardware instructions for synchronization between threads
9457 (e.g., a fence). To which extent this happens is controlled by the
9458 memory orders, which are listed here in approximately ascending order of
9459 strength. The description of each memory order is only meant to roughly
9460 illustrate the effects and is not a specification; see the C++11
9461 memory model for precise semantics.
9464 @item __ATOMIC_RELAXED
9465 Implies no inter-thread ordering constraints.
9466 @item __ATOMIC_CONSUME
9467 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9468 memory order because of a deficiency in C++11's semantics for
9469 @code{memory_order_consume}.
9470 @item __ATOMIC_ACQUIRE
9471 Creates an inter-thread happens-before constraint from the release (or
9472 stronger) semantic store to this acquire load. Can prevent hoisting
9473 of code to before the operation.
9474 @item __ATOMIC_RELEASE
9475 Creates an inter-thread happens-before constraint to acquire (or stronger)
9476 semantic loads that read from this release store. Can prevent sinking
9477 of code to after the operation.
9478 @item __ATOMIC_ACQ_REL
9479 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9480 @code{__ATOMIC_RELEASE}.
9481 @item __ATOMIC_SEQ_CST
9482 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9485 Note that in the C++11 memory model, @emph{fences} (e.g.,
9486 @samp{__atomic_thread_fence}) take effect in combination with other
9487 atomic operations on specific memory locations (e.g., atomic loads);
9488 operations on specific memory locations do not necessarily affect other
9489 operations in the same way.
9491 Target architectures are encouraged to provide their own patterns for
9492 each of the atomic built-in functions. If no target is provided, the original
9493 non-memory model set of @samp{__sync} atomic built-in functions are
9494 used, along with any required synchronization fences surrounding it in
9495 order to achieve the proper behavior. Execution in this case is subject
9496 to the same restrictions as those built-in functions.
9498 If there is no pattern or mechanism to provide a lock-free instruction
9499 sequence, a call is made to an external routine with the same parameters
9500 to be resolved at run time.
9502 When implementing patterns for these built-in functions, the memory order
9503 parameter can be ignored as long as the pattern implements the most
9504 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9505 orders execute correctly with this memory order but they may not execute as
9506 efficiently as they could with a more appropriate implementation of the
9507 relaxed requirements.
9509 Note that the C++11 standard allows for the memory order parameter to be
9510 determined at run time rather than at compile time. These built-in
9511 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9512 than invoke a runtime library call or inline a switch statement. This is
9513 standard compliant, safe, and the simplest approach for now.
9515 The memory order parameter is a signed int, but only the lower 16 bits are
9516 reserved for the memory order. The remainder of the signed int is reserved
9517 for target use and should be 0. Use of the predefined atomic values
9518 ensures proper usage.
9520 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9521 This built-in function implements an atomic load operation. It returns the
9522 contents of @code{*@var{ptr}}.
9524 The valid memory order variants are
9525 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9526 and @code{__ATOMIC_CONSUME}.
9530 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9531 This is the generic version of an atomic load. It returns the
9532 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9536 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9537 This built-in function implements an atomic store operation. It writes
9538 @code{@var{val}} into @code{*@var{ptr}}.
9540 The valid memory order variants are
9541 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9545 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9546 This is the generic version of an atomic store. It stores the value
9547 of @code{*@var{val}} into @code{*@var{ptr}}.
9551 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9552 This built-in function implements an atomic exchange operation. It writes
9553 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9556 The valid memory order variants are
9557 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9558 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9562 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9563 This is the generic version of an atomic exchange. It stores the
9564 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9565 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9569 @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)
9570 This built-in function implements an atomic compare and exchange operation.
9571 This compares the contents of @code{*@var{ptr}} with the contents of
9572 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9573 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9574 equal, the operation is a @emph{read} and the current contents of
9575 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
9576 for weak compare_exchange, which may fail spuriously, and false for
9577 the strong variation, which never fails spuriously. Many targets
9578 only offer the strong variation and ignore the parameter. When in doubt, use
9579 the strong variation.
9581 If @var{desired} is written into @code{*@var{ptr}} then true is returned
9582 and memory is affected according to the
9583 memory order specified by @var{success_memorder}. There are no
9584 restrictions on what memory order can be used here.
9586 Otherwise, false is returned and memory is affected according
9587 to @var{failure_memorder}. This memory order cannot be
9588 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9589 stronger order than that specified by @var{success_memorder}.
9593 @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)
9594 This built-in function implements the generic version of
9595 @code{__atomic_compare_exchange}. The function is virtually identical to
9596 @code{__atomic_compare_exchange_n}, except the desired value is also a
9601 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9602 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9603 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9604 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9605 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9606 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9607 These built-in functions perform the operation suggested by the name, and
9608 return the result of the operation. Operations on pointer arguments are
9609 performed as if the operands were of the @code{uintptr_t} type. That is,
9610 they are not scaled by the size of the type to which the pointer points.
9613 @{ *ptr @var{op}= val; return *ptr; @}
9616 The object pointed to by the first argument must be of integer or pointer
9617 type. It must not be a Boolean type. All memory orders are valid.
9621 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9622 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9623 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9624 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9625 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9626 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9627 These built-in functions perform the operation suggested by the name, and
9628 return the value that had previously been in @code{*@var{ptr}}. Operations
9629 on pointer arguments are performed as if the operands were of
9630 the @code{uintptr_t} type. That is, they are not scaled by the size of
9631 the type to which the pointer points.
9634 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9637 The same constraints on arguments apply as for the corresponding
9638 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
9642 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9644 This built-in function performs an atomic test-and-set operation on
9645 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9646 defined nonzero ``set'' value and the return value is @code{true} if and only
9647 if the previous contents were ``set''.
9648 It should be only used for operands of type @code{bool} or @code{char}. For
9649 other types only part of the value may be set.
9651 All memory orders are valid.
9655 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9657 This built-in function performs an atomic clear operation on
9658 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9659 It should be only used for operands of type @code{bool} or @code{char} and
9660 in conjunction with @code{__atomic_test_and_set}.
9661 For other types it may only clear partially. If the type is not @code{bool}
9662 prefer using @code{__atomic_store}.
9664 The valid memory order variants are
9665 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9666 @code{__ATOMIC_RELEASE}.
9670 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9672 This built-in function acts as a synchronization fence between threads
9673 based on the specified memory order.
9675 All memory orders are valid.
9679 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9681 This built-in function acts as a synchronization fence between a thread
9682 and signal handlers based in the same thread.
9684 All memory orders are valid.
9688 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9690 This built-in function returns true if objects of @var{size} bytes always
9691 generate lock-free atomic instructions for the target architecture.
9692 @var{size} must resolve to a compile-time constant and the result also
9693 resolves to a compile-time constant.
9695 @var{ptr} is an optional pointer to the object that may be used to determine
9696 alignment. A value of 0 indicates typical alignment should be used. The
9697 compiler may also ignore this parameter.
9700 if (__atomic_always_lock_free (sizeof (long long), 0))
9705 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9707 This built-in function returns true if objects of @var{size} bytes always
9708 generate lock-free atomic instructions for the target architecture. If
9709 the built-in function is not known to be lock-free, a call is made to a
9710 runtime routine named @code{__atomic_is_lock_free}.
9712 @var{ptr} is an optional pointer to the object that may be used to determine
9713 alignment. A value of 0 indicates typical alignment should be used. The
9714 compiler may also ignore this parameter.
9717 @node Integer Overflow Builtins
9718 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9720 The following built-in functions allow performing simple arithmetic operations
9721 together with checking whether the operations overflowed.
9723 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9724 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9725 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9726 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9727 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9728 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9729 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9731 These built-in functions promote the first two operands into infinite precision signed
9732 type and perform addition on those promoted operands. The result is then
9733 cast to the type the third pointer argument points to and stored there.
9734 If the stored result is equal to the infinite precision result, the built-in
9735 functions return false, otherwise they return true. As the addition is
9736 performed in infinite signed precision, these built-in functions have fully defined
9737 behavior for all argument values.
9739 The first built-in function allows arbitrary integral types for operands and
9740 the result type must be pointer to some integer type, the rest of the built-in
9741 functions have explicit integer types.
9743 The compiler will attempt to use hardware instructions to implement
9744 these built-in functions where possible, like conditional jump on overflow
9745 after addition, conditional jump on carry etc.
9749 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9750 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9751 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9752 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9753 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9754 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9755 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9757 These built-in functions are similar to the add overflow checking built-in
9758 functions above, except they perform subtraction, subtract the second argument
9759 from the first one, instead of addition.
9763 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9764 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9765 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9766 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9767 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9768 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9769 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9771 These built-in functions are similar to the add overflow checking built-in
9772 functions above, except they perform multiplication, instead of addition.
9776 @node x86 specific memory model extensions for transactional memory
9777 @section x86-Specific Memory Model Extensions for Transactional Memory
9779 The x86 architecture supports additional memory ordering flags
9780 to mark lock critical sections for hardware lock elision.
9781 These must be specified in addition to an existing memory order to
9785 @item __ATOMIC_HLE_ACQUIRE
9786 Start lock elision on a lock variable.
9787 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9788 @item __ATOMIC_HLE_RELEASE
9789 End lock elision on a lock variable.
9790 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9793 When a lock acquire fails, it is required for good performance to abort
9794 the transaction quickly. This can be done with a @code{_mm_pause}.
9797 #include <immintrin.h> // For _mm_pause
9801 /* Acquire lock with lock elision */
9802 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9803 _mm_pause(); /* Abort failed transaction */
9805 /* Free lock with lock elision */
9806 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9809 @node Object Size Checking
9810 @section Object Size Checking Built-in Functions
9811 @findex __builtin_object_size
9812 @findex __builtin___memcpy_chk
9813 @findex __builtin___mempcpy_chk
9814 @findex __builtin___memmove_chk
9815 @findex __builtin___memset_chk
9816 @findex __builtin___strcpy_chk
9817 @findex __builtin___stpcpy_chk
9818 @findex __builtin___strncpy_chk
9819 @findex __builtin___strcat_chk
9820 @findex __builtin___strncat_chk
9821 @findex __builtin___sprintf_chk
9822 @findex __builtin___snprintf_chk
9823 @findex __builtin___vsprintf_chk
9824 @findex __builtin___vsnprintf_chk
9825 @findex __builtin___printf_chk
9826 @findex __builtin___vprintf_chk
9827 @findex __builtin___fprintf_chk
9828 @findex __builtin___vfprintf_chk
9830 GCC implements a limited buffer overflow protection mechanism
9831 that can prevent some buffer overflow attacks.
9833 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
9834 is a built-in construct that returns a constant number of bytes from
9835 @var{ptr} to the end of the object @var{ptr} pointer points to
9836 (if known at compile time). @code{__builtin_object_size} never evaluates
9837 its arguments for side-effects. If there are any side-effects in them, it
9838 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9839 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
9840 point to and all of them are known at compile time, the returned number
9841 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
9842 0 and minimum if nonzero. If it is not possible to determine which objects
9843 @var{ptr} points to at compile time, @code{__builtin_object_size} should
9844 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
9845 for @var{type} 2 or 3.
9847 @var{type} is an integer constant from 0 to 3. If the least significant
9848 bit is clear, objects are whole variables, if it is set, a closest
9849 surrounding subobject is considered the object a pointer points to.
9850 The second bit determines if maximum or minimum of remaining bytes
9854 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
9855 char *p = &var.buf1[1], *q = &var.b;
9857 /* Here the object p points to is var. */
9858 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
9859 /* The subobject p points to is var.buf1. */
9860 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
9861 /* The object q points to is var. */
9862 assert (__builtin_object_size (q, 0)
9863 == (char *) (&var + 1) - (char *) &var.b);
9864 /* The subobject q points to is var.b. */
9865 assert (__builtin_object_size (q, 1) == sizeof (var.b));
9869 There are built-in functions added for many common string operation
9870 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
9871 built-in is provided. This built-in has an additional last argument,
9872 which is the number of bytes remaining in object the @var{dest}
9873 argument points to or @code{(size_t) -1} if the size is not known.
9875 The built-in functions are optimized into the normal string functions
9876 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
9877 it is known at compile time that the destination object will not
9878 be overflown. If the compiler can determine at compile time the
9879 object will be always overflown, it issues a warning.
9881 The intended use can be e.g.@:
9885 #define bos0(dest) __builtin_object_size (dest, 0)
9886 #define memcpy(dest, src, n) \
9887 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
9891 /* It is unknown what object p points to, so this is optimized
9892 into plain memcpy - no checking is possible. */
9893 memcpy (p, "abcde", n);
9894 /* Destination is known and length too. It is known at compile
9895 time there will be no overflow. */
9896 memcpy (&buf[5], "abcde", 5);
9897 /* Destination is known, but the length is not known at compile time.
9898 This will result in __memcpy_chk call that can check for overflow
9900 memcpy (&buf[5], "abcde", n);
9901 /* Destination is known and it is known at compile time there will
9902 be overflow. There will be a warning and __memcpy_chk call that
9903 will abort the program at run time. */
9904 memcpy (&buf[6], "abcde", 5);
9907 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
9908 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
9909 @code{strcat} and @code{strncat}.
9911 There are also checking built-in functions for formatted output functions.
9913 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
9914 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9915 const char *fmt, ...);
9916 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
9918 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
9919 const char *fmt, va_list ap);
9922 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
9923 etc.@: functions and can contain implementation specific flags on what
9924 additional security measures the checking function might take, such as
9925 handling @code{%n} differently.
9927 The @var{os} argument is the object size @var{s} points to, like in the
9928 other built-in functions. There is a small difference in the behavior
9929 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
9930 optimized into the non-checking functions only if @var{flag} is 0, otherwise
9931 the checking function is called with @var{os} argument set to
9934 In addition to this, there are checking built-in functions
9935 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
9936 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
9937 These have just one additional argument, @var{flag}, right before
9938 format string @var{fmt}. If the compiler is able to optimize them to
9939 @code{fputc} etc.@: functions, it does, otherwise the checking function
9940 is called and the @var{flag} argument passed to it.
9942 @node Pointer Bounds Checker builtins
9943 @section Pointer Bounds Checker Built-in Functions
9944 @cindex Pointer Bounds Checker builtins
9945 @findex __builtin___bnd_set_ptr_bounds
9946 @findex __builtin___bnd_narrow_ptr_bounds
9947 @findex __builtin___bnd_copy_ptr_bounds
9948 @findex __builtin___bnd_init_ptr_bounds
9949 @findex __builtin___bnd_null_ptr_bounds
9950 @findex __builtin___bnd_store_ptr_bounds
9951 @findex __builtin___bnd_chk_ptr_lbounds
9952 @findex __builtin___bnd_chk_ptr_ubounds
9953 @findex __builtin___bnd_chk_ptr_bounds
9954 @findex __builtin___bnd_get_ptr_lbound
9955 @findex __builtin___bnd_get_ptr_ubound
9957 GCC provides a set of built-in functions to control Pointer Bounds Checker
9958 instrumentation. Note that all Pointer Bounds Checker builtins can be used
9959 even if you compile with Pointer Bounds Checker off
9960 (@option{-fno-check-pointer-bounds}).
9961 The behavior may differ in such case as documented below.
9963 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
9965 This built-in function returns a new pointer with the value of @var{q}, and
9966 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
9967 Bounds Checker off, the built-in function just returns the first argument.
9970 extern void *__wrap_malloc (size_t n)
9972 void *p = (void *)__real_malloc (n);
9973 if (!p) return __builtin___bnd_null_ptr_bounds (p);
9974 return __builtin___bnd_set_ptr_bounds (p, n);
9980 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
9982 This built-in function returns a new pointer with the value of @var{p}
9983 and associates it with the narrowed bounds formed by the intersection
9984 of bounds associated with @var{q} and the bounds
9985 [@var{p}, @var{p} + @var{size} - 1].
9986 With Pointer Bounds Checker off, the built-in function just returns the first
9990 void init_objects (object *objs, size_t size)
9993 /* Initialize objects one-by-one passing pointers with bounds of
9994 an object, not the full array of objects. */
9995 for (i = 0; i < size; i++)
9996 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
10003 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
10005 This built-in function returns a new pointer with the value of @var{q},
10006 and associates it with the bounds already associated with pointer @var{r}.
10007 With Pointer Bounds Checker off, the built-in function just returns the first
10011 /* Here is a way to get pointer to object's field but
10012 still with the full object's bounds. */
10013 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10019 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10021 This built-in function returns a new pointer with the value of @var{q}, and
10022 associates it with INIT (allowing full memory access) bounds. With Pointer
10023 Bounds Checker off, the built-in function just returns the first argument.
10027 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10029 This built-in function returns a new pointer with the value of @var{q}, and
10030 associates it with NULL (allowing no memory access) bounds. With Pointer
10031 Bounds Checker off, the built-in function just returns the first argument.
10035 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10037 This built-in function stores the bounds associated with pointer @var{ptr_val}
10038 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10039 bounds from legacy code without touching the associated pointer's memory when
10040 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10041 function call is ignored.
10045 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10047 This built-in function checks if the pointer @var{q} is within the lower
10048 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10049 function call is ignored.
10052 extern void *__wrap_memset (void *dst, int c, size_t len)
10056 __builtin___bnd_chk_ptr_lbounds (dst);
10057 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10058 __real_memset (dst, c, len);
10066 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10068 This built-in function checks if the pointer @var{q} is within the upper
10069 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10070 function call is ignored.
10074 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10076 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10077 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10078 off, the built-in function call is ignored.
10081 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10085 __bnd_chk_ptr_bounds (dst, n);
10086 __bnd_chk_ptr_bounds (src, n);
10087 __real_memcpy (dst, src, n);
10095 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10097 This built-in function returns the lower bound associated
10098 with the pointer @var{q}, as a pointer value.
10099 This is useful for debugging using @code{printf}.
10100 With Pointer Bounds Checker off, the built-in function returns 0.
10103 void *lb = __builtin___bnd_get_ptr_lbound (q);
10104 void *ub = __builtin___bnd_get_ptr_ubound (q);
10105 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10110 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10112 This built-in function returns the upper bound (which is a pointer) associated
10113 with the pointer @var{q}. With Pointer Bounds Checker off,
10114 the built-in function returns -1.
10118 @node Cilk Plus Builtins
10119 @section Cilk Plus C/C++ Language Extension Built-in Functions
10121 GCC provides support for the following built-in reduction functions if Cilk Plus
10122 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10125 @item @code{__sec_implicit_index}
10126 @item @code{__sec_reduce}
10127 @item @code{__sec_reduce_add}
10128 @item @code{__sec_reduce_all_nonzero}
10129 @item @code{__sec_reduce_all_zero}
10130 @item @code{__sec_reduce_any_nonzero}
10131 @item @code{__sec_reduce_any_zero}
10132 @item @code{__sec_reduce_max}
10133 @item @code{__sec_reduce_min}
10134 @item @code{__sec_reduce_max_ind}
10135 @item @code{__sec_reduce_min_ind}
10136 @item @code{__sec_reduce_mul}
10137 @item @code{__sec_reduce_mutating}
10140 Further details and examples about these built-in functions are described
10141 in the Cilk Plus language manual which can be found at
10142 @uref{http://www.cilkplus.org}.
10144 @node Other Builtins
10145 @section Other Built-in Functions Provided by GCC
10146 @cindex built-in functions
10147 @findex __builtin_call_with_static_chain
10148 @findex __builtin_fpclassify
10149 @findex __builtin_isfinite
10150 @findex __builtin_isnormal
10151 @findex __builtin_isgreater
10152 @findex __builtin_isgreaterequal
10153 @findex __builtin_isinf_sign
10154 @findex __builtin_isless
10155 @findex __builtin_islessequal
10156 @findex __builtin_islessgreater
10157 @findex __builtin_isunordered
10158 @findex __builtin_powi
10159 @findex __builtin_powif
10160 @findex __builtin_powil
10318 @findex fprintf_unlocked
10320 @findex fputs_unlocked
10428 @findex nexttowardf
10429 @findex nexttowardl
10437 @findex printf_unlocked
10467 @findex signbitd128
10468 @findex significand
10469 @findex significandf
10470 @findex significandl
10498 @findex strncasecmp
10541 GCC provides a large number of built-in functions other than the ones
10542 mentioned above. Some of these are for internal use in the processing
10543 of exceptions or variable-length argument lists and are not
10544 documented here because they may change from time to time; we do not
10545 recommend general use of these functions.
10547 The remaining functions are provided for optimization purposes.
10549 With the exception of built-ins that have library equivalents such as
10550 the standard C library functions discussed below, or that expand to
10551 library calls, GCC built-in functions are always expanded inline and
10552 thus do not have corresponding entry points and their address cannot
10553 be obtained. Attempting to use them in an expression other than
10554 a function call results in a compile-time error.
10556 @opindex fno-builtin
10557 GCC includes built-in versions of many of the functions in the standard
10558 C library. These functions come in two forms: one whose names start with
10559 the @code{__builtin_} prefix, and the other without. Both forms have the
10560 same type (including prototype), the same address (when their address is
10561 taken), and the same meaning as the C library functions even if you specify
10562 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10563 functions are only optimized in certain cases; if they are not optimized in
10564 a particular case, a call to the library function is emitted.
10568 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10569 @option{-std=c99} or @option{-std=c11}), the functions
10570 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10571 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10572 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10573 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10574 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10575 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10576 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10577 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10578 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10579 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10580 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10581 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10582 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10583 @code{significandl}, @code{significand}, @code{sincosf},
10584 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10585 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10586 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10587 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10589 may be handled as built-in functions.
10590 All these functions have corresponding versions
10591 prefixed with @code{__builtin_}, which may be used even in strict C90
10594 The ISO C99 functions
10595 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10596 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10597 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10598 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10599 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10600 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10601 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10602 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10603 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10604 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10605 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10606 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10607 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10608 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10609 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10610 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10611 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10612 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10613 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10614 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10615 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10616 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10617 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10618 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10619 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10620 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10621 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10622 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10623 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10624 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10625 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10626 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10627 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10628 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10629 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10630 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10631 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10632 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10633 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10634 are handled as built-in functions
10635 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10637 There are also built-in versions of the ISO C99 functions
10638 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10639 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10640 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10641 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10642 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10643 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10644 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10645 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10646 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10647 that are recognized in any mode since ISO C90 reserves these names for
10648 the purpose to which ISO C99 puts them. All these functions have
10649 corresponding versions prefixed with @code{__builtin_}.
10651 The ISO C94 functions
10652 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10653 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10654 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10656 are handled as built-in functions
10657 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10659 The ISO C90 functions
10660 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10661 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10662 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10663 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10664 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10665 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10666 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10667 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10668 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10669 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10670 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10671 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10672 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10673 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10674 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10675 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10676 are all recognized as built-in functions unless
10677 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10678 is specified for an individual function). All of these functions have
10679 corresponding versions prefixed with @code{__builtin_}.
10681 GCC provides built-in versions of the ISO C99 floating-point comparison
10682 macros that avoid raising exceptions for unordered operands. They have
10683 the same names as the standard macros ( @code{isgreater},
10684 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10685 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10686 prefixed. We intend for a library implementor to be able to simply
10687 @code{#define} each standard macro to its built-in equivalent.
10688 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10689 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10690 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10691 built-in functions appear both with and without the @code{__builtin_} prefix.
10693 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10695 You can use the built-in function @code{__builtin_types_compatible_p} to
10696 determine whether two types are the same.
10698 This built-in function returns 1 if the unqualified versions of the
10699 types @var{type1} and @var{type2} (which are types, not expressions) are
10700 compatible, 0 otherwise. The result of this built-in function can be
10701 used in integer constant expressions.
10703 This built-in function ignores top level qualifiers (e.g., @code{const},
10704 @code{volatile}). For example, @code{int} is equivalent to @code{const
10707 The type @code{int[]} and @code{int[5]} are compatible. On the other
10708 hand, @code{int} and @code{char *} are not compatible, even if the size
10709 of their types, on the particular architecture are the same. Also, the
10710 amount of pointer indirection is taken into account when determining
10711 similarity. Consequently, @code{short *} is not similar to
10712 @code{short **}. Furthermore, two types that are typedefed are
10713 considered compatible if their underlying types are compatible.
10715 An @code{enum} type is not considered to be compatible with another
10716 @code{enum} type even if both are compatible with the same integer
10717 type; this is what the C standard specifies.
10718 For example, @code{enum @{foo, bar@}} is not similar to
10719 @code{enum @{hot, dog@}}.
10721 You typically use this function in code whose execution varies
10722 depending on the arguments' types. For example:
10727 typeof (x) tmp = (x); \
10728 if (__builtin_types_compatible_p (typeof (x), long double)) \
10729 tmp = foo_long_double (tmp); \
10730 else if (__builtin_types_compatible_p (typeof (x), double)) \
10731 tmp = foo_double (tmp); \
10732 else if (__builtin_types_compatible_p (typeof (x), float)) \
10733 tmp = foo_float (tmp); \
10740 @emph{Note:} This construct is only available for C@.
10744 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
10746 The @var{call_exp} expression must be a function call, and the
10747 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
10748 is passed to the function call in the target's static chain location.
10749 The result of builtin is the result of the function call.
10751 @emph{Note:} This builtin is only available for C@.
10752 This builtin can be used to call Go closures from C.
10756 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
10758 You can use the built-in function @code{__builtin_choose_expr} to
10759 evaluate code depending on the value of a constant expression. This
10760 built-in function returns @var{exp1} if @var{const_exp}, which is an
10761 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
10763 This built-in function is analogous to the @samp{? :} operator in C,
10764 except that the expression returned has its type unaltered by promotion
10765 rules. Also, the built-in function does not evaluate the expression
10766 that is not chosen. For example, if @var{const_exp} evaluates to true,
10767 @var{exp2} is not evaluated even if it has side-effects.
10769 This built-in function can return an lvalue if the chosen argument is an
10772 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
10773 type. Similarly, if @var{exp2} is returned, its return type is the same
10780 __builtin_choose_expr ( \
10781 __builtin_types_compatible_p (typeof (x), double), \
10783 __builtin_choose_expr ( \
10784 __builtin_types_compatible_p (typeof (x), float), \
10786 /* @r{The void expression results in a compile-time error} \
10787 @r{when assigning the result to something.} */ \
10791 @emph{Note:} This construct is only available for C@. Furthermore, the
10792 unused expression (@var{exp1} or @var{exp2} depending on the value of
10793 @var{const_exp}) may still generate syntax errors. This may change in
10798 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
10800 The built-in function @code{__builtin_complex} is provided for use in
10801 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
10802 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
10803 real binary floating-point type, and the result has the corresponding
10804 complex type with real and imaginary parts @var{real} and @var{imag}.
10805 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
10806 infinities, NaNs and negative zeros are involved.
10810 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
10811 You can use the built-in function @code{__builtin_constant_p} to
10812 determine if a value is known to be constant at compile time and hence
10813 that GCC can perform constant-folding on expressions involving that
10814 value. The argument of the function is the value to test. The function
10815 returns the integer 1 if the argument is known to be a compile-time
10816 constant and 0 if it is not known to be a compile-time constant. A
10817 return of 0 does not indicate that the value is @emph{not} a constant,
10818 but merely that GCC cannot prove it is a constant with the specified
10819 value of the @option{-O} option.
10821 You typically use this function in an embedded application where
10822 memory is a critical resource. If you have some complex calculation,
10823 you may want it to be folded if it involves constants, but need to call
10824 a function if it does not. For example:
10827 #define Scale_Value(X) \
10828 (__builtin_constant_p (X) \
10829 ? ((X) * SCALE + OFFSET) : Scale (X))
10832 You may use this built-in function in either a macro or an inline
10833 function. However, if you use it in an inlined function and pass an
10834 argument of the function as the argument to the built-in, GCC
10835 never returns 1 when you call the inline function with a string constant
10836 or compound literal (@pxref{Compound Literals}) and does not return 1
10837 when you pass a constant numeric value to the inline function unless you
10838 specify the @option{-O} option.
10840 You may also use @code{__builtin_constant_p} in initializers for static
10841 data. For instance, you can write
10844 static const int table[] = @{
10845 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
10851 This is an acceptable initializer even if @var{EXPRESSION} is not a
10852 constant expression, including the case where
10853 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
10854 folded to a constant but @var{EXPRESSION} contains operands that are
10855 not otherwise permitted in a static initializer (for example,
10856 @code{0 && foo ()}). GCC must be more conservative about evaluating the
10857 built-in in this case, because it has no opportunity to perform
10861 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
10862 @opindex fprofile-arcs
10863 You may use @code{__builtin_expect} to provide the compiler with
10864 branch prediction information. In general, you should prefer to
10865 use actual profile feedback for this (@option{-fprofile-arcs}), as
10866 programmers are notoriously bad at predicting how their programs
10867 actually perform. However, there are applications in which this
10868 data is hard to collect.
10870 The return value is the value of @var{exp}, which should be an integral
10871 expression. The semantics of the built-in are that it is expected that
10872 @var{exp} == @var{c}. For example:
10875 if (__builtin_expect (x, 0))
10880 indicates that we do not expect to call @code{foo}, since
10881 we expect @code{x} to be zero. Since you are limited to integral
10882 expressions for @var{exp}, you should use constructions such as
10885 if (__builtin_expect (ptr != NULL, 1))
10890 when testing pointer or floating-point values.
10893 @deftypefn {Built-in Function} void __builtin_trap (void)
10894 This function causes the program to exit abnormally. GCC implements
10895 this function by using a target-dependent mechanism (such as
10896 intentionally executing an illegal instruction) or by calling
10897 @code{abort}. The mechanism used may vary from release to release so
10898 you should not rely on any particular implementation.
10901 @deftypefn {Built-in Function} void __builtin_unreachable (void)
10902 If control flow reaches the point of the @code{__builtin_unreachable},
10903 the program is undefined. It is useful in situations where the
10904 compiler cannot deduce the unreachability of the code.
10906 One such case is immediately following an @code{asm} statement that
10907 either never terminates, or one that transfers control elsewhere
10908 and never returns. In this example, without the
10909 @code{__builtin_unreachable}, GCC issues a warning that control
10910 reaches the end of a non-void function. It also generates code
10911 to return after the @code{asm}.
10914 int f (int c, int v)
10922 asm("jmp error_handler");
10923 __builtin_unreachable ();
10929 Because the @code{asm} statement unconditionally transfers control out
10930 of the function, control never reaches the end of the function
10931 body. The @code{__builtin_unreachable} is in fact unreachable and
10932 communicates this fact to the compiler.
10934 Another use for @code{__builtin_unreachable} is following a call a
10935 function that never returns but that is not declared
10936 @code{__attribute__((noreturn))}, as in this example:
10939 void function_that_never_returns (void);
10949 function_that_never_returns ();
10950 __builtin_unreachable ();
10957 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
10958 This function returns its first argument, and allows the compiler
10959 to assume that the returned pointer is at least @var{align} bytes
10960 aligned. This built-in can have either two or three arguments,
10961 if it has three, the third argument should have integer type, and
10962 if it is nonzero means misalignment offset. For example:
10965 void *x = __builtin_assume_aligned (arg, 16);
10969 means that the compiler can assume @code{x}, set to @code{arg}, is at least
10970 16-byte aligned, while:
10973 void *x = __builtin_assume_aligned (arg, 32, 8);
10977 means that the compiler can assume for @code{x}, set to @code{arg}, that
10978 @code{(char *) x - 8} is 32-byte aligned.
10981 @deftypefn {Built-in Function} int __builtin_LINE ()
10982 This function is the equivalent to the preprocessor @code{__LINE__}
10983 macro and returns the line number of the invocation of the built-in.
10984 In a C++ default argument for a function @var{F}, it gets the line number of
10985 the call to @var{F}.
10988 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
10989 This function is the equivalent to the preprocessor @code{__FUNCTION__}
10990 macro and returns the function name the invocation of the built-in is in.
10993 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
10994 This function is the equivalent to the preprocessor @code{__FILE__}
10995 macro and returns the file name the invocation of the built-in is in.
10996 In a C++ default argument for a function @var{F}, it gets the file name of
10997 the call to @var{F}.
11000 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
11001 This function is used to flush the processor's instruction cache for
11002 the region of memory between @var{begin} inclusive and @var{end}
11003 exclusive. Some targets require that the instruction cache be
11004 flushed, after modifying memory containing code, in order to obtain
11005 deterministic behavior.
11007 If the target does not require instruction cache flushes,
11008 @code{__builtin___clear_cache} has no effect. Otherwise either
11009 instructions are emitted in-line to clear the instruction cache or a
11010 call to the @code{__clear_cache} function in libgcc is made.
11013 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
11014 This function is used to minimize cache-miss latency by moving data into
11015 a cache before it is accessed.
11016 You can insert calls to @code{__builtin_prefetch} into code for which
11017 you know addresses of data in memory that is likely to be accessed soon.
11018 If the target supports them, data prefetch instructions are generated.
11019 If the prefetch is done early enough before the access then the data will
11020 be in the cache by the time it is accessed.
11022 The value of @var{addr} is the address of the memory to prefetch.
11023 There are two optional arguments, @var{rw} and @var{locality}.
11024 The value of @var{rw} is a compile-time constant one or zero; one
11025 means that the prefetch is preparing for a write to the memory address
11026 and zero, the default, means that the prefetch is preparing for a read.
11027 The value @var{locality} must be a compile-time constant integer between
11028 zero and three. A value of zero means that the data has no temporal
11029 locality, so it need not be left in the cache after the access. A value
11030 of three means that the data has a high degree of temporal locality and
11031 should be left in all levels of cache possible. Values of one and two
11032 mean, respectively, a low or moderate degree of temporal locality. The
11036 for (i = 0; i < n; i++)
11038 a[i] = a[i] + b[i];
11039 __builtin_prefetch (&a[i+j], 1, 1);
11040 __builtin_prefetch (&b[i+j], 0, 1);
11045 Data prefetch does not generate faults if @var{addr} is invalid, but
11046 the address expression itself must be valid. For example, a prefetch
11047 of @code{p->next} does not fault if @code{p->next} is not a valid
11048 address, but evaluation faults if @code{p} is not a valid address.
11050 If the target does not support data prefetch, the address expression
11051 is evaluated if it includes side effects but no other code is generated
11052 and GCC does not issue a warning.
11055 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11056 Returns a positive infinity, if supported by the floating-point format,
11057 else @code{DBL_MAX}. This function is suitable for implementing the
11058 ISO C macro @code{HUGE_VAL}.
11061 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11062 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11065 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11066 Similar to @code{__builtin_huge_val}, except the return
11067 type is @code{long double}.
11070 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11071 This built-in implements the C99 fpclassify functionality. The first
11072 five int arguments should be the target library's notion of the
11073 possible FP classes and are used for return values. They must be
11074 constant values and they must appear in this order: @code{FP_NAN},
11075 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11076 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11077 to classify. GCC treats the last argument as type-generic, which
11078 means it does not do default promotion from float to double.
11081 @deftypefn {Built-in Function} double __builtin_inf (void)
11082 Similar to @code{__builtin_huge_val}, except a warning is generated
11083 if the target floating-point format does not support infinities.
11086 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11087 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11090 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11091 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11094 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11095 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11098 @deftypefn {Built-in Function} float __builtin_inff (void)
11099 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11100 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11103 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11104 Similar to @code{__builtin_inf}, except the return
11105 type is @code{long double}.
11108 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11109 Similar to @code{isinf}, except the return value is -1 for
11110 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11111 Note while the parameter list is an
11112 ellipsis, this function only accepts exactly one floating-point
11113 argument. GCC treats this parameter as type-generic, which means it
11114 does not do default promotion from float to double.
11117 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11118 This is an implementation of the ISO C99 function @code{nan}.
11120 Since ISO C99 defines this function in terms of @code{strtod}, which we
11121 do not implement, a description of the parsing is in order. The string
11122 is parsed as by @code{strtol}; that is, the base is recognized by
11123 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11124 in the significand such that the least significant bit of the number
11125 is at the least significant bit of the significand. The number is
11126 truncated to fit the significand field provided. The significand is
11127 forced to be a quiet NaN@.
11129 This function, if given a string literal all of which would have been
11130 consumed by @code{strtol}, is evaluated early enough that it is considered a
11131 compile-time constant.
11134 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11135 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11138 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11139 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11142 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11143 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11146 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11147 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11150 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11151 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11154 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11155 Similar to @code{__builtin_nan}, except the significand is forced
11156 to be a signaling NaN@. The @code{nans} function is proposed by
11157 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11160 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11161 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11164 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11165 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11168 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11169 Returns one plus the index of the least significant 1-bit of @var{x}, or
11170 if @var{x} is zero, returns zero.
11173 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11174 Returns the number of leading 0-bits in @var{x}, starting at the most
11175 significant bit position. If @var{x} is 0, the result is undefined.
11178 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11179 Returns the number of trailing 0-bits in @var{x}, starting at the least
11180 significant bit position. If @var{x} is 0, the result is undefined.
11183 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11184 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11185 number of bits following the most significant bit that are identical
11186 to it. There are no special cases for 0 or other values.
11189 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11190 Returns the number of 1-bits in @var{x}.
11193 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11194 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11198 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11199 Similar to @code{__builtin_ffs}, except the argument type is
11203 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11204 Similar to @code{__builtin_clz}, except the argument type is
11205 @code{unsigned long}.
11208 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11209 Similar to @code{__builtin_ctz}, except the argument type is
11210 @code{unsigned long}.
11213 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11214 Similar to @code{__builtin_clrsb}, except the argument type is
11218 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11219 Similar to @code{__builtin_popcount}, except the argument type is
11220 @code{unsigned long}.
11223 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11224 Similar to @code{__builtin_parity}, except the argument type is
11225 @code{unsigned long}.
11228 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11229 Similar to @code{__builtin_ffs}, except the argument type is
11233 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11234 Similar to @code{__builtin_clz}, except the argument type is
11235 @code{unsigned long long}.
11238 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11239 Similar to @code{__builtin_ctz}, except the argument type is
11240 @code{unsigned long long}.
11243 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11244 Similar to @code{__builtin_clrsb}, except the argument type is
11248 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11249 Similar to @code{__builtin_popcount}, except the argument type is
11250 @code{unsigned long long}.
11253 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11254 Similar to @code{__builtin_parity}, except the argument type is
11255 @code{unsigned long long}.
11258 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11259 Returns the first argument raised to the power of the second. Unlike the
11260 @code{pow} function no guarantees about precision and rounding are made.
11263 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11264 Similar to @code{__builtin_powi}, except the argument and return types
11268 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11269 Similar to @code{__builtin_powi}, except the argument and return types
11270 are @code{long double}.
11273 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11274 Returns @var{x} with the order of the bytes reversed; for example,
11275 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11279 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11280 Similar to @code{__builtin_bswap16}, except the argument and return types
11284 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11285 Similar to @code{__builtin_bswap32}, except the argument and return types
11289 @node Target Builtins
11290 @section Built-in Functions Specific to Particular Target Machines
11292 On some target machines, GCC supports many built-in functions specific
11293 to those machines. Generally these generate calls to specific machine
11294 instructions, but allow the compiler to schedule those calls.
11297 * AArch64 Built-in Functions::
11298 * Alpha Built-in Functions::
11299 * Altera Nios II Built-in Functions::
11300 * ARC Built-in Functions::
11301 * ARC SIMD Built-in Functions::
11302 * ARM iWMMXt Built-in Functions::
11303 * ARM C Language Extensions (ACLE)::
11304 * ARM Floating Point Status and Control Intrinsics::
11305 * AVR Built-in Functions::
11306 * Blackfin Built-in Functions::
11307 * FR-V Built-in Functions::
11308 * MIPS DSP Built-in Functions::
11309 * MIPS Paired-Single Support::
11310 * MIPS Loongson Built-in Functions::
11311 * Other MIPS Built-in Functions::
11312 * MSP430 Built-in Functions::
11313 * NDS32 Built-in Functions::
11314 * picoChip Built-in Functions::
11315 * PowerPC Built-in Functions::
11316 * PowerPC AltiVec/VSX Built-in Functions::
11317 * PowerPC Hardware Transactional Memory Built-in Functions::
11318 * RX Built-in Functions::
11319 * S/390 System z Built-in Functions::
11320 * SH Built-in Functions::
11321 * SPARC VIS Built-in Functions::
11322 * SPU Built-in Functions::
11323 * TI C6X Built-in Functions::
11324 * TILE-Gx Built-in Functions::
11325 * TILEPro Built-in Functions::
11326 * x86 Built-in Functions::
11327 * x86 transactional memory intrinsics::
11330 @node AArch64 Built-in Functions
11331 @subsection AArch64 Built-in Functions
11333 These built-in functions are available for the AArch64 family of
11336 unsigned int __builtin_aarch64_get_fpcr ()
11337 void __builtin_aarch64_set_fpcr (unsigned int)
11338 unsigned int __builtin_aarch64_get_fpsr ()
11339 void __builtin_aarch64_set_fpsr (unsigned int)
11342 @node Alpha Built-in Functions
11343 @subsection Alpha Built-in Functions
11345 These built-in functions are available for the Alpha family of
11346 processors, depending on the command-line switches used.
11348 The following built-in functions are always available. They
11349 all generate the machine instruction that is part of the name.
11352 long __builtin_alpha_implver (void)
11353 long __builtin_alpha_rpcc (void)
11354 long __builtin_alpha_amask (long)
11355 long __builtin_alpha_cmpbge (long, long)
11356 long __builtin_alpha_extbl (long, long)
11357 long __builtin_alpha_extwl (long, long)
11358 long __builtin_alpha_extll (long, long)
11359 long __builtin_alpha_extql (long, long)
11360 long __builtin_alpha_extwh (long, long)
11361 long __builtin_alpha_extlh (long, long)
11362 long __builtin_alpha_extqh (long, long)
11363 long __builtin_alpha_insbl (long, long)
11364 long __builtin_alpha_inswl (long, long)
11365 long __builtin_alpha_insll (long, long)
11366 long __builtin_alpha_insql (long, long)
11367 long __builtin_alpha_inswh (long, long)
11368 long __builtin_alpha_inslh (long, long)
11369 long __builtin_alpha_insqh (long, long)
11370 long __builtin_alpha_mskbl (long, long)
11371 long __builtin_alpha_mskwl (long, long)
11372 long __builtin_alpha_mskll (long, long)
11373 long __builtin_alpha_mskql (long, long)
11374 long __builtin_alpha_mskwh (long, long)
11375 long __builtin_alpha_msklh (long, long)
11376 long __builtin_alpha_mskqh (long, long)
11377 long __builtin_alpha_umulh (long, long)
11378 long __builtin_alpha_zap (long, long)
11379 long __builtin_alpha_zapnot (long, long)
11382 The following built-in functions are always with @option{-mmax}
11383 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11384 later. They all generate the machine instruction that is part
11388 long __builtin_alpha_pklb (long)
11389 long __builtin_alpha_pkwb (long)
11390 long __builtin_alpha_unpkbl (long)
11391 long __builtin_alpha_unpkbw (long)
11392 long __builtin_alpha_minub8 (long, long)
11393 long __builtin_alpha_minsb8 (long, long)
11394 long __builtin_alpha_minuw4 (long, long)
11395 long __builtin_alpha_minsw4 (long, long)
11396 long __builtin_alpha_maxub8 (long, long)
11397 long __builtin_alpha_maxsb8 (long, long)
11398 long __builtin_alpha_maxuw4 (long, long)
11399 long __builtin_alpha_maxsw4 (long, long)
11400 long __builtin_alpha_perr (long, long)
11403 The following built-in functions are always with @option{-mcix}
11404 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11405 later. They all generate the machine instruction that is part
11409 long __builtin_alpha_cttz (long)
11410 long __builtin_alpha_ctlz (long)
11411 long __builtin_alpha_ctpop (long)
11414 The following built-in functions are available on systems that use the OSF/1
11415 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11416 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11417 @code{rdval} and @code{wrval}.
11420 void *__builtin_thread_pointer (void)
11421 void __builtin_set_thread_pointer (void *)
11424 @node Altera Nios II Built-in Functions
11425 @subsection Altera Nios II Built-in Functions
11427 These built-in functions are available for the Altera Nios II
11428 family of processors.
11430 The following built-in functions are always available. They
11431 all generate the machine instruction that is part of the name.
11434 int __builtin_ldbio (volatile const void *)
11435 int __builtin_ldbuio (volatile const void *)
11436 int __builtin_ldhio (volatile const void *)
11437 int __builtin_ldhuio (volatile const void *)
11438 int __builtin_ldwio (volatile const void *)
11439 void __builtin_stbio (volatile void *, int)
11440 void __builtin_sthio (volatile void *, int)
11441 void __builtin_stwio (volatile void *, int)
11442 void __builtin_sync (void)
11443 int __builtin_rdctl (int)
11444 int __builtin_rdprs (int, int)
11445 void __builtin_wrctl (int, int)
11446 void __builtin_flushd (volatile void *)
11447 void __builtin_flushda (volatile void *)
11448 int __builtin_wrpie (int);
11449 void __builtin_eni (int);
11450 int __builtin_ldex (volatile const void *)
11451 int __builtin_stex (volatile void *, int)
11452 int __builtin_ldsex (volatile const void *)
11453 int __builtin_stsex (volatile void *, int)
11456 The following built-in functions are always available. They
11457 all generate a Nios II Custom Instruction. The name of the
11458 function represents the types that the function takes and
11459 returns. The letter before the @code{n} is the return type
11460 or void if absent. The @code{n} represents the first parameter
11461 to all the custom instructions, the custom instruction number.
11462 The two letters after the @code{n} represent the up to two
11463 parameters to the function.
11465 The letters represent the following data types:
11468 @code{void} for return type and no parameter for parameter types.
11471 @code{int} for return type and parameter type
11474 @code{float} for return type and parameter type
11477 @code{void *} for return type and parameter type
11481 And the function names are:
11483 void __builtin_custom_n (void)
11484 void __builtin_custom_ni (int)
11485 void __builtin_custom_nf (float)
11486 void __builtin_custom_np (void *)
11487 void __builtin_custom_nii (int, int)
11488 void __builtin_custom_nif (int, float)
11489 void __builtin_custom_nip (int, void *)
11490 void __builtin_custom_nfi (float, int)
11491 void __builtin_custom_nff (float, float)
11492 void __builtin_custom_nfp (float, void *)
11493 void __builtin_custom_npi (void *, int)
11494 void __builtin_custom_npf (void *, float)
11495 void __builtin_custom_npp (void *, void *)
11496 int __builtin_custom_in (void)
11497 int __builtin_custom_ini (int)
11498 int __builtin_custom_inf (float)
11499 int __builtin_custom_inp (void *)
11500 int __builtin_custom_inii (int, int)
11501 int __builtin_custom_inif (int, float)
11502 int __builtin_custom_inip (int, void *)
11503 int __builtin_custom_infi (float, int)
11504 int __builtin_custom_inff (float, float)
11505 int __builtin_custom_infp (float, void *)
11506 int __builtin_custom_inpi (void *, int)
11507 int __builtin_custom_inpf (void *, float)
11508 int __builtin_custom_inpp (void *, void *)
11509 float __builtin_custom_fn (void)
11510 float __builtin_custom_fni (int)
11511 float __builtin_custom_fnf (float)
11512 float __builtin_custom_fnp (void *)
11513 float __builtin_custom_fnii (int, int)
11514 float __builtin_custom_fnif (int, float)
11515 float __builtin_custom_fnip (int, void *)
11516 float __builtin_custom_fnfi (float, int)
11517 float __builtin_custom_fnff (float, float)
11518 float __builtin_custom_fnfp (float, void *)
11519 float __builtin_custom_fnpi (void *, int)
11520 float __builtin_custom_fnpf (void *, float)
11521 float __builtin_custom_fnpp (void *, void *)
11522 void * __builtin_custom_pn (void)
11523 void * __builtin_custom_pni (int)
11524 void * __builtin_custom_pnf (float)
11525 void * __builtin_custom_pnp (void *)
11526 void * __builtin_custom_pnii (int, int)
11527 void * __builtin_custom_pnif (int, float)
11528 void * __builtin_custom_pnip (int, void *)
11529 void * __builtin_custom_pnfi (float, int)
11530 void * __builtin_custom_pnff (float, float)
11531 void * __builtin_custom_pnfp (float, void *)
11532 void * __builtin_custom_pnpi (void *, int)
11533 void * __builtin_custom_pnpf (void *, float)
11534 void * __builtin_custom_pnpp (void *, void *)
11537 @node ARC Built-in Functions
11538 @subsection ARC Built-in Functions
11540 The following built-in functions are provided for ARC targets. The
11541 built-ins generate the corresponding assembly instructions. In the
11542 examples given below, the generated code often requires an operand or
11543 result to be in a register. Where necessary further code will be
11544 generated to ensure this is true, but for brevity this is not
11545 described in each case.
11547 @emph{Note:} Using a built-in to generate an instruction not supported
11548 by a target may cause problems. At present the compiler is not
11549 guaranteed to detect such misuse, and as a result an internal compiler
11550 error may be generated.
11552 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11553 Return 1 if @var{val} is known to have the byte alignment given
11554 by @var{alignval}, otherwise return 0.
11555 Note that this is different from
11557 __alignof__(*(char *)@var{val}) >= alignval
11559 because __alignof__ sees only the type of the dereference, whereas
11560 __builtin_arc_align uses alignment information from the pointer
11561 as well as from the pointed-to type.
11562 The information available will depend on optimization level.
11565 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11572 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11573 The operand is the number of a register to be read. Generates:
11575 mov @var{dest}, r@var{regno}
11577 where the value in @var{dest} will be the result returned from the
11581 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11582 The first operand is the number of a register to be written, the
11583 second operand is a compile time constant to write into that
11584 register. Generates:
11586 mov r@var{regno}, @var{val}
11590 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11591 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11594 divaw @var{dest}, @var{a}, @var{b}
11596 where the value in @var{dest} will be the result returned from the
11600 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11607 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11608 The operand, @var{auxv}, is the address of an auxiliary register and
11609 must be a compile time constant. Generates:
11611 lr @var{dest}, [@var{auxr}]
11613 Where the value in @var{dest} will be the result returned from the
11617 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11618 Only available with @option{-mmul64}. Generates:
11620 mul64 @var{a}, @var{b}
11624 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11625 Only available with @option{-mmul64}. Generates:
11627 mulu64 @var{a}, @var{b}
11631 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11638 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11639 Only valid if the @samp{norm} instruction is available through the
11640 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11643 norm @var{dest}, @var{src}
11645 Where the value in @var{dest} will be the result returned from the
11649 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11650 Only valid if the @samp{normw} instruction is available through the
11651 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11654 normw @var{dest}, @var{src}
11656 Where the value in @var{dest} will be the result returned from the
11660 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11667 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11674 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11675 The first argument, @var{auxv}, is the address of an auxiliary
11676 register, the second argument, @var{val}, is a compile time constant
11677 to be written to the register. Generates:
11679 sr @var{auxr}, [@var{val}]
11683 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11684 Only valid with @option{-mswap}. Generates:
11686 swap @var{dest}, @var{src}
11688 Where the value in @var{dest} will be the result returned from the
11692 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11699 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11700 Only available with @option{-mcpu=ARC700}. Generates:
11706 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
11707 Only available with @option{-mcpu=ARC700}. Generates:
11713 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
11714 Only available with @option{-mcpu=ARC700}. Generates:
11720 The instructions generated by the following builtins are not
11721 considered as candidates for scheduling. They are not moved around by
11722 the compiler during scheduling, and thus can be expected to appear
11723 where they are put in the C code:
11725 __builtin_arc_brk()
11726 __builtin_arc_core_read()
11727 __builtin_arc_core_write()
11728 __builtin_arc_flag()
11730 __builtin_arc_sleep()
11732 __builtin_arc_swi()
11735 @node ARC SIMD Built-in Functions
11736 @subsection ARC SIMD Built-in Functions
11738 SIMD builtins provided by the compiler can be used to generate the
11739 vector instructions. This section describes the available builtins
11740 and their usage in programs. With the @option{-msimd} option, the
11741 compiler provides 128-bit vector types, which can be specified using
11742 the @code{vector_size} attribute. The header file @file{arc-simd.h}
11743 can be included to use the following predefined types:
11745 typedef int __v4si __attribute__((vector_size(16)));
11746 typedef short __v8hi __attribute__((vector_size(16)));
11749 These types can be used to define 128-bit variables. The built-in
11750 functions listed in the following section can be used on these
11751 variables to generate the vector operations.
11753 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
11754 @file{arc-simd.h} also provides equivalent macros called
11755 @code{_@var{someinsn}} that can be used for programming ease and
11756 improved readability. The following macros for DMA control are also
11759 #define _setup_dma_in_channel_reg _vdiwr
11760 #define _setup_dma_out_channel_reg _vdowr
11763 The following is a complete list of all the SIMD built-ins provided
11764 for ARC, grouped by calling signature.
11766 The following take two @code{__v8hi} arguments and return a
11767 @code{__v8hi} result:
11769 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
11770 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
11771 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
11772 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
11773 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
11774 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
11775 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
11776 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
11777 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
11778 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
11779 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
11780 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
11781 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
11782 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
11783 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
11784 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
11785 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
11786 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
11787 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
11788 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
11789 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
11790 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
11791 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
11792 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
11793 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
11794 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
11795 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
11796 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
11797 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
11798 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
11799 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
11800 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
11801 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
11802 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
11803 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
11804 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
11805 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
11806 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
11807 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
11808 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
11809 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
11810 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
11811 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
11812 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
11813 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
11814 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
11815 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
11816 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
11819 The following take one @code{__v8hi} and one @code{int} argument and return a
11820 @code{__v8hi} result:
11823 __v8hi __builtin_arc_vbaddw (__v8hi, int)
11824 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
11825 __v8hi __builtin_arc_vbminw (__v8hi, int)
11826 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
11827 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
11828 __v8hi __builtin_arc_vbmulw (__v8hi, int)
11829 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
11830 __v8hi __builtin_arc_vbsubw (__v8hi, int)
11833 The following take one @code{__v8hi} argument and one @code{int} argument which
11834 must be a 3-bit compile time constant indicating a register number
11835 I0-I7. They return a @code{__v8hi} result.
11837 __v8hi __builtin_arc_vasrw (__v8hi, const int)
11838 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
11839 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
11842 The following take one @code{__v8hi} argument and one @code{int}
11843 argument which must be a 6-bit compile time constant. They return a
11844 @code{__v8hi} result.
11846 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
11847 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
11848 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
11849 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
11850 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
11851 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
11852 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
11855 The following take one @code{__v8hi} argument and one @code{int} argument which
11856 must be a 8-bit compile time constant. They return a @code{__v8hi}
11859 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
11860 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
11861 __v8hi __builtin_arc_vmvw (__v8hi, const int)
11862 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
11865 The following take two @code{int} arguments, the second of which which
11866 must be a 8-bit compile time constant. They return a @code{__v8hi}
11869 __v8hi __builtin_arc_vmovaw (int, const int)
11870 __v8hi __builtin_arc_vmovw (int, const int)
11871 __v8hi __builtin_arc_vmovzw (int, const int)
11874 The following take a single @code{__v8hi} argument and return a
11875 @code{__v8hi} result:
11877 __v8hi __builtin_arc_vabsaw (__v8hi)
11878 __v8hi __builtin_arc_vabsw (__v8hi)
11879 __v8hi __builtin_arc_vaddsuw (__v8hi)
11880 __v8hi __builtin_arc_vexch1 (__v8hi)
11881 __v8hi __builtin_arc_vexch2 (__v8hi)
11882 __v8hi __builtin_arc_vexch4 (__v8hi)
11883 __v8hi __builtin_arc_vsignw (__v8hi)
11884 __v8hi __builtin_arc_vupbaw (__v8hi)
11885 __v8hi __builtin_arc_vupbw (__v8hi)
11886 __v8hi __builtin_arc_vupsbaw (__v8hi)
11887 __v8hi __builtin_arc_vupsbw (__v8hi)
11890 The following take two @code{int} arguments and return no result:
11892 void __builtin_arc_vdirun (int, int)
11893 void __builtin_arc_vdorun (int, int)
11896 The following take two @code{int} arguments and return no result. The
11897 first argument must a 3-bit compile time constant indicating one of
11898 the DR0-DR7 DMA setup channels:
11900 void __builtin_arc_vdiwr (const int, int)
11901 void __builtin_arc_vdowr (const int, int)
11904 The following take an @code{int} argument and return no result:
11906 void __builtin_arc_vendrec (int)
11907 void __builtin_arc_vrec (int)
11908 void __builtin_arc_vrecrun (int)
11909 void __builtin_arc_vrun (int)
11912 The following take a @code{__v8hi} argument and two @code{int}
11913 arguments and return a @code{__v8hi} result. The second argument must
11914 be a 3-bit compile time constants, indicating one the registers I0-I7,
11915 and the third argument must be an 8-bit compile time constant.
11917 @emph{Note:} Although the equivalent hardware instructions do not take
11918 an SIMD register as an operand, these builtins overwrite the relevant
11919 bits of the @code{__v8hi} register provided as the first argument with
11920 the value loaded from the @code{[Ib, u8]} location in the SDM.
11923 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
11924 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
11925 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
11926 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
11929 The following take two @code{int} arguments and return a @code{__v8hi}
11930 result. The first argument must be a 3-bit compile time constants,
11931 indicating one the registers I0-I7, and the second argument must be an
11932 8-bit compile time constant.
11935 __v8hi __builtin_arc_vld128 (const int, const int)
11936 __v8hi __builtin_arc_vld64w (const int, const int)
11939 The following take a @code{__v8hi} argument and two @code{int}
11940 arguments and return no result. The second argument must be a 3-bit
11941 compile time constants, indicating one the registers I0-I7, and the
11942 third argument must be an 8-bit compile time constant.
11945 void __builtin_arc_vst128 (__v8hi, const int, const int)
11946 void __builtin_arc_vst64 (__v8hi, const int, const int)
11949 The following take a @code{__v8hi} argument and three @code{int}
11950 arguments and return no result. The second argument must be a 3-bit
11951 compile-time constant, identifying the 16-bit sub-register to be
11952 stored, the third argument must be a 3-bit compile time constants,
11953 indicating one the registers I0-I7, and the fourth argument must be an
11954 8-bit compile time constant.
11957 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
11958 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
11961 @node ARM iWMMXt Built-in Functions
11962 @subsection ARM iWMMXt Built-in Functions
11964 These built-in functions are available for the ARM family of
11965 processors when the @option{-mcpu=iwmmxt} switch is used:
11968 typedef int v2si __attribute__ ((vector_size (8)));
11969 typedef short v4hi __attribute__ ((vector_size (8)));
11970 typedef char v8qi __attribute__ ((vector_size (8)));
11972 int __builtin_arm_getwcgr0 (void)
11973 void __builtin_arm_setwcgr0 (int)
11974 int __builtin_arm_getwcgr1 (void)
11975 void __builtin_arm_setwcgr1 (int)
11976 int __builtin_arm_getwcgr2 (void)
11977 void __builtin_arm_setwcgr2 (int)
11978 int __builtin_arm_getwcgr3 (void)
11979 void __builtin_arm_setwcgr3 (int)
11980 int __builtin_arm_textrmsb (v8qi, int)
11981 int __builtin_arm_textrmsh (v4hi, int)
11982 int __builtin_arm_textrmsw (v2si, int)
11983 int __builtin_arm_textrmub (v8qi, int)
11984 int __builtin_arm_textrmuh (v4hi, int)
11985 int __builtin_arm_textrmuw (v2si, int)
11986 v8qi __builtin_arm_tinsrb (v8qi, int, int)
11987 v4hi __builtin_arm_tinsrh (v4hi, int, int)
11988 v2si __builtin_arm_tinsrw (v2si, int, int)
11989 long long __builtin_arm_tmia (long long, int, int)
11990 long long __builtin_arm_tmiabb (long long, int, int)
11991 long long __builtin_arm_tmiabt (long long, int, int)
11992 long long __builtin_arm_tmiaph (long long, int, int)
11993 long long __builtin_arm_tmiatb (long long, int, int)
11994 long long __builtin_arm_tmiatt (long long, int, int)
11995 int __builtin_arm_tmovmskb (v8qi)
11996 int __builtin_arm_tmovmskh (v4hi)
11997 int __builtin_arm_tmovmskw (v2si)
11998 long long __builtin_arm_waccb (v8qi)
11999 long long __builtin_arm_wacch (v4hi)
12000 long long __builtin_arm_waccw (v2si)
12001 v8qi __builtin_arm_waddb (v8qi, v8qi)
12002 v8qi __builtin_arm_waddbss (v8qi, v8qi)
12003 v8qi __builtin_arm_waddbus (v8qi, v8qi)
12004 v4hi __builtin_arm_waddh (v4hi, v4hi)
12005 v4hi __builtin_arm_waddhss (v4hi, v4hi)
12006 v4hi __builtin_arm_waddhus (v4hi, v4hi)
12007 v2si __builtin_arm_waddw (v2si, v2si)
12008 v2si __builtin_arm_waddwss (v2si, v2si)
12009 v2si __builtin_arm_waddwus (v2si, v2si)
12010 v8qi __builtin_arm_walign (v8qi, v8qi, int)
12011 long long __builtin_arm_wand(long long, long long)
12012 long long __builtin_arm_wandn (long long, long long)
12013 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
12014 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
12015 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
12016 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
12017 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
12018 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
12019 v2si __builtin_arm_wcmpeqw (v2si, v2si)
12020 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
12021 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
12022 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
12023 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
12024 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
12025 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
12026 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
12027 long long __builtin_arm_wmacsz (v4hi, v4hi)
12028 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
12029 long long __builtin_arm_wmacuz (v4hi, v4hi)
12030 v4hi __builtin_arm_wmadds (v4hi, v4hi)
12031 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
12032 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
12033 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
12034 v2si __builtin_arm_wmaxsw (v2si, v2si)
12035 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
12036 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
12037 v2si __builtin_arm_wmaxuw (v2si, v2si)
12038 v8qi __builtin_arm_wminsb (v8qi, v8qi)
12039 v4hi __builtin_arm_wminsh (v4hi, v4hi)
12040 v2si __builtin_arm_wminsw (v2si, v2si)
12041 v8qi __builtin_arm_wminub (v8qi, v8qi)
12042 v4hi __builtin_arm_wminuh (v4hi, v4hi)
12043 v2si __builtin_arm_wminuw (v2si, v2si)
12044 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
12045 v4hi __builtin_arm_wmulul (v4hi, v4hi)
12046 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12047 long long __builtin_arm_wor (long long, long long)
12048 v2si __builtin_arm_wpackdss (long long, long long)
12049 v2si __builtin_arm_wpackdus (long long, long long)
12050 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12051 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12052 v4hi __builtin_arm_wpackwss (v2si, v2si)
12053 v4hi __builtin_arm_wpackwus (v2si, v2si)
12054 long long __builtin_arm_wrord (long long, long long)
12055 long long __builtin_arm_wrordi (long long, int)
12056 v4hi __builtin_arm_wrorh (v4hi, long long)
12057 v4hi __builtin_arm_wrorhi (v4hi, int)
12058 v2si __builtin_arm_wrorw (v2si, long long)
12059 v2si __builtin_arm_wrorwi (v2si, int)
12060 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12061 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12062 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12063 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12064 v4hi __builtin_arm_wshufh (v4hi, int)
12065 long long __builtin_arm_wslld (long long, long long)
12066 long long __builtin_arm_wslldi (long long, int)
12067 v4hi __builtin_arm_wsllh (v4hi, long long)
12068 v4hi __builtin_arm_wsllhi (v4hi, int)
12069 v2si __builtin_arm_wsllw (v2si, long long)
12070 v2si __builtin_arm_wsllwi (v2si, int)
12071 long long __builtin_arm_wsrad (long long, long long)
12072 long long __builtin_arm_wsradi (long long, int)
12073 v4hi __builtin_arm_wsrah (v4hi, long long)
12074 v4hi __builtin_arm_wsrahi (v4hi, int)
12075 v2si __builtin_arm_wsraw (v2si, long long)
12076 v2si __builtin_arm_wsrawi (v2si, int)
12077 long long __builtin_arm_wsrld (long long, long long)
12078 long long __builtin_arm_wsrldi (long long, int)
12079 v4hi __builtin_arm_wsrlh (v4hi, long long)
12080 v4hi __builtin_arm_wsrlhi (v4hi, int)
12081 v2si __builtin_arm_wsrlw (v2si, long long)
12082 v2si __builtin_arm_wsrlwi (v2si, int)
12083 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12084 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12085 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12086 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12087 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12088 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12089 v2si __builtin_arm_wsubw (v2si, v2si)
12090 v2si __builtin_arm_wsubwss (v2si, v2si)
12091 v2si __builtin_arm_wsubwus (v2si, v2si)
12092 v4hi __builtin_arm_wunpckehsb (v8qi)
12093 v2si __builtin_arm_wunpckehsh (v4hi)
12094 long long __builtin_arm_wunpckehsw (v2si)
12095 v4hi __builtin_arm_wunpckehub (v8qi)
12096 v2si __builtin_arm_wunpckehuh (v4hi)
12097 long long __builtin_arm_wunpckehuw (v2si)
12098 v4hi __builtin_arm_wunpckelsb (v8qi)
12099 v2si __builtin_arm_wunpckelsh (v4hi)
12100 long long __builtin_arm_wunpckelsw (v2si)
12101 v4hi __builtin_arm_wunpckelub (v8qi)
12102 v2si __builtin_arm_wunpckeluh (v4hi)
12103 long long __builtin_arm_wunpckeluw (v2si)
12104 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12105 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12106 v2si __builtin_arm_wunpckihw (v2si, v2si)
12107 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12108 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12109 v2si __builtin_arm_wunpckilw (v2si, v2si)
12110 long long __builtin_arm_wxor (long long, long long)
12111 long long __builtin_arm_wzero ()
12115 @node ARM C Language Extensions (ACLE)
12116 @subsection ARM C Language Extensions (ACLE)
12118 GCC implements extensions for C as described in the ARM C Language
12119 Extensions (ACLE) specification, which can be found at
12120 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12122 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12123 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12124 intrinsics can be found at
12125 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12126 The built-in intrinsics for the Advanced SIMD extension are available when
12129 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12130 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12131 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12132 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12135 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12136 availability of extensions.
12138 @node ARM Floating Point Status and Control Intrinsics
12139 @subsection ARM Floating Point Status and Control Intrinsics
12141 These built-in functions are available for the ARM family of
12142 processors with floating-point unit.
12145 unsigned int __builtin_arm_get_fpscr ()
12146 void __builtin_arm_set_fpscr (unsigned int)
12149 @node AVR Built-in Functions
12150 @subsection AVR Built-in Functions
12152 For each built-in function for AVR, there is an equally named,
12153 uppercase built-in macro defined. That way users can easily query if
12154 or if not a specific built-in is implemented or not. For example, if
12155 @code{__builtin_avr_nop} is available the macro
12156 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12158 The following built-in functions map to the respective machine
12159 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12160 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12161 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12162 as library call if no hardware multiplier is available.
12165 void __builtin_avr_nop (void)
12166 void __builtin_avr_sei (void)
12167 void __builtin_avr_cli (void)
12168 void __builtin_avr_sleep (void)
12169 void __builtin_avr_wdr (void)
12170 unsigned char __builtin_avr_swap (unsigned char)
12171 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12172 int __builtin_avr_fmuls (char, char)
12173 int __builtin_avr_fmulsu (char, unsigned char)
12176 In order to delay execution for a specific number of cycles, GCC
12179 void __builtin_avr_delay_cycles (unsigned long ticks)
12183 @code{ticks} is the number of ticks to delay execution. Note that this
12184 built-in does not take into account the effect of interrupts that
12185 might increase delay time. @code{ticks} must be a compile-time
12186 integer constant; delays with a variable number of cycles are not supported.
12189 char __builtin_avr_flash_segment (const __memx void*)
12193 This built-in takes a byte address to the 24-bit
12194 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12195 the number of the flash segment (the 64 KiB chunk) where the address
12196 points to. Counting starts at @code{0}.
12197 If the address does not point to flash memory, return @code{-1}.
12200 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12204 Insert bits from @var{bits} into @var{val} and return the resulting
12205 value. The nibbles of @var{map} determine how the insertion is
12206 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12208 @item If @var{X} is @code{0xf},
12209 then the @var{n}-th bit of @var{val} is returned unaltered.
12211 @item If X is in the range 0@dots{}7,
12212 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12214 @item If X is in the range 8@dots{}@code{0xe},
12215 then the @var{n}-th result bit is undefined.
12219 One typical use case for this built-in is adjusting input and
12220 output values to non-contiguous port layouts. Some examples:
12223 // same as val, bits is unused
12224 __builtin_avr_insert_bits (0xffffffff, bits, val)
12228 // same as bits, val is unused
12229 __builtin_avr_insert_bits (0x76543210, bits, val)
12233 // same as rotating bits by 4
12234 __builtin_avr_insert_bits (0x32107654, bits, 0)
12238 // high nibble of result is the high nibble of val
12239 // low nibble of result is the low nibble of bits
12240 __builtin_avr_insert_bits (0xffff3210, bits, val)
12244 // reverse the bit order of bits
12245 __builtin_avr_insert_bits (0x01234567, bits, 0)
12248 @node Blackfin Built-in Functions
12249 @subsection Blackfin Built-in Functions
12251 Currently, there are two Blackfin-specific built-in functions. These are
12252 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12253 using inline assembly; by using these built-in functions the compiler can
12254 automatically add workarounds for hardware errata involving these
12255 instructions. These functions are named as follows:
12258 void __builtin_bfin_csync (void)
12259 void __builtin_bfin_ssync (void)
12262 @node FR-V Built-in Functions
12263 @subsection FR-V Built-in Functions
12265 GCC provides many FR-V-specific built-in functions. In general,
12266 these functions are intended to be compatible with those described
12267 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12268 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12269 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12270 pointer rather than by value.
12272 Most of the functions are named after specific FR-V instructions.
12273 Such functions are said to be ``directly mapped'' and are summarized
12274 here in tabular form.
12278 * Directly-mapped Integer Functions::
12279 * Directly-mapped Media Functions::
12280 * Raw read/write Functions::
12281 * Other Built-in Functions::
12284 @node Argument Types
12285 @subsubsection Argument Types
12287 The arguments to the built-in functions can be divided into three groups:
12288 register numbers, compile-time constants and run-time values. In order
12289 to make this classification clear at a glance, the arguments and return
12290 values are given the following pseudo types:
12292 @multitable @columnfractions .20 .30 .15 .35
12293 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12294 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12295 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12296 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12297 @item @code{uw2} @tab @code{unsigned long long} @tab No
12298 @tab an unsigned doubleword
12299 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12300 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12301 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12302 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12305 These pseudo types are not defined by GCC, they are simply a notational
12306 convenience used in this manual.
12308 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12309 and @code{sw2} are evaluated at run time. They correspond to
12310 register operands in the underlying FR-V instructions.
12312 @code{const} arguments represent immediate operands in the underlying
12313 FR-V instructions. They must be compile-time constants.
12315 @code{acc} arguments are evaluated at compile time and specify the number
12316 of an accumulator register. For example, an @code{acc} argument of 2
12317 selects the ACC2 register.
12319 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12320 number of an IACC register. See @pxref{Other Built-in Functions}
12323 @node Directly-mapped Integer Functions
12324 @subsubsection Directly-Mapped Integer Functions
12326 The functions listed below map directly to FR-V I-type instructions.
12328 @multitable @columnfractions .45 .32 .23
12329 @item Function prototype @tab Example usage @tab Assembly output
12330 @item @code{sw1 __ADDSS (sw1, sw1)}
12331 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12332 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12333 @item @code{sw1 __SCAN (sw1, sw1)}
12334 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12335 @tab @code{SCAN @var{a},@var{b},@var{c}}
12336 @item @code{sw1 __SCUTSS (sw1)}
12337 @tab @code{@var{b} = __SCUTSS (@var{a})}
12338 @tab @code{SCUTSS @var{a},@var{b}}
12339 @item @code{sw1 __SLASS (sw1, sw1)}
12340 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12341 @tab @code{SLASS @var{a},@var{b},@var{c}}
12342 @item @code{void __SMASS (sw1, sw1)}
12343 @tab @code{__SMASS (@var{a}, @var{b})}
12344 @tab @code{SMASS @var{a},@var{b}}
12345 @item @code{void __SMSSS (sw1, sw1)}
12346 @tab @code{__SMSSS (@var{a}, @var{b})}
12347 @tab @code{SMSSS @var{a},@var{b}}
12348 @item @code{void __SMU (sw1, sw1)}
12349 @tab @code{__SMU (@var{a}, @var{b})}
12350 @tab @code{SMU @var{a},@var{b}}
12351 @item @code{sw2 __SMUL (sw1, sw1)}
12352 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12353 @tab @code{SMUL @var{a},@var{b},@var{c}}
12354 @item @code{sw1 __SUBSS (sw1, sw1)}
12355 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12356 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12357 @item @code{uw2 __UMUL (uw1, uw1)}
12358 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12359 @tab @code{UMUL @var{a},@var{b},@var{c}}
12362 @node Directly-mapped Media Functions
12363 @subsubsection Directly-Mapped Media Functions
12365 The functions listed below map directly to FR-V M-type instructions.
12367 @multitable @columnfractions .45 .32 .23
12368 @item Function prototype @tab Example usage @tab Assembly output
12369 @item @code{uw1 __MABSHS (sw1)}
12370 @tab @code{@var{b} = __MABSHS (@var{a})}
12371 @tab @code{MABSHS @var{a},@var{b}}
12372 @item @code{void __MADDACCS (acc, acc)}
12373 @tab @code{__MADDACCS (@var{b}, @var{a})}
12374 @tab @code{MADDACCS @var{a},@var{b}}
12375 @item @code{sw1 __MADDHSS (sw1, sw1)}
12376 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12377 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12378 @item @code{uw1 __MADDHUS (uw1, uw1)}
12379 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12380 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12381 @item @code{uw1 __MAND (uw1, uw1)}
12382 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12383 @tab @code{MAND @var{a},@var{b},@var{c}}
12384 @item @code{void __MASACCS (acc, acc)}
12385 @tab @code{__MASACCS (@var{b}, @var{a})}
12386 @tab @code{MASACCS @var{a},@var{b}}
12387 @item @code{uw1 __MAVEH (uw1, uw1)}
12388 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12389 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12390 @item @code{uw2 __MBTOH (uw1)}
12391 @tab @code{@var{b} = __MBTOH (@var{a})}
12392 @tab @code{MBTOH @var{a},@var{b}}
12393 @item @code{void __MBTOHE (uw1 *, uw1)}
12394 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12395 @tab @code{MBTOHE @var{a},@var{b}}
12396 @item @code{void __MCLRACC (acc)}
12397 @tab @code{__MCLRACC (@var{a})}
12398 @tab @code{MCLRACC @var{a}}
12399 @item @code{void __MCLRACCA (void)}
12400 @tab @code{__MCLRACCA ()}
12401 @tab @code{MCLRACCA}
12402 @item @code{uw1 __Mcop1 (uw1, uw1)}
12403 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12404 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12405 @item @code{uw1 __Mcop2 (uw1, uw1)}
12406 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12407 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12408 @item @code{uw1 __MCPLHI (uw2, const)}
12409 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12410 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12411 @item @code{uw1 __MCPLI (uw2, const)}
12412 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12413 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12414 @item @code{void __MCPXIS (acc, sw1, sw1)}
12415 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12416 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12417 @item @code{void __MCPXIU (acc, uw1, uw1)}
12418 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12419 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12420 @item @code{void __MCPXRS (acc, sw1, sw1)}
12421 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12422 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12423 @item @code{void __MCPXRU (acc, uw1, uw1)}
12424 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12425 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12426 @item @code{uw1 __MCUT (acc, uw1)}
12427 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12428 @tab @code{MCUT @var{a},@var{b},@var{c}}
12429 @item @code{uw1 __MCUTSS (acc, sw1)}
12430 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12431 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12432 @item @code{void __MDADDACCS (acc, acc)}
12433 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12434 @tab @code{MDADDACCS @var{a},@var{b}}
12435 @item @code{void __MDASACCS (acc, acc)}
12436 @tab @code{__MDASACCS (@var{b}, @var{a})}
12437 @tab @code{MDASACCS @var{a},@var{b}}
12438 @item @code{uw2 __MDCUTSSI (acc, const)}
12439 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12440 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12441 @item @code{uw2 __MDPACKH (uw2, uw2)}
12442 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12443 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12444 @item @code{uw2 __MDROTLI (uw2, const)}
12445 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12446 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12447 @item @code{void __MDSUBACCS (acc, acc)}
12448 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12449 @tab @code{MDSUBACCS @var{a},@var{b}}
12450 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12451 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12452 @tab @code{MDUNPACKH @var{a},@var{b}}
12453 @item @code{uw2 __MEXPDHD (uw1, const)}
12454 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12455 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12456 @item @code{uw1 __MEXPDHW (uw1, const)}
12457 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12458 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12459 @item @code{uw1 __MHDSETH (uw1, const)}
12460 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12461 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12462 @item @code{sw1 __MHDSETS (const)}
12463 @tab @code{@var{b} = __MHDSETS (@var{a})}
12464 @tab @code{MHDSETS #@var{a},@var{b}}
12465 @item @code{uw1 __MHSETHIH (uw1, const)}
12466 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12467 @tab @code{MHSETHIH #@var{a},@var{b}}
12468 @item @code{sw1 __MHSETHIS (sw1, const)}
12469 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12470 @tab @code{MHSETHIS #@var{a},@var{b}}
12471 @item @code{uw1 __MHSETLOH (uw1, const)}
12472 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12473 @tab @code{MHSETLOH #@var{a},@var{b}}
12474 @item @code{sw1 __MHSETLOS (sw1, const)}
12475 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12476 @tab @code{MHSETLOS #@var{a},@var{b}}
12477 @item @code{uw1 __MHTOB (uw2)}
12478 @tab @code{@var{b} = __MHTOB (@var{a})}
12479 @tab @code{MHTOB @var{a},@var{b}}
12480 @item @code{void __MMACHS (acc, sw1, sw1)}
12481 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12482 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12483 @item @code{void __MMACHU (acc, uw1, uw1)}
12484 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12485 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12486 @item @code{void __MMRDHS (acc, sw1, sw1)}
12487 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12488 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12489 @item @code{void __MMRDHU (acc, uw1, uw1)}
12490 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12491 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12492 @item @code{void __MMULHS (acc, sw1, sw1)}
12493 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12494 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12495 @item @code{void __MMULHU (acc, uw1, uw1)}
12496 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12497 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12498 @item @code{void __MMULXHS (acc, sw1, sw1)}
12499 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12500 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12501 @item @code{void __MMULXHU (acc, uw1, uw1)}
12502 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12503 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12504 @item @code{uw1 __MNOT (uw1)}
12505 @tab @code{@var{b} = __MNOT (@var{a})}
12506 @tab @code{MNOT @var{a},@var{b}}
12507 @item @code{uw1 __MOR (uw1, uw1)}
12508 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12509 @tab @code{MOR @var{a},@var{b},@var{c}}
12510 @item @code{uw1 __MPACKH (uh, uh)}
12511 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12512 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12513 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12514 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12515 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12516 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12517 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12518 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12519 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12520 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12521 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12522 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12523 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12524 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12525 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12526 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12527 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12528 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12529 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12530 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12531 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12532 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12533 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12534 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12535 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12536 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12537 @item @code{void __MQMACHS (acc, sw2, sw2)}
12538 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12539 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12540 @item @code{void __MQMACHU (acc, uw2, uw2)}
12541 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12542 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12543 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12544 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12545 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12546 @item @code{void __MQMULHS (acc, sw2, sw2)}
12547 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12548 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12549 @item @code{void __MQMULHU (acc, uw2, uw2)}
12550 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12551 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12552 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12553 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12554 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12555 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12556 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12557 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12558 @item @code{sw2 __MQSATHS (sw2, sw2)}
12559 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12560 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12561 @item @code{uw2 __MQSLLHI (uw2, int)}
12562 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12563 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12564 @item @code{sw2 __MQSRAHI (sw2, int)}
12565 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12566 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12567 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12568 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12569 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12570 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12571 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12572 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12573 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12574 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12575 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12576 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12577 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12578 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12579 @item @code{uw1 __MRDACC (acc)}
12580 @tab @code{@var{b} = __MRDACC (@var{a})}
12581 @tab @code{MRDACC @var{a},@var{b}}
12582 @item @code{uw1 __MRDACCG (acc)}
12583 @tab @code{@var{b} = __MRDACCG (@var{a})}
12584 @tab @code{MRDACCG @var{a},@var{b}}
12585 @item @code{uw1 __MROTLI (uw1, const)}
12586 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12587 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12588 @item @code{uw1 __MROTRI (uw1, const)}
12589 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12590 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12591 @item @code{sw1 __MSATHS (sw1, sw1)}
12592 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12593 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12594 @item @code{uw1 __MSATHU (uw1, uw1)}
12595 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12596 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12597 @item @code{uw1 __MSLLHI (uw1, const)}
12598 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12599 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12600 @item @code{sw1 __MSRAHI (sw1, const)}
12601 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12602 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12603 @item @code{uw1 __MSRLHI (uw1, const)}
12604 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12605 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12606 @item @code{void __MSUBACCS (acc, acc)}
12607 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12608 @tab @code{MSUBACCS @var{a},@var{b}}
12609 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12610 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12611 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12612 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12613 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12614 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12615 @item @code{void __MTRAP (void)}
12616 @tab @code{__MTRAP ()}
12618 @item @code{uw2 __MUNPACKH (uw1)}
12619 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12620 @tab @code{MUNPACKH @var{a},@var{b}}
12621 @item @code{uw1 __MWCUT (uw2, uw1)}
12622 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12623 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12624 @item @code{void __MWTACC (acc, uw1)}
12625 @tab @code{__MWTACC (@var{b}, @var{a})}
12626 @tab @code{MWTACC @var{a},@var{b}}
12627 @item @code{void __MWTACCG (acc, uw1)}
12628 @tab @code{__MWTACCG (@var{b}, @var{a})}
12629 @tab @code{MWTACCG @var{a},@var{b}}
12630 @item @code{uw1 __MXOR (uw1, uw1)}
12631 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12632 @tab @code{MXOR @var{a},@var{b},@var{c}}
12635 @node Raw read/write Functions
12636 @subsubsection Raw Read/Write Functions
12638 This sections describes built-in functions related to read and write
12639 instructions to access memory. These functions generate
12640 @code{membar} instructions to flush the I/O load and stores where
12641 appropriate, as described in Fujitsu's manual described above.
12645 @item unsigned char __builtin_read8 (void *@var{data})
12646 @item unsigned short __builtin_read16 (void *@var{data})
12647 @item unsigned long __builtin_read32 (void *@var{data})
12648 @item unsigned long long __builtin_read64 (void *@var{data})
12650 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12651 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12652 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12653 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12656 @node Other Built-in Functions
12657 @subsubsection Other Built-in Functions
12659 This section describes built-in functions that are not named after
12660 a specific FR-V instruction.
12663 @item sw2 __IACCreadll (iacc @var{reg})
12664 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12665 for future expansion and must be 0.
12667 @item sw1 __IACCreadl (iacc @var{reg})
12668 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12669 Other values of @var{reg} are rejected as invalid.
12671 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12672 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12673 is reserved for future expansion and must be 0.
12675 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12676 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12677 is 1. Other values of @var{reg} are rejected as invalid.
12679 @item void __data_prefetch0 (const void *@var{x})
12680 Use the @code{dcpl} instruction to load the contents of address @var{x}
12681 into the data cache.
12683 @item void __data_prefetch (const void *@var{x})
12684 Use the @code{nldub} instruction to load the contents of address @var{x}
12685 into the data cache. The instruction is issued in slot I1@.
12688 @node MIPS DSP Built-in Functions
12689 @subsection MIPS DSP Built-in Functions
12691 The MIPS DSP Application-Specific Extension (ASE) includes new
12692 instructions that are designed to improve the performance of DSP and
12693 media applications. It provides instructions that operate on packed
12694 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12696 GCC supports MIPS DSP operations using both the generic
12697 vector extensions (@pxref{Vector Extensions}) and a collection of
12698 MIPS-specific built-in functions. Both kinds of support are
12699 enabled by the @option{-mdsp} command-line option.
12701 Revision 2 of the ASE was introduced in the second half of 2006.
12702 This revision adds extra instructions to the original ASE, but is
12703 otherwise backwards-compatible with it. You can select revision 2
12704 using the command-line option @option{-mdspr2}; this option implies
12707 The SCOUNT and POS bits of the DSP control register are global. The
12708 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
12709 POS bits. During optimization, the compiler does not delete these
12710 instructions and it does not delete calls to functions containing
12711 these instructions.
12713 At present, GCC only provides support for operations on 32-bit
12714 vectors. The vector type associated with 8-bit integer data is
12715 usually called @code{v4i8}, the vector type associated with Q7
12716 is usually called @code{v4q7}, the vector type associated with 16-bit
12717 integer data is usually called @code{v2i16}, and the vector type
12718 associated with Q15 is usually called @code{v2q15}. They can be
12719 defined in C as follows:
12722 typedef signed char v4i8 __attribute__ ((vector_size(4)));
12723 typedef signed char v4q7 __attribute__ ((vector_size(4)));
12724 typedef short v2i16 __attribute__ ((vector_size(4)));
12725 typedef short v2q15 __attribute__ ((vector_size(4)));
12728 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
12729 initialized in the same way as aggregates. For example:
12732 v4i8 a = @{1, 2, 3, 4@};
12734 b = (v4i8) @{5, 6, 7, 8@};
12736 v2q15 c = @{0x0fcb, 0x3a75@};
12738 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
12741 @emph{Note:} The CPU's endianness determines the order in which values
12742 are packed. On little-endian targets, the first value is the least
12743 significant and the last value is the most significant. The opposite
12744 order applies to big-endian targets. For example, the code above
12745 sets the lowest byte of @code{a} to @code{1} on little-endian targets
12746 and @code{4} on big-endian targets.
12748 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
12749 representation. As shown in this example, the integer representation
12750 of a Q7 value can be obtained by multiplying the fractional value by
12751 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
12752 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
12755 The table below lists the @code{v4i8} and @code{v2q15} operations for which
12756 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
12757 and @code{c} and @code{d} are @code{v2q15} values.
12759 @multitable @columnfractions .50 .50
12760 @item C code @tab MIPS instruction
12761 @item @code{a + b} @tab @code{addu.qb}
12762 @item @code{c + d} @tab @code{addq.ph}
12763 @item @code{a - b} @tab @code{subu.qb}
12764 @item @code{c - d} @tab @code{subq.ph}
12767 The table below lists the @code{v2i16} operation for which
12768 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
12769 @code{v2i16} values.
12771 @multitable @columnfractions .50 .50
12772 @item C code @tab MIPS instruction
12773 @item @code{e * f} @tab @code{mul.ph}
12776 It is easier to describe the DSP built-in functions if we first define
12777 the following types:
12782 typedef unsigned int ui32;
12783 typedef long long a64;
12786 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12787 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12788 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12789 @code{long long}, but we use @code{a64} to indicate values that are
12790 placed in one of the four DSP accumulators (@code{$ac0},
12791 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12793 Also, some built-in functions prefer or require immediate numbers as
12794 parameters, because the corresponding DSP instructions accept both immediate
12795 numbers and register operands, or accept immediate numbers only. The
12796 immediate parameters are listed as follows.
12804 imm0_255: 0 to 255.
12805 imm_n32_31: -32 to 31.
12806 imm_n512_511: -512 to 511.
12809 The following built-in functions map directly to a particular MIPS DSP
12810 instruction. Please refer to the architecture specification
12811 for details on what each instruction does.
12814 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12815 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12816 q31 __builtin_mips_addq_s_w (q31, q31)
12817 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12818 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12819 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12820 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12821 q31 __builtin_mips_subq_s_w (q31, q31)
12822 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12823 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12824 i32 __builtin_mips_addsc (i32, i32)
12825 i32 __builtin_mips_addwc (i32, i32)
12826 i32 __builtin_mips_modsub (i32, i32)
12827 i32 __builtin_mips_raddu_w_qb (v4i8)
12828 v2q15 __builtin_mips_absq_s_ph (v2q15)
12829 q31 __builtin_mips_absq_s_w (q31)
12830 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12831 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12832 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12833 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12834 q31 __builtin_mips_preceq_w_phl (v2q15)
12835 q31 __builtin_mips_preceq_w_phr (v2q15)
12836 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12837 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12838 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12839 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12840 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12841 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12842 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12843 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12844 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12845 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12846 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12847 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12848 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12849 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12850 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12851 q31 __builtin_mips_shll_s_w (q31, i32)
12852 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12853 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12854 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12855 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12856 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12857 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12858 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12859 q31 __builtin_mips_shra_r_w (q31, i32)
12860 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12861 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12862 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12863 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12864 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12865 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12866 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12867 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12868 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12869 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12870 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12871 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12872 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12873 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12874 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12875 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12876 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12877 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12878 i32 __builtin_mips_bitrev (i32)
12879 i32 __builtin_mips_insv (i32, i32)
12880 v4i8 __builtin_mips_repl_qb (imm0_255)
12881 v4i8 __builtin_mips_repl_qb (i32)
12882 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12883 v2q15 __builtin_mips_repl_ph (i32)
12884 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12885 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12886 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12887 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12888 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12889 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12890 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12891 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12892 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12893 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12894 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12895 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12896 i32 __builtin_mips_extr_w (a64, imm0_31)
12897 i32 __builtin_mips_extr_w (a64, i32)
12898 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12899 i32 __builtin_mips_extr_s_h (a64, i32)
12900 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12901 i32 __builtin_mips_extr_rs_w (a64, i32)
12902 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12903 i32 __builtin_mips_extr_r_w (a64, i32)
12904 i32 __builtin_mips_extp (a64, imm0_31)
12905 i32 __builtin_mips_extp (a64, i32)
12906 i32 __builtin_mips_extpdp (a64, imm0_31)
12907 i32 __builtin_mips_extpdp (a64, i32)
12908 a64 __builtin_mips_shilo (a64, imm_n32_31)
12909 a64 __builtin_mips_shilo (a64, i32)
12910 a64 __builtin_mips_mthlip (a64, i32)
12911 void __builtin_mips_wrdsp (i32, imm0_63)
12912 i32 __builtin_mips_rddsp (imm0_63)
12913 i32 __builtin_mips_lbux (void *, i32)
12914 i32 __builtin_mips_lhx (void *, i32)
12915 i32 __builtin_mips_lwx (void *, i32)
12916 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12917 i32 __builtin_mips_bposge32 (void)
12918 a64 __builtin_mips_madd (a64, i32, i32);
12919 a64 __builtin_mips_maddu (a64, ui32, ui32);
12920 a64 __builtin_mips_msub (a64, i32, i32);
12921 a64 __builtin_mips_msubu (a64, ui32, ui32);
12922 a64 __builtin_mips_mult (i32, i32);
12923 a64 __builtin_mips_multu (ui32, ui32);
12926 The following built-in functions map directly to a particular MIPS DSP REV 2
12927 instruction. Please refer to the architecture specification
12928 for details on what each instruction does.
12931 v4q7 __builtin_mips_absq_s_qb (v4q7);
12932 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12933 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12934 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12935 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12936 i32 __builtin_mips_append (i32, i32, imm0_31);
12937 i32 __builtin_mips_balign (i32, i32, imm0_3);
12938 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12939 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12940 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12941 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12942 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12943 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12944 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12945 q31 __builtin_mips_mulq_rs_w (q31, q31);
12946 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12947 q31 __builtin_mips_mulq_s_w (q31, q31);
12948 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12949 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12950 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12951 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12952 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12953 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12954 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12955 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12956 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12957 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12958 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12959 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12960 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12961 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12962 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12963 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12964 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12965 q31 __builtin_mips_addqh_w (q31, q31);
12966 q31 __builtin_mips_addqh_r_w (q31, q31);
12967 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12968 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12969 q31 __builtin_mips_subqh_w (q31, q31);
12970 q31 __builtin_mips_subqh_r_w (q31, q31);
12971 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12972 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12973 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12974 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12975 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12976 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12980 @node MIPS Paired-Single Support
12981 @subsection MIPS Paired-Single Support
12983 The MIPS64 architecture includes a number of instructions that
12984 operate on pairs of single-precision floating-point values.
12985 Each pair is packed into a 64-bit floating-point register,
12986 with one element being designated the ``upper half'' and
12987 the other being designated the ``lower half''.
12989 GCC supports paired-single operations using both the generic
12990 vector extensions (@pxref{Vector Extensions}) and a collection of
12991 MIPS-specific built-in functions. Both kinds of support are
12992 enabled by the @option{-mpaired-single} command-line option.
12994 The vector type associated with paired-single values is usually
12995 called @code{v2sf}. It can be defined in C as follows:
12998 typedef float v2sf __attribute__ ((vector_size (8)));
13001 @code{v2sf} values are initialized in the same way as aggregates.
13005 v2sf a = @{1.5, 9.1@};
13008 b = (v2sf) @{e, f@};
13011 @emph{Note:} The CPU's endianness determines which value is stored in
13012 the upper half of a register and which value is stored in the lower half.
13013 On little-endian targets, the first value is the lower one and the second
13014 value is the upper one. The opposite order applies to big-endian targets.
13015 For example, the code above sets the lower half of @code{a} to
13016 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
13018 @node MIPS Loongson Built-in Functions
13019 @subsection MIPS Loongson Built-in Functions
13021 GCC provides intrinsics to access the SIMD instructions provided by the
13022 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
13023 available after inclusion of the @code{loongson.h} header file,
13024 operate on the following 64-bit vector types:
13027 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
13028 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
13029 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
13030 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
13031 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
13032 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
13035 The intrinsics provided are listed below; each is named after the
13036 machine instruction to which it corresponds, with suffixes added as
13037 appropriate to distinguish intrinsics that expand to the same machine
13038 instruction yet have different argument types. Refer to the architecture
13039 documentation for a description of the functionality of each
13043 int16x4_t packsswh (int32x2_t s, int32x2_t t);
13044 int8x8_t packsshb (int16x4_t s, int16x4_t t);
13045 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
13046 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13047 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13048 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13049 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13050 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13051 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13052 uint64_t paddd_u (uint64_t s, uint64_t t);
13053 int64_t paddd_s (int64_t s, int64_t t);
13054 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13055 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13056 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13057 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13058 uint64_t pandn_ud (uint64_t s, uint64_t t);
13059 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13060 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13061 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13062 int64_t pandn_sd (int64_t s, int64_t t);
13063 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13064 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13065 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13066 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13067 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13068 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13069 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13070 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13071 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13072 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13073 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13074 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13075 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13076 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13077 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13078 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13079 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13080 uint16x4_t pextrh_u (uint16x4_t s, int field);
13081 int16x4_t pextrh_s (int16x4_t s, int field);
13082 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13083 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13084 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13085 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13086 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13087 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13088 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13089 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13090 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13091 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13092 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13093 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13094 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13095 uint8x8_t pmovmskb_u (uint8x8_t s);
13096 int8x8_t pmovmskb_s (int8x8_t s);
13097 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13098 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13099 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13100 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13101 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13102 uint16x4_t biadd (uint8x8_t s);
13103 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13104 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13105 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13106 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13107 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13108 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13109 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13110 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13111 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13112 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13113 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13114 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13115 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13116 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13117 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13118 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13119 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13120 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13121 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13122 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13123 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13124 uint64_t psubd_u (uint64_t s, uint64_t t);
13125 int64_t psubd_s (int64_t s, int64_t t);
13126 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13127 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13128 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13129 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13130 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13131 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13132 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13133 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13134 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13135 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13136 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13137 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13138 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13139 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13140 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13141 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13145 * Paired-Single Arithmetic::
13146 * Paired-Single Built-in Functions::
13147 * MIPS-3D Built-in Functions::
13150 @node Paired-Single Arithmetic
13151 @subsubsection Paired-Single Arithmetic
13153 The table below lists the @code{v2sf} operations for which hardware
13154 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13155 values and @code{x} is an integral value.
13157 @multitable @columnfractions .50 .50
13158 @item C code @tab MIPS instruction
13159 @item @code{a + b} @tab @code{add.ps}
13160 @item @code{a - b} @tab @code{sub.ps}
13161 @item @code{-a} @tab @code{neg.ps}
13162 @item @code{a * b} @tab @code{mul.ps}
13163 @item @code{a * b + c} @tab @code{madd.ps}
13164 @item @code{a * b - c} @tab @code{msub.ps}
13165 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13166 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13167 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13170 Note that the multiply-accumulate instructions can be disabled
13171 using the command-line option @code{-mno-fused-madd}.
13173 @node Paired-Single Built-in Functions
13174 @subsubsection Paired-Single Built-in Functions
13176 The following paired-single functions map directly to a particular
13177 MIPS instruction. Please refer to the architecture specification
13178 for details on what each instruction does.
13181 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13182 Pair lower lower (@code{pll.ps}).
13184 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13185 Pair upper lower (@code{pul.ps}).
13187 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13188 Pair lower upper (@code{plu.ps}).
13190 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13191 Pair upper upper (@code{puu.ps}).
13193 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13194 Convert pair to paired single (@code{cvt.ps.s}).
13196 @item float __builtin_mips_cvt_s_pl (v2sf)
13197 Convert pair lower to single (@code{cvt.s.pl}).
13199 @item float __builtin_mips_cvt_s_pu (v2sf)
13200 Convert pair upper to single (@code{cvt.s.pu}).
13202 @item v2sf __builtin_mips_abs_ps (v2sf)
13203 Absolute value (@code{abs.ps}).
13205 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13206 Align variable (@code{alnv.ps}).
13208 @emph{Note:} The value of the third parameter must be 0 or 4
13209 modulo 8, otherwise the result is unpredictable. Please read the
13210 instruction description for details.
13213 The following multi-instruction functions are also available.
13214 In each case, @var{cond} can be any of the 16 floating-point conditions:
13215 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13216 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13217 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13220 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13221 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13222 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13223 @code{movt.ps}/@code{movf.ps}).
13225 The @code{movt} functions return the value @var{x} computed by:
13228 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13229 mov.ps @var{x},@var{c}
13230 movt.ps @var{x},@var{d},@var{cc}
13233 The @code{movf} functions are similar but use @code{movf.ps} instead
13236 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13237 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13238 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13239 @code{bc1t}/@code{bc1f}).
13241 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13242 and return either the upper or lower half of the result. For example:
13246 if (__builtin_mips_upper_c_eq_ps (a, b))
13247 upper_halves_are_equal ();
13249 upper_halves_are_unequal ();
13251 if (__builtin_mips_lower_c_eq_ps (a, b))
13252 lower_halves_are_equal ();
13254 lower_halves_are_unequal ();
13258 @node MIPS-3D Built-in Functions
13259 @subsubsection MIPS-3D Built-in Functions
13261 The MIPS-3D Application-Specific Extension (ASE) includes additional
13262 paired-single instructions that are designed to improve the performance
13263 of 3D graphics operations. Support for these instructions is controlled
13264 by the @option{-mips3d} command-line option.
13266 The functions listed below map directly to a particular MIPS-3D
13267 instruction. Please refer to the architecture specification for
13268 more details on what each instruction does.
13271 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13272 Reduction add (@code{addr.ps}).
13274 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13275 Reduction multiply (@code{mulr.ps}).
13277 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13278 Convert paired single to paired word (@code{cvt.pw.ps}).
13280 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13281 Convert paired word to paired single (@code{cvt.ps.pw}).
13283 @item float __builtin_mips_recip1_s (float)
13284 @itemx double __builtin_mips_recip1_d (double)
13285 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13286 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13288 @item float __builtin_mips_recip2_s (float, float)
13289 @itemx double __builtin_mips_recip2_d (double, double)
13290 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13291 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13293 @item float __builtin_mips_rsqrt1_s (float)
13294 @itemx double __builtin_mips_rsqrt1_d (double)
13295 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13296 Reduced-precision reciprocal square root (sequence step 1)
13297 (@code{rsqrt1.@var{fmt}}).
13299 @item float __builtin_mips_rsqrt2_s (float, float)
13300 @itemx double __builtin_mips_rsqrt2_d (double, double)
13301 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13302 Reduced-precision reciprocal square root (sequence step 2)
13303 (@code{rsqrt2.@var{fmt}}).
13306 The following multi-instruction functions are also available.
13307 In each case, @var{cond} can be any of the 16 floating-point conditions:
13308 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13309 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13310 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13313 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13314 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13315 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13316 @code{bc1t}/@code{bc1f}).
13318 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13319 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13324 if (__builtin_mips_cabs_eq_s (a, b))
13330 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13331 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13332 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13333 @code{bc1t}/@code{bc1f}).
13335 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13336 and return either the upper or lower half of the result. For example:
13340 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13341 upper_halves_are_equal ();
13343 upper_halves_are_unequal ();
13345 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13346 lower_halves_are_equal ();
13348 lower_halves_are_unequal ();
13351 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13352 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13353 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13354 @code{movt.ps}/@code{movf.ps}).
13356 The @code{movt} functions return the value @var{x} computed by:
13359 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13360 mov.ps @var{x},@var{c}
13361 movt.ps @var{x},@var{d},@var{cc}
13364 The @code{movf} functions are similar but use @code{movf.ps} instead
13367 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13368 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13369 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13370 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13371 Comparison of two paired-single values
13372 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13373 @code{bc1any2t}/@code{bc1any2f}).
13375 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13376 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13377 result is true and the @code{all} forms return true if both results are true.
13382 if (__builtin_mips_any_c_eq_ps (a, b))
13387 if (__builtin_mips_all_c_eq_ps (a, b))
13393 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13394 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13395 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13396 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13397 Comparison of four paired-single values
13398 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13399 @code{bc1any4t}/@code{bc1any4f}).
13401 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13402 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13403 The @code{any} forms return true if any of the four results are true
13404 and the @code{all} forms return true if all four results are true.
13409 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13414 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13421 @node Other MIPS Built-in Functions
13422 @subsection Other MIPS Built-in Functions
13424 GCC provides other MIPS-specific built-in functions:
13427 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
13428 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
13429 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
13430 when this function is available.
13432 @item unsigned int __builtin_mips_get_fcsr (void)
13433 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
13434 Get and set the contents of the floating-point control and status register
13435 (FPU control register 31). These functions are only available in hard-float
13436 code but can be called in both MIPS16 and non-MIPS16 contexts.
13438 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
13439 register except the condition codes, which GCC assumes are preserved.
13442 @node MSP430 Built-in Functions
13443 @subsection MSP430 Built-in Functions
13445 GCC provides a couple of special builtin functions to aid in the
13446 writing of interrupt handlers in C.
13449 @item __bic_SR_register_on_exit (int @var{mask})
13450 This clears the indicated bits in the saved copy of the status register
13451 currently residing on the stack. This only works inside interrupt
13452 handlers and the changes to the status register will only take affect
13453 once the handler returns.
13455 @item __bis_SR_register_on_exit (int @var{mask})
13456 This sets the indicated bits in the saved copy of the status register
13457 currently residing on the stack. This only works inside interrupt
13458 handlers and the changes to the status register will only take affect
13459 once the handler returns.
13461 @item __delay_cycles (long long @var{cycles})
13462 This inserts an instruction sequence that takes exactly @var{cycles}
13463 cycles (between 0 and about 17E9) to complete. The inserted sequence
13464 may use jumps, loops, or no-ops, and does not interfere with any other
13465 instructions. Note that @var{cycles} must be a compile-time constant
13466 integer - that is, you must pass a number, not a variable that may be
13467 optimized to a constant later. The number of cycles delayed by this
13471 @node NDS32 Built-in Functions
13472 @subsection NDS32 Built-in Functions
13474 These built-in functions are available for the NDS32 target:
13476 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
13477 Insert an ISYNC instruction into the instruction stream where
13478 @var{addr} is an instruction address for serialization.
13481 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
13482 Insert an ISB instruction into the instruction stream.
13485 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
13486 Return the content of a system register which is mapped by @var{sr}.
13489 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
13490 Return the content of a user space register which is mapped by @var{usr}.
13493 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
13494 Move the @var{value} to a system register which is mapped by @var{sr}.
13497 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
13498 Move the @var{value} to a user space register which is mapped by @var{usr}.
13501 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
13502 Enable global interrupt.
13505 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
13506 Disable global interrupt.
13509 @node picoChip Built-in Functions
13510 @subsection picoChip Built-in Functions
13512 GCC provides an interface to selected machine instructions from the
13513 picoChip instruction set.
13516 @item int __builtin_sbc (int @var{value})
13517 Sign bit count. Return the number of consecutive bits in @var{value}
13518 that have the same value as the sign bit. The result is the number of
13519 leading sign bits minus one, giving the number of redundant sign bits in
13522 @item int __builtin_byteswap (int @var{value})
13523 Byte swap. Return the result of swapping the upper and lower bytes of
13526 @item int __builtin_brev (int @var{value})
13527 Bit reversal. Return the result of reversing the bits in
13528 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
13531 @item int __builtin_adds (int @var{x}, int @var{y})
13532 Saturating addition. Return the result of adding @var{x} and @var{y},
13533 storing the value 32767 if the result overflows.
13535 @item int __builtin_subs (int @var{x}, int @var{y})
13536 Saturating subtraction. Return the result of subtracting @var{y} from
13537 @var{x}, storing the value @minus{}32768 if the result overflows.
13539 @item void __builtin_halt (void)
13540 Halt. The processor stops execution. This built-in is useful for
13541 implementing assertions.
13545 @node PowerPC Built-in Functions
13546 @subsection PowerPC Built-in Functions
13548 The following built-in functions are always available and can be used to
13549 check the PowerPC target platform type:
13551 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
13552 This function is a @code{nop} on the PowerPC platform and is included solely
13553 to maintain API compatibility with the x86 builtins.
13556 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
13557 This function returns a value of @code{1} if the run-time CPU is of type
13558 @var{cpuname} and returns @code{0} otherwise. The following CPU names can be
13563 IBM POWER9 Server CPU.
13565 IBM POWER8 Server CPU.
13567 IBM POWER7 Server CPU.
13569 IBM POWER6 Server CPU (RAW mode).
13571 IBM POWER6 Server CPU (Architected mode).
13573 IBM POWER5+ Server CPU.
13575 IBM POWER5 Server CPU.
13577 IBM 970 Server CPU (ie, Apple G5).
13579 IBM POWER4 Server CPU.
13581 IBM A2 64-bit Embedded CPU
13583 IBM PowerPC 476FP 32-bit Embedded CPU.
13585 IBM PowerPC 464 32-bit Embedded CPU.
13587 PowerPC 440 32-bit Embedded CPU.
13589 PowerPC 405 32-bit Embedded CPU.
13591 IBM PowerPC Cell Broadband Engine Architecture CPU.
13594 Here is an example:
13596 if (__builtin_cpu_is ("power8"))
13598 do_power8 (); // POWER8 specific implementation.
13602 do_generic (); // Generic implementation.
13607 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
13608 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
13609 feature @var{feature} and returns @code{0} otherwise. The following features can be
13614 4xx CPU has a Multiply Accumulator.
13616 CPU has a SIMD/Vector Unit.
13618 CPU supports ISA 2.05 (eg, POWER6)
13620 CPU supports ISA 2.06 (eg, POWER7)
13622 CPU supports ISA 2.07 (eg, POWER8)
13624 CPU supports ISA 3.00 (eg, POWER9)
13626 CPU supports the set of compatible performance monitoring events.
13628 CPU supports the Embedded ISA category.
13630 CPU has a CELL broadband engine.
13632 CPU has a decimal floating point unit.
13634 CPU supports the data stream control register.
13636 CPU supports event base branching.
13638 CPU has a SPE double precision floating point unit.
13640 CPU has a SPE single precision floating point unit.
13642 CPU has a floating point unit.
13644 CPU has hardware transaction memory instructions.
13646 Kernel aborts hardware transactions when a syscall is made.
13648 CPU supports icache snooping capabilities.
13650 CPU supports 128-bit IEEE binary floating point instructions.
13652 CPU supports the integer select instruction.
13654 CPU has a memory management unit.
13656 CPU does not have a timebase (eg, 601 and 403gx).
13658 CPU supports the PA Semi 6T CORE ISA.
13660 CPU supports ISA 2.00 (eg, POWER4)
13662 CPU supports ISA 2.02 (eg, POWER5)
13664 CPU supports ISA 2.03 (eg, POWER5+)
13666 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
13668 CPU supports 32-bit mode execution.
13670 CPU supports the old POWER ISA (eg, 601)
13672 CPU supports 64-bit mode execution.
13674 CPU supports a little-endian mode that uses address swizzling.
13676 CPU support simultaneous multi-threading.
13678 CPU has a signal processing extension unit.
13680 CPU supports the target address register.
13682 CPU supports true little-endian mode.
13684 CPU has unified I/D cache.
13686 CPU supports the vector cryptography instructions.
13688 CPU supports the vector-scalar extension.
13691 Here is an example:
13693 if (__builtin_cpu_supports ("fpu"))
13695 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
13699 dst = __fadd (src1, src2); // Software FP addition function.
13704 These built-in functions are available for the PowerPC family of
13707 float __builtin_recipdivf (float, float);
13708 float __builtin_rsqrtf (float);
13709 double __builtin_recipdiv (double, double);
13710 double __builtin_rsqrt (double);
13711 uint64_t __builtin_ppc_get_timebase ();
13712 unsigned long __builtin_ppc_mftb ();
13713 double __builtin_unpack_longdouble (long double, int);
13714 long double __builtin_pack_longdouble (double, double);
13717 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13718 @code{__builtin_rsqrtf} functions generate multiple instructions to
13719 implement the reciprocal sqrt functionality using reciprocal sqrt
13720 estimate instructions.
13722 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13723 functions generate multiple instructions to implement division using
13724 the reciprocal estimate instructions.
13726 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
13727 functions generate instructions to read the Time Base Register. The
13728 @code{__builtin_ppc_get_timebase} function may generate multiple
13729 instructions and always returns the 64 bits of the Time Base Register.
13730 The @code{__builtin_ppc_mftb} function always generates one instruction and
13731 returns the Time Base Register value as an unsigned long, throwing away
13732 the most significant word on 32-bit environments.
13734 The following built-in functions are available for the PowerPC family
13735 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
13736 or @option{-mpopcntd}):
13738 long __builtin_bpermd (long, long);
13739 int __builtin_divwe (int, int);
13740 int __builtin_divweo (int, int);
13741 unsigned int __builtin_divweu (unsigned int, unsigned int);
13742 unsigned int __builtin_divweuo (unsigned int, unsigned int);
13743 long __builtin_divde (long, long);
13744 long __builtin_divdeo (long, long);
13745 unsigned long __builtin_divdeu (unsigned long, unsigned long);
13746 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
13747 unsigned int cdtbcd (unsigned int);
13748 unsigned int cbcdtd (unsigned int);
13749 unsigned int addg6s (unsigned int, unsigned int);
13752 The @code{__builtin_divde}, @code{__builtin_divdeo},
13753 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
13754 64-bit environment support ISA 2.06 or later.
13756 The following built-in functions are available for the PowerPC family
13757 of processors when hardware decimal floating point
13758 (@option{-mhard-dfp}) is available:
13760 _Decimal64 __builtin_dxex (_Decimal64);
13761 _Decimal128 __builtin_dxexq (_Decimal128);
13762 _Decimal64 __builtin_ddedpd (int, _Decimal64);
13763 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
13764 _Decimal64 __builtin_denbcd (int, _Decimal64);
13765 _Decimal128 __builtin_denbcdq (int, _Decimal128);
13766 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
13767 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
13768 _Decimal64 __builtin_dscli (_Decimal64, int);
13769 _Decimal128 __builtin_dscliq (_Decimal128, int);
13770 _Decimal64 __builtin_dscri (_Decimal64, int);
13771 _Decimal128 __builtin_dscriq (_Decimal128, int);
13772 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
13773 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
13776 The following built-in functions are available for the PowerPC family
13777 of processors when the Vector Scalar (vsx) instruction set is
13780 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
13781 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
13782 unsigned long long);
13785 @node PowerPC AltiVec/VSX Built-in Functions
13786 @subsection PowerPC AltiVec Built-in Functions
13788 GCC provides an interface for the PowerPC family of processors to access
13789 the AltiVec operations described in Motorola's AltiVec Programming
13790 Interface Manual. The interface is made available by including
13791 @code{<altivec.h>} and using @option{-maltivec} and
13792 @option{-mabi=altivec}. The interface supports the following vector
13796 vector unsigned char
13800 vector unsigned short
13801 vector signed short
13805 vector unsigned int
13811 If @option{-mvsx} is used the following additional vector types are
13815 vector unsigned long
13820 The long types are only implemented for 64-bit code generation, and
13821 the long type is only used in the floating point/integer conversion
13824 GCC's implementation of the high-level language interface available from
13825 C and C++ code differs from Motorola's documentation in several ways.
13830 A vector constant is a list of constant expressions within curly braces.
13833 A vector initializer requires no cast if the vector constant is of the
13834 same type as the variable it is initializing.
13837 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13838 vector type is the default signedness of the base type. The default
13839 varies depending on the operating system, so a portable program should
13840 always specify the signedness.
13843 Compiling with @option{-maltivec} adds keywords @code{__vector},
13844 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
13845 @code{bool}. When compiling ISO C, the context-sensitive substitution
13846 of the keywords @code{vector}, @code{pixel} and @code{bool} is
13847 disabled. To use them, you must include @code{<altivec.h>} instead.
13850 GCC allows using a @code{typedef} name as the type specifier for a
13854 For C, overloaded functions are implemented with macros so the following
13858 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13862 Since @code{vec_add} is a macro, the vector constant in the example
13863 is treated as four separate arguments. Wrap the entire argument in
13864 parentheses for this to work.
13867 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
13868 Internally, GCC uses built-in functions to achieve the functionality in
13869 the aforementioned header file, but they are not supported and are
13870 subject to change without notice.
13872 The following interfaces are supported for the generic and specific
13873 AltiVec operations and the AltiVec predicates. In cases where there
13874 is a direct mapping between generic and specific operations, only the
13875 generic names are shown here, although the specific operations can also
13878 Arguments that are documented as @code{const int} require literal
13879 integral values within the range required for that operation.
13882 vector signed char vec_abs (vector signed char);
13883 vector signed short vec_abs (vector signed short);
13884 vector signed int vec_abs (vector signed int);
13885 vector float vec_abs (vector float);
13887 vector signed char vec_abss (vector signed char);
13888 vector signed short vec_abss (vector signed short);
13889 vector signed int vec_abss (vector signed int);
13891 vector signed char vec_add (vector bool char, vector signed char);
13892 vector signed char vec_add (vector signed char, vector bool char);
13893 vector signed char vec_add (vector signed char, vector signed char);
13894 vector unsigned char vec_add (vector bool char, vector unsigned char);
13895 vector unsigned char vec_add (vector unsigned char, vector bool char);
13896 vector unsigned char vec_add (vector unsigned char,
13897 vector unsigned char);
13898 vector signed short vec_add (vector bool short, vector signed short);
13899 vector signed short vec_add (vector signed short, vector bool short);
13900 vector signed short vec_add (vector signed short, vector signed short);
13901 vector unsigned short vec_add (vector bool short,
13902 vector unsigned short);
13903 vector unsigned short vec_add (vector unsigned short,
13904 vector bool short);
13905 vector unsigned short vec_add (vector unsigned short,
13906 vector unsigned short);
13907 vector signed int vec_add (vector bool int, vector signed int);
13908 vector signed int vec_add (vector signed int, vector bool int);
13909 vector signed int vec_add (vector signed int, vector signed int);
13910 vector unsigned int vec_add (vector bool int, vector unsigned int);
13911 vector unsigned int vec_add (vector unsigned int, vector bool int);
13912 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
13913 vector float vec_add (vector float, vector float);
13915 vector float vec_vaddfp (vector float, vector float);
13917 vector signed int vec_vadduwm (vector bool int, vector signed int);
13918 vector signed int vec_vadduwm (vector signed int, vector bool int);
13919 vector signed int vec_vadduwm (vector signed int, vector signed int);
13920 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
13921 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
13922 vector unsigned int vec_vadduwm (vector unsigned int,
13923 vector unsigned int);
13925 vector signed short vec_vadduhm (vector bool short,
13926 vector signed short);
13927 vector signed short vec_vadduhm (vector signed short,
13928 vector bool short);
13929 vector signed short vec_vadduhm (vector signed short,
13930 vector signed short);
13931 vector unsigned short vec_vadduhm (vector bool short,
13932 vector unsigned short);
13933 vector unsigned short vec_vadduhm (vector unsigned short,
13934 vector bool short);
13935 vector unsigned short vec_vadduhm (vector unsigned short,
13936 vector unsigned short);
13938 vector signed char vec_vaddubm (vector bool char, vector signed char);
13939 vector signed char vec_vaddubm (vector signed char, vector bool char);
13940 vector signed char vec_vaddubm (vector signed char, vector signed char);
13941 vector unsigned char vec_vaddubm (vector bool char,
13942 vector unsigned char);
13943 vector unsigned char vec_vaddubm (vector unsigned char,
13945 vector unsigned char vec_vaddubm (vector unsigned char,
13946 vector unsigned char);
13948 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
13950 vector unsigned char vec_adds (vector bool char, vector unsigned char);
13951 vector unsigned char vec_adds (vector unsigned char, vector bool char);
13952 vector unsigned char vec_adds (vector unsigned char,
13953 vector unsigned char);
13954 vector signed char vec_adds (vector bool char, vector signed char);
13955 vector signed char vec_adds (vector signed char, vector bool char);
13956 vector signed char vec_adds (vector signed char, vector signed char);
13957 vector unsigned short vec_adds (vector bool short,
13958 vector unsigned short);
13959 vector unsigned short vec_adds (vector unsigned short,
13960 vector bool short);
13961 vector unsigned short vec_adds (vector unsigned short,
13962 vector unsigned short);
13963 vector signed short vec_adds (vector bool short, vector signed short);
13964 vector signed short vec_adds (vector signed short, vector bool short);
13965 vector signed short vec_adds (vector signed short, vector signed short);
13966 vector unsigned int vec_adds (vector bool int, vector unsigned int);
13967 vector unsigned int vec_adds (vector unsigned int, vector bool int);
13968 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
13969 vector signed int vec_adds (vector bool int, vector signed int);
13970 vector signed int vec_adds (vector signed int, vector bool int);
13971 vector signed int vec_adds (vector signed int, vector signed int);
13973 vector signed int vec_vaddsws (vector bool int, vector signed int);
13974 vector signed int vec_vaddsws (vector signed int, vector bool int);
13975 vector signed int vec_vaddsws (vector signed int, vector signed int);
13977 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
13978 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
13979 vector unsigned int vec_vadduws (vector unsigned int,
13980 vector unsigned int);
13982 vector signed short vec_vaddshs (vector bool short,
13983 vector signed short);
13984 vector signed short vec_vaddshs (vector signed short,
13985 vector bool short);
13986 vector signed short vec_vaddshs (vector signed short,
13987 vector signed short);
13989 vector unsigned short vec_vadduhs (vector bool short,
13990 vector unsigned short);
13991 vector unsigned short vec_vadduhs (vector unsigned short,
13992 vector bool short);
13993 vector unsigned short vec_vadduhs (vector unsigned short,
13994 vector unsigned short);
13996 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13997 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13998 vector signed char vec_vaddsbs (vector signed char, vector signed char);
14000 vector unsigned char vec_vaddubs (vector bool char,
14001 vector unsigned char);
14002 vector unsigned char vec_vaddubs (vector unsigned char,
14004 vector unsigned char vec_vaddubs (vector unsigned char,
14005 vector unsigned char);
14007 vector float vec_and (vector float, vector float);
14008 vector float vec_and (vector float, vector bool int);
14009 vector float vec_and (vector bool int, vector float);
14010 vector bool int vec_and (vector bool int, vector bool int);
14011 vector signed int vec_and (vector bool int, vector signed int);
14012 vector signed int vec_and (vector signed int, vector bool int);
14013 vector signed int vec_and (vector signed int, vector signed int);
14014 vector unsigned int vec_and (vector bool int, vector unsigned int);
14015 vector unsigned int vec_and (vector unsigned int, vector bool int);
14016 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
14017 vector bool short vec_and (vector bool short, vector bool short);
14018 vector signed short vec_and (vector bool short, vector signed short);
14019 vector signed short vec_and (vector signed short, vector bool short);
14020 vector signed short vec_and (vector signed short, vector signed short);
14021 vector unsigned short vec_and (vector bool short,
14022 vector unsigned short);
14023 vector unsigned short vec_and (vector unsigned short,
14024 vector bool short);
14025 vector unsigned short vec_and (vector unsigned short,
14026 vector unsigned short);
14027 vector signed char vec_and (vector bool char, vector signed char);
14028 vector bool char vec_and (vector bool char, vector bool char);
14029 vector signed char vec_and (vector signed char, vector bool char);
14030 vector signed char vec_and (vector signed char, vector signed char);
14031 vector unsigned char vec_and (vector bool char, vector unsigned char);
14032 vector unsigned char vec_and (vector unsigned char, vector bool char);
14033 vector unsigned char vec_and (vector unsigned char,
14034 vector unsigned char);
14036 vector float vec_andc (vector float, vector float);
14037 vector float vec_andc (vector float, vector bool int);
14038 vector float vec_andc (vector bool int, vector float);
14039 vector bool int vec_andc (vector bool int, vector bool int);
14040 vector signed int vec_andc (vector bool int, vector signed int);
14041 vector signed int vec_andc (vector signed int, vector bool int);
14042 vector signed int vec_andc (vector signed int, vector signed int);
14043 vector unsigned int vec_andc (vector bool int, vector unsigned int);
14044 vector unsigned int vec_andc (vector unsigned int, vector bool int);
14045 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
14046 vector bool short vec_andc (vector bool short, vector bool short);
14047 vector signed short vec_andc (vector bool short, vector signed short);
14048 vector signed short vec_andc (vector signed short, vector bool short);
14049 vector signed short vec_andc (vector signed short, vector signed short);
14050 vector unsigned short vec_andc (vector bool short,
14051 vector unsigned short);
14052 vector unsigned short vec_andc (vector unsigned short,
14053 vector bool short);
14054 vector unsigned short vec_andc (vector unsigned short,
14055 vector unsigned short);
14056 vector signed char vec_andc (vector bool char, vector signed char);
14057 vector bool char vec_andc (vector bool char, vector bool char);
14058 vector signed char vec_andc (vector signed char, vector bool char);
14059 vector signed char vec_andc (vector signed char, vector signed char);
14060 vector unsigned char vec_andc (vector bool char, vector unsigned char);
14061 vector unsigned char vec_andc (vector unsigned char, vector bool char);
14062 vector unsigned char vec_andc (vector unsigned char,
14063 vector unsigned char);
14065 vector unsigned char vec_avg (vector unsigned char,
14066 vector unsigned char);
14067 vector signed char vec_avg (vector signed char, vector signed char);
14068 vector unsigned short vec_avg (vector unsigned short,
14069 vector unsigned short);
14070 vector signed short vec_avg (vector signed short, vector signed short);
14071 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
14072 vector signed int vec_avg (vector signed int, vector signed int);
14074 vector signed int vec_vavgsw (vector signed int, vector signed int);
14076 vector unsigned int vec_vavguw (vector unsigned int,
14077 vector unsigned int);
14079 vector signed short vec_vavgsh (vector signed short,
14080 vector signed short);
14082 vector unsigned short vec_vavguh (vector unsigned short,
14083 vector unsigned short);
14085 vector signed char vec_vavgsb (vector signed char, vector signed char);
14087 vector unsigned char vec_vavgub (vector unsigned char,
14088 vector unsigned char);
14090 vector float vec_copysign (vector float);
14092 vector float vec_ceil (vector float);
14094 vector signed int vec_cmpb (vector float, vector float);
14096 vector bool char vec_cmpeq (vector signed char, vector signed char);
14097 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
14098 vector bool short vec_cmpeq (vector signed short, vector signed short);
14099 vector bool short vec_cmpeq (vector unsigned short,
14100 vector unsigned short);
14101 vector bool int vec_cmpeq (vector signed int, vector signed int);
14102 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
14103 vector bool int vec_cmpeq (vector float, vector float);
14105 vector bool int vec_vcmpeqfp (vector float, vector float);
14107 vector bool int vec_vcmpequw (vector signed int, vector signed int);
14108 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
14110 vector bool short vec_vcmpequh (vector signed short,
14111 vector signed short);
14112 vector bool short vec_vcmpequh (vector unsigned short,
14113 vector unsigned short);
14115 vector bool char vec_vcmpequb (vector signed char, vector signed char);
14116 vector bool char vec_vcmpequb (vector unsigned char,
14117 vector unsigned char);
14119 vector bool int vec_cmpge (vector float, vector float);
14121 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
14122 vector bool char vec_cmpgt (vector signed char, vector signed char);
14123 vector bool short vec_cmpgt (vector unsigned short,
14124 vector unsigned short);
14125 vector bool short vec_cmpgt (vector signed short, vector signed short);
14126 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
14127 vector bool int vec_cmpgt (vector signed int, vector signed int);
14128 vector bool int vec_cmpgt (vector float, vector float);
14130 vector bool int vec_vcmpgtfp (vector float, vector float);
14132 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
14134 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
14136 vector bool short vec_vcmpgtsh (vector signed short,
14137 vector signed short);
14139 vector bool short vec_vcmpgtuh (vector unsigned short,
14140 vector unsigned short);
14142 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
14144 vector bool char vec_vcmpgtub (vector unsigned char,
14145 vector unsigned char);
14147 vector bool int vec_cmple (vector float, vector float);
14149 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
14150 vector bool char vec_cmplt (vector signed char, vector signed char);
14151 vector bool short vec_cmplt (vector unsigned short,
14152 vector unsigned short);
14153 vector bool short vec_cmplt (vector signed short, vector signed short);
14154 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
14155 vector bool int vec_cmplt (vector signed int, vector signed int);
14156 vector bool int vec_cmplt (vector float, vector float);
14158 vector float vec_cpsgn (vector float, vector float);
14160 vector float vec_ctf (vector unsigned int, const int);
14161 vector float vec_ctf (vector signed int, const int);
14162 vector double vec_ctf (vector unsigned long, const int);
14163 vector double vec_ctf (vector signed long, const int);
14165 vector float vec_vcfsx (vector signed int, const int);
14167 vector float vec_vcfux (vector unsigned int, const int);
14169 vector signed int vec_cts (vector float, const int);
14170 vector signed long vec_cts (vector double, const int);
14172 vector unsigned int vec_ctu (vector float, const int);
14173 vector unsigned long vec_ctu (vector double, const int);
14175 void vec_dss (const int);
14177 void vec_dssall (void);
14179 void vec_dst (const vector unsigned char *, int, const int);
14180 void vec_dst (const vector signed char *, int, const int);
14181 void vec_dst (const vector bool char *, int, const int);
14182 void vec_dst (const vector unsigned short *, int, const int);
14183 void vec_dst (const vector signed short *, int, const int);
14184 void vec_dst (const vector bool short *, int, const int);
14185 void vec_dst (const vector pixel *, int, const int);
14186 void vec_dst (const vector unsigned int *, int, const int);
14187 void vec_dst (const vector signed int *, int, const int);
14188 void vec_dst (const vector bool int *, int, const int);
14189 void vec_dst (const vector float *, int, const int);
14190 void vec_dst (const unsigned char *, int, const int);
14191 void vec_dst (const signed char *, int, const int);
14192 void vec_dst (const unsigned short *, int, const int);
14193 void vec_dst (const short *, int, const int);
14194 void vec_dst (const unsigned int *, int, const int);
14195 void vec_dst (const int *, int, const int);
14196 void vec_dst (const unsigned long *, int, const int);
14197 void vec_dst (const long *, int, const int);
14198 void vec_dst (const float *, int, const int);
14200 void vec_dstst (const vector unsigned char *, int, const int);
14201 void vec_dstst (const vector signed char *, int, const int);
14202 void vec_dstst (const vector bool char *, int, const int);
14203 void vec_dstst (const vector unsigned short *, int, const int);
14204 void vec_dstst (const vector signed short *, int, const int);
14205 void vec_dstst (const vector bool short *, int, const int);
14206 void vec_dstst (const vector pixel *, int, const int);
14207 void vec_dstst (const vector unsigned int *, int, const int);
14208 void vec_dstst (const vector signed int *, int, const int);
14209 void vec_dstst (const vector bool int *, int, const int);
14210 void vec_dstst (const vector float *, int, const int);
14211 void vec_dstst (const unsigned char *, int, const int);
14212 void vec_dstst (const signed char *, int, const int);
14213 void vec_dstst (const unsigned short *, int, const int);
14214 void vec_dstst (const short *, int, const int);
14215 void vec_dstst (const unsigned int *, int, const int);
14216 void vec_dstst (const int *, int, const int);
14217 void vec_dstst (const unsigned long *, int, const int);
14218 void vec_dstst (const long *, int, const int);
14219 void vec_dstst (const float *, int, const int);
14221 void vec_dststt (const vector unsigned char *, int, const int);
14222 void vec_dststt (const vector signed char *, int, const int);
14223 void vec_dststt (const vector bool char *, int, const int);
14224 void vec_dststt (const vector unsigned short *, int, const int);
14225 void vec_dststt (const vector signed short *, int, const int);
14226 void vec_dststt (const vector bool short *, int, const int);
14227 void vec_dststt (const vector pixel *, int, const int);
14228 void vec_dststt (const vector unsigned int *, int, const int);
14229 void vec_dststt (const vector signed int *, int, const int);
14230 void vec_dststt (const vector bool int *, int, const int);
14231 void vec_dststt (const vector float *, int, const int);
14232 void vec_dststt (const unsigned char *, int, const int);
14233 void vec_dststt (const signed char *, int, const int);
14234 void vec_dststt (const unsigned short *, int, const int);
14235 void vec_dststt (const short *, int, const int);
14236 void vec_dststt (const unsigned int *, int, const int);
14237 void vec_dststt (const int *, int, const int);
14238 void vec_dststt (const unsigned long *, int, const int);
14239 void vec_dststt (const long *, int, const int);
14240 void vec_dststt (const float *, int, const int);
14242 void vec_dstt (const vector unsigned char *, int, const int);
14243 void vec_dstt (const vector signed char *, int, const int);
14244 void vec_dstt (const vector bool char *, int, const int);
14245 void vec_dstt (const vector unsigned short *, int, const int);
14246 void vec_dstt (const vector signed short *, int, const int);
14247 void vec_dstt (const vector bool short *, int, const int);
14248 void vec_dstt (const vector pixel *, int, const int);
14249 void vec_dstt (const vector unsigned int *, int, const int);
14250 void vec_dstt (const vector signed int *, int, const int);
14251 void vec_dstt (const vector bool int *, int, const int);
14252 void vec_dstt (const vector float *, int, const int);
14253 void vec_dstt (const unsigned char *, int, const int);
14254 void vec_dstt (const signed char *, int, const int);
14255 void vec_dstt (const unsigned short *, int, const int);
14256 void vec_dstt (const short *, int, const int);
14257 void vec_dstt (const unsigned int *, int, const int);
14258 void vec_dstt (const int *, int, const int);
14259 void vec_dstt (const unsigned long *, int, const int);
14260 void vec_dstt (const long *, int, const int);
14261 void vec_dstt (const float *, int, const int);
14263 vector float vec_expte (vector float);
14265 vector float vec_floor (vector float);
14267 vector float vec_ld (int, const vector float *);
14268 vector float vec_ld (int, const float *);
14269 vector bool int vec_ld (int, const vector bool int *);
14270 vector signed int vec_ld (int, const vector signed int *);
14271 vector signed int vec_ld (int, const int *);
14272 vector signed int vec_ld (int, const long *);
14273 vector unsigned int vec_ld (int, const vector unsigned int *);
14274 vector unsigned int vec_ld (int, const unsigned int *);
14275 vector unsigned int vec_ld (int, const unsigned long *);
14276 vector bool short vec_ld (int, const vector bool short *);
14277 vector pixel vec_ld (int, const vector pixel *);
14278 vector signed short vec_ld (int, const vector signed short *);
14279 vector signed short vec_ld (int, const short *);
14280 vector unsigned short vec_ld (int, const vector unsigned short *);
14281 vector unsigned short vec_ld (int, const unsigned short *);
14282 vector bool char vec_ld (int, const vector bool char *);
14283 vector signed char vec_ld (int, const vector signed char *);
14284 vector signed char vec_ld (int, const signed char *);
14285 vector unsigned char vec_ld (int, const vector unsigned char *);
14286 vector unsigned char vec_ld (int, const unsigned char *);
14288 vector signed char vec_lde (int, const signed char *);
14289 vector unsigned char vec_lde (int, const unsigned char *);
14290 vector signed short vec_lde (int, const short *);
14291 vector unsigned short vec_lde (int, const unsigned short *);
14292 vector float vec_lde (int, const float *);
14293 vector signed int vec_lde (int, const int *);
14294 vector unsigned int vec_lde (int, const unsigned int *);
14295 vector signed int vec_lde (int, const long *);
14296 vector unsigned int vec_lde (int, const unsigned long *);
14298 vector float vec_lvewx (int, float *);
14299 vector signed int vec_lvewx (int, int *);
14300 vector unsigned int vec_lvewx (int, unsigned int *);
14301 vector signed int vec_lvewx (int, long *);
14302 vector unsigned int vec_lvewx (int, unsigned long *);
14304 vector signed short vec_lvehx (int, short *);
14305 vector unsigned short vec_lvehx (int, unsigned short *);
14307 vector signed char vec_lvebx (int, char *);
14308 vector unsigned char vec_lvebx (int, unsigned char *);
14310 vector float vec_ldl (int, const vector float *);
14311 vector float vec_ldl (int, const float *);
14312 vector bool int vec_ldl (int, const vector bool int *);
14313 vector signed int vec_ldl (int, const vector signed int *);
14314 vector signed int vec_ldl (int, const int *);
14315 vector signed int vec_ldl (int, const long *);
14316 vector unsigned int vec_ldl (int, const vector unsigned int *);
14317 vector unsigned int vec_ldl (int, const unsigned int *);
14318 vector unsigned int vec_ldl (int, const unsigned long *);
14319 vector bool short vec_ldl (int, const vector bool short *);
14320 vector pixel vec_ldl (int, const vector pixel *);
14321 vector signed short vec_ldl (int, const vector signed short *);
14322 vector signed short vec_ldl (int, const short *);
14323 vector unsigned short vec_ldl (int, const vector unsigned short *);
14324 vector unsigned short vec_ldl (int, const unsigned short *);
14325 vector bool char vec_ldl (int, const vector bool char *);
14326 vector signed char vec_ldl (int, const vector signed char *);
14327 vector signed char vec_ldl (int, const signed char *);
14328 vector unsigned char vec_ldl (int, const vector unsigned char *);
14329 vector unsigned char vec_ldl (int, const unsigned char *);
14331 vector float vec_loge (vector float);
14333 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
14334 vector unsigned char vec_lvsl (int, const volatile signed char *);
14335 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
14336 vector unsigned char vec_lvsl (int, const volatile short *);
14337 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
14338 vector unsigned char vec_lvsl (int, const volatile int *);
14339 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
14340 vector unsigned char vec_lvsl (int, const volatile long *);
14341 vector unsigned char vec_lvsl (int, const volatile float *);
14343 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
14344 vector unsigned char vec_lvsr (int, const volatile signed char *);
14345 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
14346 vector unsigned char vec_lvsr (int, const volatile short *);
14347 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
14348 vector unsigned char vec_lvsr (int, const volatile int *);
14349 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
14350 vector unsigned char vec_lvsr (int, const volatile long *);
14351 vector unsigned char vec_lvsr (int, const volatile float *);
14353 vector float vec_madd (vector float, vector float, vector float);
14355 vector signed short vec_madds (vector signed short,
14356 vector signed short,
14357 vector signed short);
14359 vector unsigned char vec_max (vector bool char, vector unsigned char);
14360 vector unsigned char vec_max (vector unsigned char, vector bool char);
14361 vector unsigned char vec_max (vector unsigned char,
14362 vector unsigned char);
14363 vector signed char vec_max (vector bool char, vector signed char);
14364 vector signed char vec_max (vector signed char, vector bool char);
14365 vector signed char vec_max (vector signed char, vector signed char);
14366 vector unsigned short vec_max (vector bool short,
14367 vector unsigned short);
14368 vector unsigned short vec_max (vector unsigned short,
14369 vector bool short);
14370 vector unsigned short vec_max (vector unsigned short,
14371 vector unsigned short);
14372 vector signed short vec_max (vector bool short, vector signed short);
14373 vector signed short vec_max (vector signed short, vector bool short);
14374 vector signed short vec_max (vector signed short, vector signed short);
14375 vector unsigned int vec_max (vector bool int, vector unsigned int);
14376 vector unsigned int vec_max (vector unsigned int, vector bool int);
14377 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
14378 vector signed int vec_max (vector bool int, vector signed int);
14379 vector signed int vec_max (vector signed int, vector bool int);
14380 vector signed int vec_max (vector signed int, vector signed int);
14381 vector float vec_max (vector float, vector float);
14383 vector float vec_vmaxfp (vector float, vector float);
14385 vector signed int vec_vmaxsw (vector bool int, vector signed int);
14386 vector signed int vec_vmaxsw (vector signed int, vector bool int);
14387 vector signed int vec_vmaxsw (vector signed int, vector signed int);
14389 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
14390 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
14391 vector unsigned int vec_vmaxuw (vector unsigned int,
14392 vector unsigned int);
14394 vector signed short vec_vmaxsh (vector bool short, vector signed short);
14395 vector signed short vec_vmaxsh (vector signed short, vector bool short);
14396 vector signed short vec_vmaxsh (vector signed short,
14397 vector signed short);
14399 vector unsigned short vec_vmaxuh (vector bool short,
14400 vector unsigned short);
14401 vector unsigned short vec_vmaxuh (vector unsigned short,
14402 vector bool short);
14403 vector unsigned short vec_vmaxuh (vector unsigned short,
14404 vector unsigned short);
14406 vector signed char vec_vmaxsb (vector bool char, vector signed char);
14407 vector signed char vec_vmaxsb (vector signed char, vector bool char);
14408 vector signed char vec_vmaxsb (vector signed char, vector signed char);
14410 vector unsigned char vec_vmaxub (vector bool char,
14411 vector unsigned char);
14412 vector unsigned char vec_vmaxub (vector unsigned char,
14414 vector unsigned char vec_vmaxub (vector unsigned char,
14415 vector unsigned char);
14417 vector bool char vec_mergeh (vector bool char, vector bool char);
14418 vector signed char vec_mergeh (vector signed char, vector signed char);
14419 vector unsigned char vec_mergeh (vector unsigned char,
14420 vector unsigned char);
14421 vector bool short vec_mergeh (vector bool short, vector bool short);
14422 vector pixel vec_mergeh (vector pixel, vector pixel);
14423 vector signed short vec_mergeh (vector signed short,
14424 vector signed short);
14425 vector unsigned short vec_mergeh (vector unsigned short,
14426 vector unsigned short);
14427 vector float vec_mergeh (vector float, vector float);
14428 vector bool int vec_mergeh (vector bool int, vector bool int);
14429 vector signed int vec_mergeh (vector signed int, vector signed int);
14430 vector unsigned int vec_mergeh (vector unsigned int,
14431 vector unsigned int);
14433 vector float vec_vmrghw (vector float, vector float);
14434 vector bool int vec_vmrghw (vector bool int, vector bool int);
14435 vector signed int vec_vmrghw (vector signed int, vector signed int);
14436 vector unsigned int vec_vmrghw (vector unsigned int,
14437 vector unsigned int);
14439 vector bool short vec_vmrghh (vector bool short, vector bool short);
14440 vector signed short vec_vmrghh (vector signed short,
14441 vector signed short);
14442 vector unsigned short vec_vmrghh (vector unsigned short,
14443 vector unsigned short);
14444 vector pixel vec_vmrghh (vector pixel, vector pixel);
14446 vector bool char vec_vmrghb (vector bool char, vector bool char);
14447 vector signed char vec_vmrghb (vector signed char, vector signed char);
14448 vector unsigned char vec_vmrghb (vector unsigned char,
14449 vector unsigned char);
14451 vector bool char vec_mergel (vector bool char, vector bool char);
14452 vector signed char vec_mergel (vector signed char, vector signed char);
14453 vector unsigned char vec_mergel (vector unsigned char,
14454 vector unsigned char);
14455 vector bool short vec_mergel (vector bool short, vector bool short);
14456 vector pixel vec_mergel (vector pixel, vector pixel);
14457 vector signed short vec_mergel (vector signed short,
14458 vector signed short);
14459 vector unsigned short vec_mergel (vector unsigned short,
14460 vector unsigned short);
14461 vector float vec_mergel (vector float, vector float);
14462 vector bool int vec_mergel (vector bool int, vector bool int);
14463 vector signed int vec_mergel (vector signed int, vector signed int);
14464 vector unsigned int vec_mergel (vector unsigned int,
14465 vector unsigned int);
14467 vector float vec_vmrglw (vector float, vector float);
14468 vector signed int vec_vmrglw (vector signed int, vector signed int);
14469 vector unsigned int vec_vmrglw (vector unsigned int,
14470 vector unsigned int);
14471 vector bool int vec_vmrglw (vector bool int, vector bool int);
14473 vector bool short vec_vmrglh (vector bool short, vector bool short);
14474 vector signed short vec_vmrglh (vector signed short,
14475 vector signed short);
14476 vector unsigned short vec_vmrglh (vector unsigned short,
14477 vector unsigned short);
14478 vector pixel vec_vmrglh (vector pixel, vector pixel);
14480 vector bool char vec_vmrglb (vector bool char, vector bool char);
14481 vector signed char vec_vmrglb (vector signed char, vector signed char);
14482 vector unsigned char vec_vmrglb (vector unsigned char,
14483 vector unsigned char);
14485 vector unsigned short vec_mfvscr (void);
14487 vector unsigned char vec_min (vector bool char, vector unsigned char);
14488 vector unsigned char vec_min (vector unsigned char, vector bool char);
14489 vector unsigned char vec_min (vector unsigned char,
14490 vector unsigned char);
14491 vector signed char vec_min (vector bool char, vector signed char);
14492 vector signed char vec_min (vector signed char, vector bool char);
14493 vector signed char vec_min (vector signed char, vector signed char);
14494 vector unsigned short vec_min (vector bool short,
14495 vector unsigned short);
14496 vector unsigned short vec_min (vector unsigned short,
14497 vector bool short);
14498 vector unsigned short vec_min (vector unsigned short,
14499 vector unsigned short);
14500 vector signed short vec_min (vector bool short, vector signed short);
14501 vector signed short vec_min (vector signed short, vector bool short);
14502 vector signed short vec_min (vector signed short, vector signed short);
14503 vector unsigned int vec_min (vector bool int, vector unsigned int);
14504 vector unsigned int vec_min (vector unsigned int, vector bool int);
14505 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
14506 vector signed int vec_min (vector bool int, vector signed int);
14507 vector signed int vec_min (vector signed int, vector bool int);
14508 vector signed int vec_min (vector signed int, vector signed int);
14509 vector float vec_min (vector float, vector float);
14511 vector float vec_vminfp (vector float, vector float);
14513 vector signed int vec_vminsw (vector bool int, vector signed int);
14514 vector signed int vec_vminsw (vector signed int, vector bool int);
14515 vector signed int vec_vminsw (vector signed int, vector signed int);
14517 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
14518 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
14519 vector unsigned int vec_vminuw (vector unsigned int,
14520 vector unsigned int);
14522 vector signed short vec_vminsh (vector bool short, vector signed short);
14523 vector signed short vec_vminsh (vector signed short, vector bool short);
14524 vector signed short vec_vminsh (vector signed short,
14525 vector signed short);
14527 vector unsigned short vec_vminuh (vector bool short,
14528 vector unsigned short);
14529 vector unsigned short vec_vminuh (vector unsigned short,
14530 vector bool short);
14531 vector unsigned short vec_vminuh (vector unsigned short,
14532 vector unsigned short);
14534 vector signed char vec_vminsb (vector bool char, vector signed char);
14535 vector signed char vec_vminsb (vector signed char, vector bool char);
14536 vector signed char vec_vminsb (vector signed char, vector signed char);
14538 vector unsigned char vec_vminub (vector bool char,
14539 vector unsigned char);
14540 vector unsigned char vec_vminub (vector unsigned char,
14542 vector unsigned char vec_vminub (vector unsigned char,
14543 vector unsigned char);
14545 vector signed short vec_mladd (vector signed short,
14546 vector signed short,
14547 vector signed short);
14548 vector signed short vec_mladd (vector signed short,
14549 vector unsigned short,
14550 vector unsigned short);
14551 vector signed short vec_mladd (vector unsigned short,
14552 vector signed short,
14553 vector signed short);
14554 vector unsigned short vec_mladd (vector unsigned short,
14555 vector unsigned short,
14556 vector unsigned short);
14558 vector signed short vec_mradds (vector signed short,
14559 vector signed short,
14560 vector signed short);
14562 vector unsigned int vec_msum (vector unsigned char,
14563 vector unsigned char,
14564 vector unsigned int);
14565 vector signed int vec_msum (vector signed char,
14566 vector unsigned char,
14567 vector signed int);
14568 vector unsigned int vec_msum (vector unsigned short,
14569 vector unsigned short,
14570 vector unsigned int);
14571 vector signed int vec_msum (vector signed short,
14572 vector signed short,
14573 vector signed int);
14575 vector signed int vec_vmsumshm (vector signed short,
14576 vector signed short,
14577 vector signed int);
14579 vector unsigned int vec_vmsumuhm (vector unsigned short,
14580 vector unsigned short,
14581 vector unsigned int);
14583 vector signed int vec_vmsummbm (vector signed char,
14584 vector unsigned char,
14585 vector signed int);
14587 vector unsigned int vec_vmsumubm (vector unsigned char,
14588 vector unsigned char,
14589 vector unsigned int);
14591 vector unsigned int vec_msums (vector unsigned short,
14592 vector unsigned short,
14593 vector unsigned int);
14594 vector signed int vec_msums (vector signed short,
14595 vector signed short,
14596 vector signed int);
14598 vector signed int vec_vmsumshs (vector signed short,
14599 vector signed short,
14600 vector signed int);
14602 vector unsigned int vec_vmsumuhs (vector unsigned short,
14603 vector unsigned short,
14604 vector unsigned int);
14606 void vec_mtvscr (vector signed int);
14607 void vec_mtvscr (vector unsigned int);
14608 void vec_mtvscr (vector bool int);
14609 void vec_mtvscr (vector signed short);
14610 void vec_mtvscr (vector unsigned short);
14611 void vec_mtvscr (vector bool short);
14612 void vec_mtvscr (vector pixel);
14613 void vec_mtvscr (vector signed char);
14614 void vec_mtvscr (vector unsigned char);
14615 void vec_mtvscr (vector bool char);
14617 vector unsigned short vec_mule (vector unsigned char,
14618 vector unsigned char);
14619 vector signed short vec_mule (vector signed char,
14620 vector signed char);
14621 vector unsigned int vec_mule (vector unsigned short,
14622 vector unsigned short);
14623 vector signed int vec_mule (vector signed short, vector signed short);
14625 vector signed int vec_vmulesh (vector signed short,
14626 vector signed short);
14628 vector unsigned int vec_vmuleuh (vector unsigned short,
14629 vector unsigned short);
14631 vector signed short vec_vmulesb (vector signed char,
14632 vector signed char);
14634 vector unsigned short vec_vmuleub (vector unsigned char,
14635 vector unsigned char);
14637 vector unsigned short vec_mulo (vector unsigned char,
14638 vector unsigned char);
14639 vector signed short vec_mulo (vector signed char, vector signed char);
14640 vector unsigned int vec_mulo (vector unsigned short,
14641 vector unsigned short);
14642 vector signed int vec_mulo (vector signed short, vector signed short);
14644 vector signed int vec_vmulosh (vector signed short,
14645 vector signed short);
14647 vector unsigned int vec_vmulouh (vector unsigned short,
14648 vector unsigned short);
14650 vector signed short vec_vmulosb (vector signed char,
14651 vector signed char);
14653 vector unsigned short vec_vmuloub (vector unsigned char,
14654 vector unsigned char);
14656 vector float vec_nmsub (vector float, vector float, vector float);
14658 vector float vec_nor (vector float, vector float);
14659 vector signed int vec_nor (vector signed int, vector signed int);
14660 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
14661 vector bool int vec_nor (vector bool int, vector bool int);
14662 vector signed short vec_nor (vector signed short, vector signed short);
14663 vector unsigned short vec_nor (vector unsigned short,
14664 vector unsigned short);
14665 vector bool short vec_nor (vector bool short, vector bool short);
14666 vector signed char vec_nor (vector signed char, vector signed char);
14667 vector unsigned char vec_nor (vector unsigned char,
14668 vector unsigned char);
14669 vector bool char vec_nor (vector bool char, vector bool char);
14671 vector float vec_or (vector float, vector float);
14672 vector float vec_or (vector float, vector bool int);
14673 vector float vec_or (vector bool int, vector float);
14674 vector bool int vec_or (vector bool int, vector bool int);
14675 vector signed int vec_or (vector bool int, vector signed int);
14676 vector signed int vec_or (vector signed int, vector bool int);
14677 vector signed int vec_or (vector signed int, vector signed int);
14678 vector unsigned int vec_or (vector bool int, vector unsigned int);
14679 vector unsigned int vec_or (vector unsigned int, vector bool int);
14680 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
14681 vector bool short vec_or (vector bool short, vector bool short);
14682 vector signed short vec_or (vector bool short, vector signed short);
14683 vector signed short vec_or (vector signed short, vector bool short);
14684 vector signed short vec_or (vector signed short, vector signed short);
14685 vector unsigned short vec_or (vector bool short, vector unsigned short);
14686 vector unsigned short vec_or (vector unsigned short, vector bool short);
14687 vector unsigned short vec_or (vector unsigned short,
14688 vector unsigned short);
14689 vector signed char vec_or (vector bool char, vector signed char);
14690 vector bool char vec_or (vector bool char, vector bool char);
14691 vector signed char vec_or (vector signed char, vector bool char);
14692 vector signed char vec_or (vector signed char, vector signed char);
14693 vector unsigned char vec_or (vector bool char, vector unsigned char);
14694 vector unsigned char vec_or (vector unsigned char, vector bool char);
14695 vector unsigned char vec_or (vector unsigned char,
14696 vector unsigned char);
14698 vector signed char vec_pack (vector signed short, vector signed short);
14699 vector unsigned char vec_pack (vector unsigned short,
14700 vector unsigned short);
14701 vector bool char vec_pack (vector bool short, vector bool short);
14702 vector signed short vec_pack (vector signed int, vector signed int);
14703 vector unsigned short vec_pack (vector unsigned int,
14704 vector unsigned int);
14705 vector bool short vec_pack (vector bool int, vector bool int);
14707 vector bool short vec_vpkuwum (vector bool int, vector bool int);
14708 vector signed short vec_vpkuwum (vector signed int, vector signed int);
14709 vector unsigned short vec_vpkuwum (vector unsigned int,
14710 vector unsigned int);
14712 vector bool char vec_vpkuhum (vector bool short, vector bool short);
14713 vector signed char vec_vpkuhum (vector signed short,
14714 vector signed short);
14715 vector unsigned char vec_vpkuhum (vector unsigned short,
14716 vector unsigned short);
14718 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
14720 vector unsigned char vec_packs (vector unsigned short,
14721 vector unsigned short);
14722 vector signed char vec_packs (vector signed short, vector signed short);
14723 vector unsigned short vec_packs (vector unsigned int,
14724 vector unsigned int);
14725 vector signed short vec_packs (vector signed int, vector signed int);
14727 vector signed short vec_vpkswss (vector signed int, vector signed int);
14729 vector unsigned short vec_vpkuwus (vector unsigned int,
14730 vector unsigned int);
14732 vector signed char vec_vpkshss (vector signed short,
14733 vector signed short);
14735 vector unsigned char vec_vpkuhus (vector unsigned short,
14736 vector unsigned short);
14738 vector unsigned char vec_packsu (vector unsigned short,
14739 vector unsigned short);
14740 vector unsigned char vec_packsu (vector signed short,
14741 vector signed short);
14742 vector unsigned short vec_packsu (vector unsigned int,
14743 vector unsigned int);
14744 vector unsigned short vec_packsu (vector signed int, vector signed int);
14746 vector unsigned short vec_vpkswus (vector signed int,
14747 vector signed int);
14749 vector unsigned char vec_vpkshus (vector signed short,
14750 vector signed short);
14752 vector float vec_perm (vector float,
14754 vector unsigned char);
14755 vector signed int vec_perm (vector signed int,
14757 vector unsigned char);
14758 vector unsigned int vec_perm (vector unsigned int,
14759 vector unsigned int,
14760 vector unsigned char);
14761 vector bool int vec_perm (vector bool int,
14763 vector unsigned char);
14764 vector signed short vec_perm (vector signed short,
14765 vector signed short,
14766 vector unsigned char);
14767 vector unsigned short vec_perm (vector unsigned short,
14768 vector unsigned short,
14769 vector unsigned char);
14770 vector bool short vec_perm (vector bool short,
14772 vector unsigned char);
14773 vector pixel vec_perm (vector pixel,
14775 vector unsigned char);
14776 vector signed char vec_perm (vector signed char,
14777 vector signed char,
14778 vector unsigned char);
14779 vector unsigned char vec_perm (vector unsigned char,
14780 vector unsigned char,
14781 vector unsigned char);
14782 vector bool char vec_perm (vector bool char,
14784 vector unsigned char);
14786 vector float vec_re (vector float);
14788 vector signed char vec_rl (vector signed char,
14789 vector unsigned char);
14790 vector unsigned char vec_rl (vector unsigned char,
14791 vector unsigned char);
14792 vector signed short vec_rl (vector signed short, vector unsigned short);
14793 vector unsigned short vec_rl (vector unsigned short,
14794 vector unsigned short);
14795 vector signed int vec_rl (vector signed int, vector unsigned int);
14796 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
14798 vector signed int vec_vrlw (vector signed int, vector unsigned int);
14799 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
14801 vector signed short vec_vrlh (vector signed short,
14802 vector unsigned short);
14803 vector unsigned short vec_vrlh (vector unsigned short,
14804 vector unsigned short);
14806 vector signed char vec_vrlb (vector signed char, vector unsigned char);
14807 vector unsigned char vec_vrlb (vector unsigned char,
14808 vector unsigned char);
14810 vector float vec_round (vector float);
14812 vector float vec_recip (vector float, vector float);
14814 vector float vec_rsqrt (vector float);
14816 vector float vec_rsqrte (vector float);
14818 vector float vec_sel (vector float, vector float, vector bool int);
14819 vector float vec_sel (vector float, vector float, vector unsigned int);
14820 vector signed int vec_sel (vector signed int,
14823 vector signed int vec_sel (vector signed int,
14825 vector unsigned int);
14826 vector unsigned int vec_sel (vector unsigned int,
14827 vector unsigned int,
14829 vector unsigned int vec_sel (vector unsigned int,
14830 vector unsigned int,
14831 vector unsigned int);
14832 vector bool int vec_sel (vector bool int,
14835 vector bool int vec_sel (vector bool int,
14837 vector unsigned int);
14838 vector signed short vec_sel (vector signed short,
14839 vector signed short,
14840 vector bool short);
14841 vector signed short vec_sel (vector signed short,
14842 vector signed short,
14843 vector unsigned short);
14844 vector unsigned short vec_sel (vector unsigned short,
14845 vector unsigned short,
14846 vector bool short);
14847 vector unsigned short vec_sel (vector unsigned short,
14848 vector unsigned short,
14849 vector unsigned short);
14850 vector bool short vec_sel (vector bool short,
14852 vector bool short);
14853 vector bool short vec_sel (vector bool short,
14855 vector unsigned short);
14856 vector signed char vec_sel (vector signed char,
14857 vector signed char,
14859 vector signed char vec_sel (vector signed char,
14860 vector signed char,
14861 vector unsigned char);
14862 vector unsigned char vec_sel (vector unsigned char,
14863 vector unsigned char,
14865 vector unsigned char vec_sel (vector unsigned char,
14866 vector unsigned char,
14867 vector unsigned char);
14868 vector bool char vec_sel (vector bool char,
14871 vector bool char vec_sel (vector bool char,
14873 vector unsigned char);
14875 vector signed char vec_sl (vector signed char,
14876 vector unsigned char);
14877 vector unsigned char vec_sl (vector unsigned char,
14878 vector unsigned char);
14879 vector signed short vec_sl (vector signed short, vector unsigned short);
14880 vector unsigned short vec_sl (vector unsigned short,
14881 vector unsigned short);
14882 vector signed int vec_sl (vector signed int, vector unsigned int);
14883 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
14885 vector signed int vec_vslw (vector signed int, vector unsigned int);
14886 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
14888 vector signed short vec_vslh (vector signed short,
14889 vector unsigned short);
14890 vector unsigned short vec_vslh (vector unsigned short,
14891 vector unsigned short);
14893 vector signed char vec_vslb (vector signed char, vector unsigned char);
14894 vector unsigned char vec_vslb (vector unsigned char,
14895 vector unsigned char);
14897 vector float vec_sld (vector float, vector float, const int);
14898 vector signed int vec_sld (vector signed int,
14901 vector unsigned int vec_sld (vector unsigned int,
14902 vector unsigned int,
14904 vector bool int vec_sld (vector bool int,
14907 vector signed short vec_sld (vector signed short,
14908 vector signed short,
14910 vector unsigned short vec_sld (vector unsigned short,
14911 vector unsigned short,
14913 vector bool short vec_sld (vector bool short,
14916 vector pixel vec_sld (vector pixel,
14919 vector signed char vec_sld (vector signed char,
14920 vector signed char,
14922 vector unsigned char vec_sld (vector unsigned char,
14923 vector unsigned char,
14925 vector bool char vec_sld (vector bool char,
14929 vector signed int vec_sll (vector signed int,
14930 vector unsigned int);
14931 vector signed int vec_sll (vector signed int,
14932 vector unsigned short);
14933 vector signed int vec_sll (vector signed int,
14934 vector unsigned char);
14935 vector unsigned int vec_sll (vector unsigned int,
14936 vector unsigned int);
14937 vector unsigned int vec_sll (vector unsigned int,
14938 vector unsigned short);
14939 vector unsigned int vec_sll (vector unsigned int,
14940 vector unsigned char);
14941 vector bool int vec_sll (vector bool int,
14942 vector unsigned int);
14943 vector bool int vec_sll (vector bool int,
14944 vector unsigned short);
14945 vector bool int vec_sll (vector bool int,
14946 vector unsigned char);
14947 vector signed short vec_sll (vector signed short,
14948 vector unsigned int);
14949 vector signed short vec_sll (vector signed short,
14950 vector unsigned short);
14951 vector signed short vec_sll (vector signed short,
14952 vector unsigned char);
14953 vector unsigned short vec_sll (vector unsigned short,
14954 vector unsigned int);
14955 vector unsigned short vec_sll (vector unsigned short,
14956 vector unsigned short);
14957 vector unsigned short vec_sll (vector unsigned short,
14958 vector unsigned char);
14959 vector bool short vec_sll (vector bool short, vector unsigned int);
14960 vector bool short vec_sll (vector bool short, vector unsigned short);
14961 vector bool short vec_sll (vector bool short, vector unsigned char);
14962 vector pixel vec_sll (vector pixel, vector unsigned int);
14963 vector pixel vec_sll (vector pixel, vector unsigned short);
14964 vector pixel vec_sll (vector pixel, vector unsigned char);
14965 vector signed char vec_sll (vector signed char, vector unsigned int);
14966 vector signed char vec_sll (vector signed char, vector unsigned short);
14967 vector signed char vec_sll (vector signed char, vector unsigned char);
14968 vector unsigned char vec_sll (vector unsigned char,
14969 vector unsigned int);
14970 vector unsigned char vec_sll (vector unsigned char,
14971 vector unsigned short);
14972 vector unsigned char vec_sll (vector unsigned char,
14973 vector unsigned char);
14974 vector bool char vec_sll (vector bool char, vector unsigned int);
14975 vector bool char vec_sll (vector bool char, vector unsigned short);
14976 vector bool char vec_sll (vector bool char, vector unsigned char);
14978 vector float vec_slo (vector float, vector signed char);
14979 vector float vec_slo (vector float, vector unsigned char);
14980 vector signed int vec_slo (vector signed int, vector signed char);
14981 vector signed int vec_slo (vector signed int, vector unsigned char);
14982 vector unsigned int vec_slo (vector unsigned int, vector signed char);
14983 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
14984 vector signed short vec_slo (vector signed short, vector signed char);
14985 vector signed short vec_slo (vector signed short, vector unsigned char);
14986 vector unsigned short vec_slo (vector unsigned short,
14987 vector signed char);
14988 vector unsigned short vec_slo (vector unsigned short,
14989 vector unsigned char);
14990 vector pixel vec_slo (vector pixel, vector signed char);
14991 vector pixel vec_slo (vector pixel, vector unsigned char);
14992 vector signed char vec_slo (vector signed char, vector signed char);
14993 vector signed char vec_slo (vector signed char, vector unsigned char);
14994 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14995 vector unsigned char vec_slo (vector unsigned char,
14996 vector unsigned char);
14998 vector signed char vec_splat (vector signed char, const int);
14999 vector unsigned char vec_splat (vector unsigned char, const int);
15000 vector bool char vec_splat (vector bool char, const int);
15001 vector signed short vec_splat (vector signed short, const int);
15002 vector unsigned short vec_splat (vector unsigned short, const int);
15003 vector bool short vec_splat (vector bool short, const int);
15004 vector pixel vec_splat (vector pixel, const int);
15005 vector float vec_splat (vector float, const int);
15006 vector signed int vec_splat (vector signed int, const int);
15007 vector unsigned int vec_splat (vector unsigned int, const int);
15008 vector bool int vec_splat (vector bool int, const int);
15009 vector signed long vec_splat (vector signed long, const int);
15010 vector unsigned long vec_splat (vector unsigned long, const int);
15012 vector signed char vec_splats (signed char);
15013 vector unsigned char vec_splats (unsigned char);
15014 vector signed short vec_splats (signed short);
15015 vector unsigned short vec_splats (unsigned short);
15016 vector signed int vec_splats (signed int);
15017 vector unsigned int vec_splats (unsigned int);
15018 vector float vec_splats (float);
15020 vector float vec_vspltw (vector float, const int);
15021 vector signed int vec_vspltw (vector signed int, const int);
15022 vector unsigned int vec_vspltw (vector unsigned int, const int);
15023 vector bool int vec_vspltw (vector bool int, const int);
15025 vector bool short vec_vsplth (vector bool short, const int);
15026 vector signed short vec_vsplth (vector signed short, const int);
15027 vector unsigned short vec_vsplth (vector unsigned short, const int);
15028 vector pixel vec_vsplth (vector pixel, const int);
15030 vector signed char vec_vspltb (vector signed char, const int);
15031 vector unsigned char vec_vspltb (vector unsigned char, const int);
15032 vector bool char vec_vspltb (vector bool char, const int);
15034 vector signed char vec_splat_s8 (const int);
15036 vector signed short vec_splat_s16 (const int);
15038 vector signed int vec_splat_s32 (const int);
15040 vector unsigned char vec_splat_u8 (const int);
15042 vector unsigned short vec_splat_u16 (const int);
15044 vector unsigned int vec_splat_u32 (const int);
15046 vector signed char vec_sr (vector signed char, vector unsigned char);
15047 vector unsigned char vec_sr (vector unsigned char,
15048 vector unsigned char);
15049 vector signed short vec_sr (vector signed short,
15050 vector unsigned short);
15051 vector unsigned short vec_sr (vector unsigned short,
15052 vector unsigned short);
15053 vector signed int vec_sr (vector signed int, vector unsigned int);
15054 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
15056 vector signed int vec_vsrw (vector signed int, vector unsigned int);
15057 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
15059 vector signed short vec_vsrh (vector signed short,
15060 vector unsigned short);
15061 vector unsigned short vec_vsrh (vector unsigned short,
15062 vector unsigned short);
15064 vector signed char vec_vsrb (vector signed char, vector unsigned char);
15065 vector unsigned char vec_vsrb (vector unsigned char,
15066 vector unsigned char);
15068 vector signed char vec_sra (vector signed char, vector unsigned char);
15069 vector unsigned char vec_sra (vector unsigned char,
15070 vector unsigned char);
15071 vector signed short vec_sra (vector signed short,
15072 vector unsigned short);
15073 vector unsigned short vec_sra (vector unsigned short,
15074 vector unsigned short);
15075 vector signed int vec_sra (vector signed int, vector unsigned int);
15076 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
15078 vector signed int vec_vsraw (vector signed int, vector unsigned int);
15079 vector unsigned int vec_vsraw (vector unsigned int,
15080 vector unsigned int);
15082 vector signed short vec_vsrah (vector signed short,
15083 vector unsigned short);
15084 vector unsigned short vec_vsrah (vector unsigned short,
15085 vector unsigned short);
15087 vector signed char vec_vsrab (vector signed char, vector unsigned char);
15088 vector unsigned char vec_vsrab (vector unsigned char,
15089 vector unsigned char);
15091 vector signed int vec_srl (vector signed int, vector unsigned int);
15092 vector signed int vec_srl (vector signed int, vector unsigned short);
15093 vector signed int vec_srl (vector signed int, vector unsigned char);
15094 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
15095 vector unsigned int vec_srl (vector unsigned int,
15096 vector unsigned short);
15097 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
15098 vector bool int vec_srl (vector bool int, vector unsigned int);
15099 vector bool int vec_srl (vector bool int, vector unsigned short);
15100 vector bool int vec_srl (vector bool int, vector unsigned char);
15101 vector signed short vec_srl (vector signed short, vector unsigned int);
15102 vector signed short vec_srl (vector signed short,
15103 vector unsigned short);
15104 vector signed short vec_srl (vector signed short, vector unsigned char);
15105 vector unsigned short vec_srl (vector unsigned short,
15106 vector unsigned int);
15107 vector unsigned short vec_srl (vector unsigned short,
15108 vector unsigned short);
15109 vector unsigned short vec_srl (vector unsigned short,
15110 vector unsigned char);
15111 vector bool short vec_srl (vector bool short, vector unsigned int);
15112 vector bool short vec_srl (vector bool short, vector unsigned short);
15113 vector bool short vec_srl (vector bool short, vector unsigned char);
15114 vector pixel vec_srl (vector pixel, vector unsigned int);
15115 vector pixel vec_srl (vector pixel, vector unsigned short);
15116 vector pixel vec_srl (vector pixel, vector unsigned char);
15117 vector signed char vec_srl (vector signed char, vector unsigned int);
15118 vector signed char vec_srl (vector signed char, vector unsigned short);
15119 vector signed char vec_srl (vector signed char, vector unsigned char);
15120 vector unsigned char vec_srl (vector unsigned char,
15121 vector unsigned int);
15122 vector unsigned char vec_srl (vector unsigned char,
15123 vector unsigned short);
15124 vector unsigned char vec_srl (vector unsigned char,
15125 vector unsigned char);
15126 vector bool char vec_srl (vector bool char, vector unsigned int);
15127 vector bool char vec_srl (vector bool char, vector unsigned short);
15128 vector bool char vec_srl (vector bool char, vector unsigned char);
15130 vector float vec_sro (vector float, vector signed char);
15131 vector float vec_sro (vector float, vector unsigned char);
15132 vector signed int vec_sro (vector signed int, vector signed char);
15133 vector signed int vec_sro (vector signed int, vector unsigned char);
15134 vector unsigned int vec_sro (vector unsigned int, vector signed char);
15135 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
15136 vector signed short vec_sro (vector signed short, vector signed char);
15137 vector signed short vec_sro (vector signed short, vector unsigned char);
15138 vector unsigned short vec_sro (vector unsigned short,
15139 vector signed char);
15140 vector unsigned short vec_sro (vector unsigned short,
15141 vector unsigned char);
15142 vector pixel vec_sro (vector pixel, vector signed char);
15143 vector pixel vec_sro (vector pixel, vector unsigned char);
15144 vector signed char vec_sro (vector signed char, vector signed char);
15145 vector signed char vec_sro (vector signed char, vector unsigned char);
15146 vector unsigned char vec_sro (vector unsigned char, vector signed char);
15147 vector unsigned char vec_sro (vector unsigned char,
15148 vector unsigned char);
15150 void vec_st (vector float, int, vector float *);
15151 void vec_st (vector float, int, float *);
15152 void vec_st (vector signed int, int, vector signed int *);
15153 void vec_st (vector signed int, int, int *);
15154 void vec_st (vector unsigned int, int, vector unsigned int *);
15155 void vec_st (vector unsigned int, int, unsigned int *);
15156 void vec_st (vector bool int, int, vector bool int *);
15157 void vec_st (vector bool int, int, unsigned int *);
15158 void vec_st (vector bool int, int, int *);
15159 void vec_st (vector signed short, int, vector signed short *);
15160 void vec_st (vector signed short, int, short *);
15161 void vec_st (vector unsigned short, int, vector unsigned short *);
15162 void vec_st (vector unsigned short, int, unsigned short *);
15163 void vec_st (vector bool short, int, vector bool short *);
15164 void vec_st (vector bool short, int, unsigned short *);
15165 void vec_st (vector pixel, int, vector pixel *);
15166 void vec_st (vector pixel, int, unsigned short *);
15167 void vec_st (vector pixel, int, short *);
15168 void vec_st (vector bool short, int, short *);
15169 void vec_st (vector signed char, int, vector signed char *);
15170 void vec_st (vector signed char, int, signed char *);
15171 void vec_st (vector unsigned char, int, vector unsigned char *);
15172 void vec_st (vector unsigned char, int, unsigned char *);
15173 void vec_st (vector bool char, int, vector bool char *);
15174 void vec_st (vector bool char, int, unsigned char *);
15175 void vec_st (vector bool char, int, signed char *);
15177 void vec_ste (vector signed char, int, signed char *);
15178 void vec_ste (vector unsigned char, int, unsigned char *);
15179 void vec_ste (vector bool char, int, signed char *);
15180 void vec_ste (vector bool char, int, unsigned char *);
15181 void vec_ste (vector signed short, int, short *);
15182 void vec_ste (vector unsigned short, int, unsigned short *);
15183 void vec_ste (vector bool short, int, short *);
15184 void vec_ste (vector bool short, int, unsigned short *);
15185 void vec_ste (vector pixel, int, short *);
15186 void vec_ste (vector pixel, int, unsigned short *);
15187 void vec_ste (vector float, int, float *);
15188 void vec_ste (vector signed int, int, int *);
15189 void vec_ste (vector unsigned int, int, unsigned int *);
15190 void vec_ste (vector bool int, int, int *);
15191 void vec_ste (vector bool int, int, unsigned int *);
15193 void vec_stvewx (vector float, int, float *);
15194 void vec_stvewx (vector signed int, int, int *);
15195 void vec_stvewx (vector unsigned int, int, unsigned int *);
15196 void vec_stvewx (vector bool int, int, int *);
15197 void vec_stvewx (vector bool int, int, unsigned int *);
15199 void vec_stvehx (vector signed short, int, short *);
15200 void vec_stvehx (vector unsigned short, int, unsigned short *);
15201 void vec_stvehx (vector bool short, int, short *);
15202 void vec_stvehx (vector bool short, int, unsigned short *);
15203 void vec_stvehx (vector pixel, int, short *);
15204 void vec_stvehx (vector pixel, int, unsigned short *);
15206 void vec_stvebx (vector signed char, int, signed char *);
15207 void vec_stvebx (vector unsigned char, int, unsigned char *);
15208 void vec_stvebx (vector bool char, int, signed char *);
15209 void vec_stvebx (vector bool char, int, unsigned char *);
15211 void vec_stl (vector float, int, vector float *);
15212 void vec_stl (vector float, int, float *);
15213 void vec_stl (vector signed int, int, vector signed int *);
15214 void vec_stl (vector signed int, int, int *);
15215 void vec_stl (vector unsigned int, int, vector unsigned int *);
15216 void vec_stl (vector unsigned int, int, unsigned int *);
15217 void vec_stl (vector bool int, int, vector bool int *);
15218 void vec_stl (vector bool int, int, unsigned int *);
15219 void vec_stl (vector bool int, int, int *);
15220 void vec_stl (vector signed short, int, vector signed short *);
15221 void vec_stl (vector signed short, int, short *);
15222 void vec_stl (vector unsigned short, int, vector unsigned short *);
15223 void vec_stl (vector unsigned short, int, unsigned short *);
15224 void vec_stl (vector bool short, int, vector bool short *);
15225 void vec_stl (vector bool short, int, unsigned short *);
15226 void vec_stl (vector bool short, int, short *);
15227 void vec_stl (vector pixel, int, vector pixel *);
15228 void vec_stl (vector pixel, int, unsigned short *);
15229 void vec_stl (vector pixel, int, short *);
15230 void vec_stl (vector signed char, int, vector signed char *);
15231 void vec_stl (vector signed char, int, signed char *);
15232 void vec_stl (vector unsigned char, int, vector unsigned char *);
15233 void vec_stl (vector unsigned char, int, unsigned char *);
15234 void vec_stl (vector bool char, int, vector bool char *);
15235 void vec_stl (vector bool char, int, unsigned char *);
15236 void vec_stl (vector bool char, int, signed char *);
15238 vector signed char vec_sub (vector bool char, vector signed char);
15239 vector signed char vec_sub (vector signed char, vector bool char);
15240 vector signed char vec_sub (vector signed char, vector signed char);
15241 vector unsigned char vec_sub (vector bool char, vector unsigned char);
15242 vector unsigned char vec_sub (vector unsigned char, vector bool char);
15243 vector unsigned char vec_sub (vector unsigned char,
15244 vector unsigned char);
15245 vector signed short vec_sub (vector bool short, vector signed short);
15246 vector signed short vec_sub (vector signed short, vector bool short);
15247 vector signed short vec_sub (vector signed short, vector signed short);
15248 vector unsigned short vec_sub (vector bool short,
15249 vector unsigned short);
15250 vector unsigned short vec_sub (vector unsigned short,
15251 vector bool short);
15252 vector unsigned short vec_sub (vector unsigned short,
15253 vector unsigned short);
15254 vector signed int vec_sub (vector bool int, vector signed int);
15255 vector signed int vec_sub (vector signed int, vector bool int);
15256 vector signed int vec_sub (vector signed int, vector signed int);
15257 vector unsigned int vec_sub (vector bool int, vector unsigned int);
15258 vector unsigned int vec_sub (vector unsigned int, vector bool int);
15259 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
15260 vector float vec_sub (vector float, vector float);
15262 vector float vec_vsubfp (vector float, vector float);
15264 vector signed int vec_vsubuwm (vector bool int, vector signed int);
15265 vector signed int vec_vsubuwm (vector signed int, vector bool int);
15266 vector signed int vec_vsubuwm (vector signed int, vector signed int);
15267 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
15268 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
15269 vector unsigned int vec_vsubuwm (vector unsigned int,
15270 vector unsigned int);
15272 vector signed short vec_vsubuhm (vector bool short,
15273 vector signed short);
15274 vector signed short vec_vsubuhm (vector signed short,
15275 vector bool short);
15276 vector signed short vec_vsubuhm (vector signed short,
15277 vector signed short);
15278 vector unsigned short vec_vsubuhm (vector bool short,
15279 vector unsigned short);
15280 vector unsigned short vec_vsubuhm (vector unsigned short,
15281 vector bool short);
15282 vector unsigned short vec_vsubuhm (vector unsigned short,
15283 vector unsigned short);
15285 vector signed char vec_vsububm (vector bool char, vector signed char);
15286 vector signed char vec_vsububm (vector signed char, vector bool char);
15287 vector signed char vec_vsububm (vector signed char, vector signed char);
15288 vector unsigned char vec_vsububm (vector bool char,
15289 vector unsigned char);
15290 vector unsigned char vec_vsububm (vector unsigned char,
15292 vector unsigned char vec_vsububm (vector unsigned char,
15293 vector unsigned char);
15295 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
15297 vector unsigned char vec_subs (vector bool char, vector unsigned char);
15298 vector unsigned char vec_subs (vector unsigned char, vector bool char);
15299 vector unsigned char vec_subs (vector unsigned char,
15300 vector unsigned char);
15301 vector signed char vec_subs (vector bool char, vector signed char);
15302 vector signed char vec_subs (vector signed char, vector bool char);
15303 vector signed char vec_subs (vector signed char, vector signed char);
15304 vector unsigned short vec_subs (vector bool short,
15305 vector unsigned short);
15306 vector unsigned short vec_subs (vector unsigned short,
15307 vector bool short);
15308 vector unsigned short vec_subs (vector unsigned short,
15309 vector unsigned short);
15310 vector signed short vec_subs (vector bool short, vector signed short);
15311 vector signed short vec_subs (vector signed short, vector bool short);
15312 vector signed short vec_subs (vector signed short, vector signed short);
15313 vector unsigned int vec_subs (vector bool int, vector unsigned int);
15314 vector unsigned int vec_subs (vector unsigned int, vector bool int);
15315 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
15316 vector signed int vec_subs (vector bool int, vector signed int);
15317 vector signed int vec_subs (vector signed int, vector bool int);
15318 vector signed int vec_subs (vector signed int, vector signed int);
15320 vector signed int vec_vsubsws (vector bool int, vector signed int);
15321 vector signed int vec_vsubsws (vector signed int, vector bool int);
15322 vector signed int vec_vsubsws (vector signed int, vector signed int);
15324 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
15325 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
15326 vector unsigned int vec_vsubuws (vector unsigned int,
15327 vector unsigned int);
15329 vector signed short vec_vsubshs (vector bool short,
15330 vector signed short);
15331 vector signed short vec_vsubshs (vector signed short,
15332 vector bool short);
15333 vector signed short vec_vsubshs (vector signed short,
15334 vector signed short);
15336 vector unsigned short vec_vsubuhs (vector bool short,
15337 vector unsigned short);
15338 vector unsigned short vec_vsubuhs (vector unsigned short,
15339 vector bool short);
15340 vector unsigned short vec_vsubuhs (vector unsigned short,
15341 vector unsigned short);
15343 vector signed char vec_vsubsbs (vector bool char, vector signed char);
15344 vector signed char vec_vsubsbs (vector signed char, vector bool char);
15345 vector signed char vec_vsubsbs (vector signed char, vector signed char);
15347 vector unsigned char vec_vsububs (vector bool char,
15348 vector unsigned char);
15349 vector unsigned char vec_vsububs (vector unsigned char,
15351 vector unsigned char vec_vsububs (vector unsigned char,
15352 vector unsigned char);
15354 vector unsigned int vec_sum4s (vector unsigned char,
15355 vector unsigned int);
15356 vector signed int vec_sum4s (vector signed char, vector signed int);
15357 vector signed int vec_sum4s (vector signed short, vector signed int);
15359 vector signed int vec_vsum4shs (vector signed short, vector signed int);
15361 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
15363 vector unsigned int vec_vsum4ubs (vector unsigned char,
15364 vector unsigned int);
15366 vector signed int vec_sum2s (vector signed int, vector signed int);
15368 vector signed int vec_sums (vector signed int, vector signed int);
15370 vector float vec_trunc (vector float);
15372 vector signed short vec_unpackh (vector signed char);
15373 vector bool short vec_unpackh (vector bool char);
15374 vector signed int vec_unpackh (vector signed short);
15375 vector bool int vec_unpackh (vector bool short);
15376 vector unsigned int vec_unpackh (vector pixel);
15378 vector bool int vec_vupkhsh (vector bool short);
15379 vector signed int vec_vupkhsh (vector signed short);
15381 vector unsigned int vec_vupkhpx (vector pixel);
15383 vector bool short vec_vupkhsb (vector bool char);
15384 vector signed short vec_vupkhsb (vector signed char);
15386 vector signed short vec_unpackl (vector signed char);
15387 vector bool short vec_unpackl (vector bool char);
15388 vector unsigned int vec_unpackl (vector pixel);
15389 vector signed int vec_unpackl (vector signed short);
15390 vector bool int vec_unpackl (vector bool short);
15392 vector unsigned int vec_vupklpx (vector pixel);
15394 vector bool int vec_vupklsh (vector bool short);
15395 vector signed int vec_vupklsh (vector signed short);
15397 vector bool short vec_vupklsb (vector bool char);
15398 vector signed short vec_vupklsb (vector signed char);
15400 vector float vec_xor (vector float, vector float);
15401 vector float vec_xor (vector float, vector bool int);
15402 vector float vec_xor (vector bool int, vector float);
15403 vector bool int vec_xor (vector bool int, vector bool int);
15404 vector signed int vec_xor (vector bool int, vector signed int);
15405 vector signed int vec_xor (vector signed int, vector bool int);
15406 vector signed int vec_xor (vector signed int, vector signed int);
15407 vector unsigned int vec_xor (vector bool int, vector unsigned int);
15408 vector unsigned int vec_xor (vector unsigned int, vector bool int);
15409 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
15410 vector bool short vec_xor (vector bool short, vector bool short);
15411 vector signed short vec_xor (vector bool short, vector signed short);
15412 vector signed short vec_xor (vector signed short, vector bool short);
15413 vector signed short vec_xor (vector signed short, vector signed short);
15414 vector unsigned short vec_xor (vector bool short,
15415 vector unsigned short);
15416 vector unsigned short vec_xor (vector unsigned short,
15417 vector bool short);
15418 vector unsigned short vec_xor (vector unsigned short,
15419 vector unsigned short);
15420 vector signed char vec_xor (vector bool char, vector signed char);
15421 vector bool char vec_xor (vector bool char, vector bool char);
15422 vector signed char vec_xor (vector signed char, vector bool char);
15423 vector signed char vec_xor (vector signed char, vector signed char);
15424 vector unsigned char vec_xor (vector bool char, vector unsigned char);
15425 vector unsigned char vec_xor (vector unsigned char, vector bool char);
15426 vector unsigned char vec_xor (vector unsigned char,
15427 vector unsigned char);
15429 int vec_all_eq (vector signed char, vector bool char);
15430 int vec_all_eq (vector signed char, vector signed char);
15431 int vec_all_eq (vector unsigned char, vector bool char);
15432 int vec_all_eq (vector unsigned char, vector unsigned char);
15433 int vec_all_eq (vector bool char, vector bool char);
15434 int vec_all_eq (vector bool char, vector unsigned char);
15435 int vec_all_eq (vector bool char, vector signed char);
15436 int vec_all_eq (vector signed short, vector bool short);
15437 int vec_all_eq (vector signed short, vector signed short);
15438 int vec_all_eq (vector unsigned short, vector bool short);
15439 int vec_all_eq (vector unsigned short, vector unsigned short);
15440 int vec_all_eq (vector bool short, vector bool short);
15441 int vec_all_eq (vector bool short, vector unsigned short);
15442 int vec_all_eq (vector bool short, vector signed short);
15443 int vec_all_eq (vector pixel, vector pixel);
15444 int vec_all_eq (vector signed int, vector bool int);
15445 int vec_all_eq (vector signed int, vector signed int);
15446 int vec_all_eq (vector unsigned int, vector bool int);
15447 int vec_all_eq (vector unsigned int, vector unsigned int);
15448 int vec_all_eq (vector bool int, vector bool int);
15449 int vec_all_eq (vector bool int, vector unsigned int);
15450 int vec_all_eq (vector bool int, vector signed int);
15451 int vec_all_eq (vector float, vector float);
15453 int vec_all_ge (vector bool char, vector unsigned char);
15454 int vec_all_ge (vector unsigned char, vector bool char);
15455 int vec_all_ge (vector unsigned char, vector unsigned char);
15456 int vec_all_ge (vector bool char, vector signed char);
15457 int vec_all_ge (vector signed char, vector bool char);
15458 int vec_all_ge (vector signed char, vector signed char);
15459 int vec_all_ge (vector bool short, vector unsigned short);
15460 int vec_all_ge (vector unsigned short, vector bool short);
15461 int vec_all_ge (vector unsigned short, vector unsigned short);
15462 int vec_all_ge (vector signed short, vector signed short);
15463 int vec_all_ge (vector bool short, vector signed short);
15464 int vec_all_ge (vector signed short, vector bool short);
15465 int vec_all_ge (vector bool int, vector unsigned int);
15466 int vec_all_ge (vector unsigned int, vector bool int);
15467 int vec_all_ge (vector unsigned int, vector unsigned int);
15468 int vec_all_ge (vector bool int, vector signed int);
15469 int vec_all_ge (vector signed int, vector bool int);
15470 int vec_all_ge (vector signed int, vector signed int);
15471 int vec_all_ge (vector float, vector float);
15473 int vec_all_gt (vector bool char, vector unsigned char);
15474 int vec_all_gt (vector unsigned char, vector bool char);
15475 int vec_all_gt (vector unsigned char, vector unsigned char);
15476 int vec_all_gt (vector bool char, vector signed char);
15477 int vec_all_gt (vector signed char, vector bool char);
15478 int vec_all_gt (vector signed char, vector signed char);
15479 int vec_all_gt (vector bool short, vector unsigned short);
15480 int vec_all_gt (vector unsigned short, vector bool short);
15481 int vec_all_gt (vector unsigned short, vector unsigned short);
15482 int vec_all_gt (vector bool short, vector signed short);
15483 int vec_all_gt (vector signed short, vector bool short);
15484 int vec_all_gt (vector signed short, vector signed short);
15485 int vec_all_gt (vector bool int, vector unsigned int);
15486 int vec_all_gt (vector unsigned int, vector bool int);
15487 int vec_all_gt (vector unsigned int, vector unsigned int);
15488 int vec_all_gt (vector bool int, vector signed int);
15489 int vec_all_gt (vector signed int, vector bool int);
15490 int vec_all_gt (vector signed int, vector signed int);
15491 int vec_all_gt (vector float, vector float);
15493 int vec_all_in (vector float, vector float);
15495 int vec_all_le (vector bool char, vector unsigned char);
15496 int vec_all_le (vector unsigned char, vector bool char);
15497 int vec_all_le (vector unsigned char, vector unsigned char);
15498 int vec_all_le (vector bool char, vector signed char);
15499 int vec_all_le (vector signed char, vector bool char);
15500 int vec_all_le (vector signed char, vector signed char);
15501 int vec_all_le (vector bool short, vector unsigned short);
15502 int vec_all_le (vector unsigned short, vector bool short);
15503 int vec_all_le (vector unsigned short, vector unsigned short);
15504 int vec_all_le (vector bool short, vector signed short);
15505 int vec_all_le (vector signed short, vector bool short);
15506 int vec_all_le (vector signed short, vector signed short);
15507 int vec_all_le (vector bool int, vector unsigned int);
15508 int vec_all_le (vector unsigned int, vector bool int);
15509 int vec_all_le (vector unsigned int, vector unsigned int);
15510 int vec_all_le (vector bool int, vector signed int);
15511 int vec_all_le (vector signed int, vector bool int);
15512 int vec_all_le (vector signed int, vector signed int);
15513 int vec_all_le (vector float, vector float);
15515 int vec_all_lt (vector bool char, vector unsigned char);
15516 int vec_all_lt (vector unsigned char, vector bool char);
15517 int vec_all_lt (vector unsigned char, vector unsigned char);
15518 int vec_all_lt (vector bool char, vector signed char);
15519 int vec_all_lt (vector signed char, vector bool char);
15520 int vec_all_lt (vector signed char, vector signed char);
15521 int vec_all_lt (vector bool short, vector unsigned short);
15522 int vec_all_lt (vector unsigned short, vector bool short);
15523 int vec_all_lt (vector unsigned short, vector unsigned short);
15524 int vec_all_lt (vector bool short, vector signed short);
15525 int vec_all_lt (vector signed short, vector bool short);
15526 int vec_all_lt (vector signed short, vector signed short);
15527 int vec_all_lt (vector bool int, vector unsigned int);
15528 int vec_all_lt (vector unsigned int, vector bool int);
15529 int vec_all_lt (vector unsigned int, vector unsigned int);
15530 int vec_all_lt (vector bool int, vector signed int);
15531 int vec_all_lt (vector signed int, vector bool int);
15532 int vec_all_lt (vector signed int, vector signed int);
15533 int vec_all_lt (vector float, vector float);
15535 int vec_all_nan (vector float);
15537 int vec_all_ne (vector signed char, vector bool char);
15538 int vec_all_ne (vector signed char, vector signed char);
15539 int vec_all_ne (vector unsigned char, vector bool char);
15540 int vec_all_ne (vector unsigned char, vector unsigned char);
15541 int vec_all_ne (vector bool char, vector bool char);
15542 int vec_all_ne (vector bool char, vector unsigned char);
15543 int vec_all_ne (vector bool char, vector signed char);
15544 int vec_all_ne (vector signed short, vector bool short);
15545 int vec_all_ne (vector signed short, vector signed short);
15546 int vec_all_ne (vector unsigned short, vector bool short);
15547 int vec_all_ne (vector unsigned short, vector unsigned short);
15548 int vec_all_ne (vector bool short, vector bool short);
15549 int vec_all_ne (vector bool short, vector unsigned short);
15550 int vec_all_ne (vector bool short, vector signed short);
15551 int vec_all_ne (vector pixel, vector pixel);
15552 int vec_all_ne (vector signed int, vector bool int);
15553 int vec_all_ne (vector signed int, vector signed int);
15554 int vec_all_ne (vector unsigned int, vector bool int);
15555 int vec_all_ne (vector unsigned int, vector unsigned int);
15556 int vec_all_ne (vector bool int, vector bool int);
15557 int vec_all_ne (vector bool int, vector unsigned int);
15558 int vec_all_ne (vector bool int, vector signed int);
15559 int vec_all_ne (vector float, vector float);
15561 int vec_all_nge (vector float, vector float);
15563 int vec_all_ngt (vector float, vector float);
15565 int vec_all_nle (vector float, vector float);
15567 int vec_all_nlt (vector float, vector float);
15569 int vec_all_numeric (vector float);
15571 int vec_any_eq (vector signed char, vector bool char);
15572 int vec_any_eq (vector signed char, vector signed char);
15573 int vec_any_eq (vector unsigned char, vector bool char);
15574 int vec_any_eq (vector unsigned char, vector unsigned char);
15575 int vec_any_eq (vector bool char, vector bool char);
15576 int vec_any_eq (vector bool char, vector unsigned char);
15577 int vec_any_eq (vector bool char, vector signed char);
15578 int vec_any_eq (vector signed short, vector bool short);
15579 int vec_any_eq (vector signed short, vector signed short);
15580 int vec_any_eq (vector unsigned short, vector bool short);
15581 int vec_any_eq (vector unsigned short, vector unsigned short);
15582 int vec_any_eq (vector bool short, vector bool short);
15583 int vec_any_eq (vector bool short, vector unsigned short);
15584 int vec_any_eq (vector bool short, vector signed short);
15585 int vec_any_eq (vector pixel, vector pixel);
15586 int vec_any_eq (vector signed int, vector bool int);
15587 int vec_any_eq (vector signed int, vector signed int);
15588 int vec_any_eq (vector unsigned int, vector bool int);
15589 int vec_any_eq (vector unsigned int, vector unsigned int);
15590 int vec_any_eq (vector bool int, vector bool int);
15591 int vec_any_eq (vector bool int, vector unsigned int);
15592 int vec_any_eq (vector bool int, vector signed int);
15593 int vec_any_eq (vector float, vector float);
15595 int vec_any_ge (vector signed char, vector bool char);
15596 int vec_any_ge (vector unsigned char, vector bool char);
15597 int vec_any_ge (vector unsigned char, vector unsigned char);
15598 int vec_any_ge (vector signed char, vector signed char);
15599 int vec_any_ge (vector bool char, vector unsigned char);
15600 int vec_any_ge (vector bool char, vector signed char);
15601 int vec_any_ge (vector unsigned short, vector bool short);
15602 int vec_any_ge (vector unsigned short, vector unsigned short);
15603 int vec_any_ge (vector signed short, vector signed short);
15604 int vec_any_ge (vector signed short, vector bool short);
15605 int vec_any_ge (vector bool short, vector unsigned short);
15606 int vec_any_ge (vector bool short, vector signed short);
15607 int vec_any_ge (vector signed int, vector bool int);
15608 int vec_any_ge (vector unsigned int, vector bool int);
15609 int vec_any_ge (vector unsigned int, vector unsigned int);
15610 int vec_any_ge (vector signed int, vector signed int);
15611 int vec_any_ge (vector bool int, vector unsigned int);
15612 int vec_any_ge (vector bool int, vector signed int);
15613 int vec_any_ge (vector float, vector float);
15615 int vec_any_gt (vector bool char, vector unsigned char);
15616 int vec_any_gt (vector unsigned char, vector bool char);
15617 int vec_any_gt (vector unsigned char, vector unsigned char);
15618 int vec_any_gt (vector bool char, vector signed char);
15619 int vec_any_gt (vector signed char, vector bool char);
15620 int vec_any_gt (vector signed char, vector signed char);
15621 int vec_any_gt (vector bool short, vector unsigned short);
15622 int vec_any_gt (vector unsigned short, vector bool short);
15623 int vec_any_gt (vector unsigned short, vector unsigned short);
15624 int vec_any_gt (vector bool short, vector signed short);
15625 int vec_any_gt (vector signed short, vector bool short);
15626 int vec_any_gt (vector signed short, vector signed short);
15627 int vec_any_gt (vector bool int, vector unsigned int);
15628 int vec_any_gt (vector unsigned int, vector bool int);
15629 int vec_any_gt (vector unsigned int, vector unsigned int);
15630 int vec_any_gt (vector bool int, vector signed int);
15631 int vec_any_gt (vector signed int, vector bool int);
15632 int vec_any_gt (vector signed int, vector signed int);
15633 int vec_any_gt (vector float, vector float);
15635 int vec_any_le (vector bool char, vector unsigned char);
15636 int vec_any_le (vector unsigned char, vector bool char);
15637 int vec_any_le (vector unsigned char, vector unsigned char);
15638 int vec_any_le (vector bool char, vector signed char);
15639 int vec_any_le (vector signed char, vector bool char);
15640 int vec_any_le (vector signed char, vector signed char);
15641 int vec_any_le (vector bool short, vector unsigned short);
15642 int vec_any_le (vector unsigned short, vector bool short);
15643 int vec_any_le (vector unsigned short, vector unsigned short);
15644 int vec_any_le (vector bool short, vector signed short);
15645 int vec_any_le (vector signed short, vector bool short);
15646 int vec_any_le (vector signed short, vector signed short);
15647 int vec_any_le (vector bool int, vector unsigned int);
15648 int vec_any_le (vector unsigned int, vector bool int);
15649 int vec_any_le (vector unsigned int, vector unsigned int);
15650 int vec_any_le (vector bool int, vector signed int);
15651 int vec_any_le (vector signed int, vector bool int);
15652 int vec_any_le (vector signed int, vector signed int);
15653 int vec_any_le (vector float, vector float);
15655 int vec_any_lt (vector bool char, vector unsigned char);
15656 int vec_any_lt (vector unsigned char, vector bool char);
15657 int vec_any_lt (vector unsigned char, vector unsigned char);
15658 int vec_any_lt (vector bool char, vector signed char);
15659 int vec_any_lt (vector signed char, vector bool char);
15660 int vec_any_lt (vector signed char, vector signed char);
15661 int vec_any_lt (vector bool short, vector unsigned short);
15662 int vec_any_lt (vector unsigned short, vector bool short);
15663 int vec_any_lt (vector unsigned short, vector unsigned short);
15664 int vec_any_lt (vector bool short, vector signed short);
15665 int vec_any_lt (vector signed short, vector bool short);
15666 int vec_any_lt (vector signed short, vector signed short);
15667 int vec_any_lt (vector bool int, vector unsigned int);
15668 int vec_any_lt (vector unsigned int, vector bool int);
15669 int vec_any_lt (vector unsigned int, vector unsigned int);
15670 int vec_any_lt (vector bool int, vector signed int);
15671 int vec_any_lt (vector signed int, vector bool int);
15672 int vec_any_lt (vector signed int, vector signed int);
15673 int vec_any_lt (vector float, vector float);
15675 int vec_any_nan (vector float);
15677 int vec_any_ne (vector signed char, vector bool char);
15678 int vec_any_ne (vector signed char, vector signed char);
15679 int vec_any_ne (vector unsigned char, vector bool char);
15680 int vec_any_ne (vector unsigned char, vector unsigned char);
15681 int vec_any_ne (vector bool char, vector bool char);
15682 int vec_any_ne (vector bool char, vector unsigned char);
15683 int vec_any_ne (vector bool char, vector signed char);
15684 int vec_any_ne (vector signed short, vector bool short);
15685 int vec_any_ne (vector signed short, vector signed short);
15686 int vec_any_ne (vector unsigned short, vector bool short);
15687 int vec_any_ne (vector unsigned short, vector unsigned short);
15688 int vec_any_ne (vector bool short, vector bool short);
15689 int vec_any_ne (vector bool short, vector unsigned short);
15690 int vec_any_ne (vector bool short, vector signed short);
15691 int vec_any_ne (vector pixel, vector pixel);
15692 int vec_any_ne (vector signed int, vector bool int);
15693 int vec_any_ne (vector signed int, vector signed int);
15694 int vec_any_ne (vector unsigned int, vector bool int);
15695 int vec_any_ne (vector unsigned int, vector unsigned int);
15696 int vec_any_ne (vector bool int, vector bool int);
15697 int vec_any_ne (vector bool int, vector unsigned int);
15698 int vec_any_ne (vector bool int, vector signed int);
15699 int vec_any_ne (vector float, vector float);
15701 int vec_any_nge (vector float, vector float);
15703 int vec_any_ngt (vector float, vector float);
15705 int vec_any_nle (vector float, vector float);
15707 int vec_any_nlt (vector float, vector float);
15709 int vec_any_numeric (vector float);
15711 int vec_any_out (vector float, vector float);
15714 If the vector/scalar (VSX) instruction set is available, the following
15715 additional functions are available:
15718 vector double vec_abs (vector double);
15719 vector double vec_add (vector double, vector double);
15720 vector double vec_and (vector double, vector double);
15721 vector double vec_and (vector double, vector bool long);
15722 vector double vec_and (vector bool long, vector double);
15723 vector long vec_and (vector long, vector long);
15724 vector long vec_and (vector long, vector bool long);
15725 vector long vec_and (vector bool long, vector long);
15726 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
15727 vector unsigned long vec_and (vector unsigned long, vector bool long);
15728 vector unsigned long vec_and (vector bool long, vector unsigned long);
15729 vector double vec_andc (vector double, vector double);
15730 vector double vec_andc (vector double, vector bool long);
15731 vector double vec_andc (vector bool long, vector double);
15732 vector long vec_andc (vector long, vector long);
15733 vector long vec_andc (vector long, vector bool long);
15734 vector long vec_andc (vector bool long, vector long);
15735 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
15736 vector unsigned long vec_andc (vector unsigned long, vector bool long);
15737 vector unsigned long vec_andc (vector bool long, vector unsigned long);
15738 vector double vec_ceil (vector double);
15739 vector bool long vec_cmpeq (vector double, vector double);
15740 vector bool long vec_cmpge (vector double, vector double);
15741 vector bool long vec_cmpgt (vector double, vector double);
15742 vector bool long vec_cmple (vector double, vector double);
15743 vector bool long vec_cmplt (vector double, vector double);
15744 vector double vec_cpsgn (vector double, vector double);
15745 vector float vec_div (vector float, vector float);
15746 vector double vec_div (vector double, vector double);
15747 vector long vec_div (vector long, vector long);
15748 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
15749 vector double vec_floor (vector double);
15750 vector double vec_ld (int, const vector double *);
15751 vector double vec_ld (int, const double *);
15752 vector double vec_ldl (int, const vector double *);
15753 vector double vec_ldl (int, const double *);
15754 vector unsigned char vec_lvsl (int, const volatile double *);
15755 vector unsigned char vec_lvsr (int, const volatile double *);
15756 vector double vec_madd (vector double, vector double, vector double);
15757 vector double vec_max (vector double, vector double);
15758 vector signed long vec_mergeh (vector signed long, vector signed long);
15759 vector signed long vec_mergeh (vector signed long, vector bool long);
15760 vector signed long vec_mergeh (vector bool long, vector signed long);
15761 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
15762 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
15763 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
15764 vector signed long vec_mergel (vector signed long, vector signed long);
15765 vector signed long vec_mergel (vector signed long, vector bool long);
15766 vector signed long vec_mergel (vector bool long, vector signed long);
15767 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
15768 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
15769 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
15770 vector double vec_min (vector double, vector double);
15771 vector float vec_msub (vector float, vector float, vector float);
15772 vector double vec_msub (vector double, vector double, vector double);
15773 vector float vec_mul (vector float, vector float);
15774 vector double vec_mul (vector double, vector double);
15775 vector long vec_mul (vector long, vector long);
15776 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
15777 vector float vec_nearbyint (vector float);
15778 vector double vec_nearbyint (vector double);
15779 vector float vec_nmadd (vector float, vector float, vector float);
15780 vector double vec_nmadd (vector double, vector double, vector double);
15781 vector double vec_nmsub (vector double, vector double, vector double);
15782 vector double vec_nor (vector double, vector double);
15783 vector long vec_nor (vector long, vector long);
15784 vector long vec_nor (vector long, vector bool long);
15785 vector long vec_nor (vector bool long, vector long);
15786 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
15787 vector unsigned long vec_nor (vector unsigned long, vector bool long);
15788 vector unsigned long vec_nor (vector bool long, vector unsigned long);
15789 vector double vec_or (vector double, vector double);
15790 vector double vec_or (vector double, vector bool long);
15791 vector double vec_or (vector bool long, vector double);
15792 vector long vec_or (vector long, vector long);
15793 vector long vec_or (vector long, vector bool long);
15794 vector long vec_or (vector bool long, vector long);
15795 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
15796 vector unsigned long vec_or (vector unsigned long, vector bool long);
15797 vector unsigned long vec_or (vector bool long, vector unsigned long);
15798 vector double vec_perm (vector double, vector double, vector unsigned char);
15799 vector long vec_perm (vector long, vector long, vector unsigned char);
15800 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
15801 vector unsigned char);
15802 vector double vec_rint (vector double);
15803 vector double vec_recip (vector double, vector double);
15804 vector double vec_rsqrt (vector double);
15805 vector double vec_rsqrte (vector double);
15806 vector double vec_sel (vector double, vector double, vector bool long);
15807 vector double vec_sel (vector double, vector double, vector unsigned long);
15808 vector long vec_sel (vector long, vector long, vector long);
15809 vector long vec_sel (vector long, vector long, vector unsigned long);
15810 vector long vec_sel (vector long, vector long, vector bool long);
15811 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15813 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15814 vector unsigned long);
15815 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
15817 vector double vec_splats (double);
15818 vector signed long vec_splats (signed long);
15819 vector unsigned long vec_splats (unsigned long);
15820 vector float vec_sqrt (vector float);
15821 vector double vec_sqrt (vector double);
15822 void vec_st (vector double, int, vector double *);
15823 void vec_st (vector double, int, double *);
15824 vector double vec_sub (vector double, vector double);
15825 vector double vec_trunc (vector double);
15826 vector double vec_xor (vector double, vector double);
15827 vector double vec_xor (vector double, vector bool long);
15828 vector double vec_xor (vector bool long, vector double);
15829 vector long vec_xor (vector long, vector long);
15830 vector long vec_xor (vector long, vector bool long);
15831 vector long vec_xor (vector bool long, vector long);
15832 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
15833 vector unsigned long vec_xor (vector unsigned long, vector bool long);
15834 vector unsigned long vec_xor (vector bool long, vector unsigned long);
15835 int vec_all_eq (vector double, vector double);
15836 int vec_all_ge (vector double, vector double);
15837 int vec_all_gt (vector double, vector double);
15838 int vec_all_le (vector double, vector double);
15839 int vec_all_lt (vector double, vector double);
15840 int vec_all_nan (vector double);
15841 int vec_all_ne (vector double, vector double);
15842 int vec_all_nge (vector double, vector double);
15843 int vec_all_ngt (vector double, vector double);
15844 int vec_all_nle (vector double, vector double);
15845 int vec_all_nlt (vector double, vector double);
15846 int vec_all_numeric (vector double);
15847 int vec_any_eq (vector double, vector double);
15848 int vec_any_ge (vector double, vector double);
15849 int vec_any_gt (vector double, vector double);
15850 int vec_any_le (vector double, vector double);
15851 int vec_any_lt (vector double, vector double);
15852 int vec_any_nan (vector double);
15853 int vec_any_ne (vector double, vector double);
15854 int vec_any_nge (vector double, vector double);
15855 int vec_any_ngt (vector double, vector double);
15856 int vec_any_nle (vector double, vector double);
15857 int vec_any_nlt (vector double, vector double);
15858 int vec_any_numeric (vector double);
15860 vector double vec_vsx_ld (int, const vector double *);
15861 vector double vec_vsx_ld (int, const double *);
15862 vector float vec_vsx_ld (int, const vector float *);
15863 vector float vec_vsx_ld (int, const float *);
15864 vector bool int vec_vsx_ld (int, const vector bool int *);
15865 vector signed int vec_vsx_ld (int, const vector signed int *);
15866 vector signed int vec_vsx_ld (int, const int *);
15867 vector signed int vec_vsx_ld (int, const long *);
15868 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
15869 vector unsigned int vec_vsx_ld (int, const unsigned int *);
15870 vector unsigned int vec_vsx_ld (int, const unsigned long *);
15871 vector bool short vec_vsx_ld (int, const vector bool short *);
15872 vector pixel vec_vsx_ld (int, const vector pixel *);
15873 vector signed short vec_vsx_ld (int, const vector signed short *);
15874 vector signed short vec_vsx_ld (int, const short *);
15875 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
15876 vector unsigned short vec_vsx_ld (int, const unsigned short *);
15877 vector bool char vec_vsx_ld (int, const vector bool char *);
15878 vector signed char vec_vsx_ld (int, const vector signed char *);
15879 vector signed char vec_vsx_ld (int, const signed char *);
15880 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
15881 vector unsigned char vec_vsx_ld (int, const unsigned char *);
15883 void vec_vsx_st (vector double, int, vector double *);
15884 void vec_vsx_st (vector double, int, double *);
15885 void vec_vsx_st (vector float, int, vector float *);
15886 void vec_vsx_st (vector float, int, float *);
15887 void vec_vsx_st (vector signed int, int, vector signed int *);
15888 void vec_vsx_st (vector signed int, int, int *);
15889 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
15890 void vec_vsx_st (vector unsigned int, int, unsigned int *);
15891 void vec_vsx_st (vector bool int, int, vector bool int *);
15892 void vec_vsx_st (vector bool int, int, unsigned int *);
15893 void vec_vsx_st (vector bool int, int, int *);
15894 void vec_vsx_st (vector signed short, int, vector signed short *);
15895 void vec_vsx_st (vector signed short, int, short *);
15896 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
15897 void vec_vsx_st (vector unsigned short, int, unsigned short *);
15898 void vec_vsx_st (vector bool short, int, vector bool short *);
15899 void vec_vsx_st (vector bool short, int, unsigned short *);
15900 void vec_vsx_st (vector pixel, int, vector pixel *);
15901 void vec_vsx_st (vector pixel, int, unsigned short *);
15902 void vec_vsx_st (vector pixel, int, short *);
15903 void vec_vsx_st (vector bool short, int, short *);
15904 void vec_vsx_st (vector signed char, int, vector signed char *);
15905 void vec_vsx_st (vector signed char, int, signed char *);
15906 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
15907 void vec_vsx_st (vector unsigned char, int, unsigned char *);
15908 void vec_vsx_st (vector bool char, int, vector bool char *);
15909 void vec_vsx_st (vector bool char, int, unsigned char *);
15910 void vec_vsx_st (vector bool char, int, signed char *);
15912 vector double vec_xxpermdi (vector double, vector double, int);
15913 vector float vec_xxpermdi (vector float, vector float, int);
15914 vector long long vec_xxpermdi (vector long long, vector long long, int);
15915 vector unsigned long long vec_xxpermdi (vector unsigned long long,
15916 vector unsigned long long, int);
15917 vector int vec_xxpermdi (vector int, vector int, int);
15918 vector unsigned int vec_xxpermdi (vector unsigned int,
15919 vector unsigned int, int);
15920 vector short vec_xxpermdi (vector short, vector short, int);
15921 vector unsigned short vec_xxpermdi (vector unsigned short,
15922 vector unsigned short, int);
15923 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
15924 vector unsigned char vec_xxpermdi (vector unsigned char,
15925 vector unsigned char, int);
15927 vector double vec_xxsldi (vector double, vector double, int);
15928 vector float vec_xxsldi (vector float, vector float, int);
15929 vector long long vec_xxsldi (vector long long, vector long long, int);
15930 vector unsigned long long vec_xxsldi (vector unsigned long long,
15931 vector unsigned long long, int);
15932 vector int vec_xxsldi (vector int, vector int, int);
15933 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
15934 vector short vec_xxsldi (vector short, vector short, int);
15935 vector unsigned short vec_xxsldi (vector unsigned short,
15936 vector unsigned short, int);
15937 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
15938 vector unsigned char vec_xxsldi (vector unsigned char,
15939 vector unsigned char, int);
15942 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
15943 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
15944 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
15945 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
15946 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
15948 If the ISA 2.07 additions to the vector/scalar (power8-vector)
15949 instruction set is available, the following additional functions are
15950 available for both 32-bit and 64-bit targets. For 64-bit targets, you
15951 can use @var{vector long} instead of @var{vector long long},
15952 @var{vector bool long} instead of @var{vector bool long long}, and
15953 @var{vector unsigned long} instead of @var{vector unsigned long long}.
15956 vector long long vec_abs (vector long long);
15958 vector long long vec_add (vector long long, vector long long);
15959 vector unsigned long long vec_add (vector unsigned long long,
15960 vector unsigned long long);
15962 int vec_all_eq (vector long long, vector long long);
15963 int vec_all_eq (vector unsigned long long, vector unsigned long long);
15964 int vec_all_ge (vector long long, vector long long);
15965 int vec_all_ge (vector unsigned long long, vector unsigned long long);
15966 int vec_all_gt (vector long long, vector long long);
15967 int vec_all_gt (vector unsigned long long, vector unsigned long long);
15968 int vec_all_le (vector long long, vector long long);
15969 int vec_all_le (vector unsigned long long, vector unsigned long long);
15970 int vec_all_lt (vector long long, vector long long);
15971 int vec_all_lt (vector unsigned long long, vector unsigned long long);
15972 int vec_all_ne (vector long long, vector long long);
15973 int vec_all_ne (vector unsigned long long, vector unsigned long long);
15975 int vec_any_eq (vector long long, vector long long);
15976 int vec_any_eq (vector unsigned long long, vector unsigned long long);
15977 int vec_any_ge (vector long long, vector long long);
15978 int vec_any_ge (vector unsigned long long, vector unsigned long long);
15979 int vec_any_gt (vector long long, vector long long);
15980 int vec_any_gt (vector unsigned long long, vector unsigned long long);
15981 int vec_any_le (vector long long, vector long long);
15982 int vec_any_le (vector unsigned long long, vector unsigned long long);
15983 int vec_any_lt (vector long long, vector long long);
15984 int vec_any_lt (vector unsigned long long, vector unsigned long long);
15985 int vec_any_ne (vector long long, vector long long);
15986 int vec_any_ne (vector unsigned long long, vector unsigned long long);
15988 vector long long vec_eqv (vector long long, vector long long);
15989 vector long long vec_eqv (vector bool long long, vector long long);
15990 vector long long vec_eqv (vector long long, vector bool long long);
15991 vector unsigned long long vec_eqv (vector unsigned long long,
15992 vector unsigned long long);
15993 vector unsigned long long vec_eqv (vector bool long long,
15994 vector unsigned long long);
15995 vector unsigned long long vec_eqv (vector unsigned long long,
15996 vector bool long long);
15997 vector int vec_eqv (vector int, vector int);
15998 vector int vec_eqv (vector bool int, vector int);
15999 vector int vec_eqv (vector int, vector bool int);
16000 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
16001 vector unsigned int vec_eqv (vector bool unsigned int,
16002 vector unsigned int);
16003 vector unsigned int vec_eqv (vector unsigned int,
16004 vector bool unsigned int);
16005 vector short vec_eqv (vector short, vector short);
16006 vector short vec_eqv (vector bool short, vector short);
16007 vector short vec_eqv (vector short, vector bool short);
16008 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
16009 vector unsigned short vec_eqv (vector bool unsigned short,
16010 vector unsigned short);
16011 vector unsigned short vec_eqv (vector unsigned short,
16012 vector bool unsigned short);
16013 vector signed char vec_eqv (vector signed char, vector signed char);
16014 vector signed char vec_eqv (vector bool signed char, vector signed char);
16015 vector signed char vec_eqv (vector signed char, vector bool signed char);
16016 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
16017 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
16018 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
16020 vector long long vec_max (vector long long, vector long long);
16021 vector unsigned long long vec_max (vector unsigned long long,
16022 vector unsigned long long);
16024 vector signed int vec_mergee (vector signed int, vector signed int);
16025 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
16026 vector bool int vec_mergee (vector bool int, vector bool int);
16028 vector signed int vec_mergeo (vector signed int, vector signed int);
16029 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
16030 vector bool int vec_mergeo (vector bool int, vector bool int);
16032 vector long long vec_min (vector long long, vector long long);
16033 vector unsigned long long vec_min (vector unsigned long long,
16034 vector unsigned long long);
16036 vector long long vec_nand (vector long long, vector long long);
16037 vector long long vec_nand (vector bool long long, vector long long);
16038 vector long long vec_nand (vector long long, vector bool long long);
16039 vector unsigned long long vec_nand (vector unsigned long long,
16040 vector unsigned long long);
16041 vector unsigned long long vec_nand (vector bool long long,
16042 vector unsigned long long);
16043 vector unsigned long long vec_nand (vector unsigned long long,
16044 vector bool long long);
16045 vector int vec_nand (vector int, vector int);
16046 vector int vec_nand (vector bool int, vector int);
16047 vector int vec_nand (vector int, vector bool int);
16048 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
16049 vector unsigned int vec_nand (vector bool unsigned int,
16050 vector unsigned int);
16051 vector unsigned int vec_nand (vector unsigned int,
16052 vector bool unsigned int);
16053 vector short vec_nand (vector short, vector short);
16054 vector short vec_nand (vector bool short, vector short);
16055 vector short vec_nand (vector short, vector bool short);
16056 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
16057 vector unsigned short vec_nand (vector bool unsigned short,
16058 vector unsigned short);
16059 vector unsigned short vec_nand (vector unsigned short,
16060 vector bool unsigned short);
16061 vector signed char vec_nand (vector signed char, vector signed char);
16062 vector signed char vec_nand (vector bool signed char, vector signed char);
16063 vector signed char vec_nand (vector signed char, vector bool signed char);
16064 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
16065 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
16066 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
16068 vector long long vec_orc (vector long long, vector long long);
16069 vector long long vec_orc (vector bool long long, vector long long);
16070 vector long long vec_orc (vector long long, vector bool long long);
16071 vector unsigned long long vec_orc (vector unsigned long long,
16072 vector unsigned long long);
16073 vector unsigned long long vec_orc (vector bool long long,
16074 vector unsigned long long);
16075 vector unsigned long long vec_orc (vector unsigned long long,
16076 vector bool long long);
16077 vector int vec_orc (vector int, vector int);
16078 vector int vec_orc (vector bool int, vector int);
16079 vector int vec_orc (vector int, vector bool int);
16080 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
16081 vector unsigned int vec_orc (vector bool unsigned int,
16082 vector unsigned int);
16083 vector unsigned int vec_orc (vector unsigned int,
16084 vector bool unsigned int);
16085 vector short vec_orc (vector short, vector short);
16086 vector short vec_orc (vector bool short, vector short);
16087 vector short vec_orc (vector short, vector bool short);
16088 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
16089 vector unsigned short vec_orc (vector bool unsigned short,
16090 vector unsigned short);
16091 vector unsigned short vec_orc (vector unsigned short,
16092 vector bool unsigned short);
16093 vector signed char vec_orc (vector signed char, vector signed char);
16094 vector signed char vec_orc (vector bool signed char, vector signed char);
16095 vector signed char vec_orc (vector signed char, vector bool signed char);
16096 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
16097 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
16098 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
16100 vector int vec_pack (vector long long, vector long long);
16101 vector unsigned int vec_pack (vector unsigned long long,
16102 vector unsigned long long);
16103 vector bool int vec_pack (vector bool long long, vector bool long long);
16105 vector int vec_packs (vector long long, vector long long);
16106 vector unsigned int vec_packs (vector unsigned long long,
16107 vector unsigned long long);
16109 vector unsigned int vec_packsu (vector long long, vector long long);
16110 vector unsigned int vec_packsu (vector unsigned long long,
16111 vector unsigned long long);
16113 vector long long vec_rl (vector long long,
16114 vector unsigned long long);
16115 vector long long vec_rl (vector unsigned long long,
16116 vector unsigned long long);
16118 vector long long vec_sl (vector long long, vector unsigned long long);
16119 vector long long vec_sl (vector unsigned long long,
16120 vector unsigned long long);
16122 vector long long vec_sr (vector long long, vector unsigned long long);
16123 vector unsigned long long char vec_sr (vector unsigned long long,
16124 vector unsigned long long);
16126 vector long long vec_sra (vector long long, vector unsigned long long);
16127 vector unsigned long long vec_sra (vector unsigned long long,
16128 vector unsigned long long);
16130 vector long long vec_sub (vector long long, vector long long);
16131 vector unsigned long long vec_sub (vector unsigned long long,
16132 vector unsigned long long);
16134 vector long long vec_unpackh (vector int);
16135 vector unsigned long long vec_unpackh (vector unsigned int);
16137 vector long long vec_unpackl (vector int);
16138 vector unsigned long long vec_unpackl (vector unsigned int);
16140 vector long long vec_vaddudm (vector long long, vector long long);
16141 vector long long vec_vaddudm (vector bool long long, vector long long);
16142 vector long long vec_vaddudm (vector long long, vector bool long long);
16143 vector unsigned long long vec_vaddudm (vector unsigned long long,
16144 vector unsigned long long);
16145 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
16146 vector unsigned long long);
16147 vector unsigned long long vec_vaddudm (vector unsigned long long,
16148 vector bool unsigned long long);
16150 vector long long vec_vbpermq (vector signed char, vector signed char);
16151 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
16153 vector long long vec_cntlz (vector long long);
16154 vector unsigned long long vec_cntlz (vector unsigned long long);
16155 vector int vec_cntlz (vector int);
16156 vector unsigned int vec_cntlz (vector int);
16157 vector short vec_cntlz (vector short);
16158 vector unsigned short vec_cntlz (vector unsigned short);
16159 vector signed char vec_cntlz (vector signed char);
16160 vector unsigned char vec_cntlz (vector unsigned char);
16162 vector long long vec_vclz (vector long long);
16163 vector unsigned long long vec_vclz (vector unsigned long long);
16164 vector int vec_vclz (vector int);
16165 vector unsigned int vec_vclz (vector int);
16166 vector short vec_vclz (vector short);
16167 vector unsigned short vec_vclz (vector unsigned short);
16168 vector signed char vec_vclz (vector signed char);
16169 vector unsigned char vec_vclz (vector unsigned char);
16171 vector signed char vec_vclzb (vector signed char);
16172 vector unsigned char vec_vclzb (vector unsigned char);
16174 vector long long vec_vclzd (vector long long);
16175 vector unsigned long long vec_vclzd (vector unsigned long long);
16177 vector short vec_vclzh (vector short);
16178 vector unsigned short vec_vclzh (vector unsigned short);
16180 vector int vec_vclzw (vector int);
16181 vector unsigned int vec_vclzw (vector int);
16183 vector signed char vec_vgbbd (vector signed char);
16184 vector unsigned char vec_vgbbd (vector unsigned char);
16186 vector long long vec_vmaxsd (vector long long, vector long long);
16188 vector unsigned long long vec_vmaxud (vector unsigned long long,
16189 unsigned vector long long);
16191 vector long long vec_vminsd (vector long long, vector long long);
16193 vector unsigned long long vec_vminud (vector long long,
16196 vector int vec_vpksdss (vector long long, vector long long);
16197 vector unsigned int vec_vpksdss (vector long long, vector long long);
16199 vector unsigned int vec_vpkudus (vector unsigned long long,
16200 vector unsigned long long);
16202 vector int vec_vpkudum (vector long long, vector long long);
16203 vector unsigned int vec_vpkudum (vector unsigned long long,
16204 vector unsigned long long);
16205 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
16207 vector long long vec_vpopcnt (vector long long);
16208 vector unsigned long long vec_vpopcnt (vector unsigned long long);
16209 vector int vec_vpopcnt (vector int);
16210 vector unsigned int vec_vpopcnt (vector int);
16211 vector short vec_vpopcnt (vector short);
16212 vector unsigned short vec_vpopcnt (vector unsigned short);
16213 vector signed char vec_vpopcnt (vector signed char);
16214 vector unsigned char vec_vpopcnt (vector unsigned char);
16216 vector signed char vec_vpopcntb (vector signed char);
16217 vector unsigned char vec_vpopcntb (vector unsigned char);
16219 vector long long vec_vpopcntd (vector long long);
16220 vector unsigned long long vec_vpopcntd (vector unsigned long long);
16222 vector short vec_vpopcnth (vector short);
16223 vector unsigned short vec_vpopcnth (vector unsigned short);
16225 vector int vec_vpopcntw (vector int);
16226 vector unsigned int vec_vpopcntw (vector int);
16228 vector long long vec_vrld (vector long long, vector unsigned long long);
16229 vector unsigned long long vec_vrld (vector unsigned long long,
16230 vector unsigned long long);
16232 vector long long vec_vsld (vector long long, vector unsigned long long);
16233 vector long long vec_vsld (vector unsigned long long,
16234 vector unsigned long long);
16236 vector long long vec_vsrad (vector long long, vector unsigned long long);
16237 vector unsigned long long vec_vsrad (vector unsigned long long,
16238 vector unsigned long long);
16240 vector long long vec_vsrd (vector long long, vector unsigned long long);
16241 vector unsigned long long char vec_vsrd (vector unsigned long long,
16242 vector unsigned long long);
16244 vector long long vec_vsubudm (vector long long, vector long long);
16245 vector long long vec_vsubudm (vector bool long long, vector long long);
16246 vector long long vec_vsubudm (vector long long, vector bool long long);
16247 vector unsigned long long vec_vsubudm (vector unsigned long long,
16248 vector unsigned long long);
16249 vector unsigned long long vec_vsubudm (vector bool long long,
16250 vector unsigned long long);
16251 vector unsigned long long vec_vsubudm (vector unsigned long long,
16252 vector bool long long);
16254 vector long long vec_vupkhsw (vector int);
16255 vector unsigned long long vec_vupkhsw (vector unsigned int);
16257 vector long long vec_vupklsw (vector int);
16258 vector unsigned long long vec_vupklsw (vector int);
16261 If the ISA 2.07 additions to the vector/scalar (power8-vector)
16262 instruction set is available, the following additional functions are
16263 available for 64-bit targets. New vector types
16264 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
16265 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
16268 The normal vector extract, and set operations work on
16269 @var{vector __int128_t} and @var{vector __uint128_t} types,
16270 but the index value must be 0.
16273 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
16274 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
16276 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
16277 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
16279 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
16280 vector __int128_t);
16281 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
16282 vector __uint128_t);
16284 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
16285 vector __int128_t);
16286 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
16287 vector __uint128_t);
16289 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
16290 vector __int128_t);
16291 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
16292 vector __uint128_t);
16294 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
16295 vector __int128_t);
16296 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
16297 vector __uint128_t);
16299 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
16300 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
16302 __int128_t vec_vsubuqm (__int128_t, __int128_t);
16303 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
16305 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
16306 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
16307 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
16308 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
16309 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
16310 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
16311 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
16312 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
16313 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
16314 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
16317 If the cryptographic instructions are enabled (@option{-mcrypto} or
16318 @option{-mcpu=power8}), the following builtins are enabled.
16321 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
16323 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
16324 vector unsigned long long);
16326 vector unsigned long long __builtin_crypto_vcipherlast
16327 (vector unsigned long long,
16328 vector unsigned long long);
16330 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
16331 vector unsigned long long);
16333 vector unsigned long long __builtin_crypto_vncipherlast
16334 (vector unsigned long long,
16335 vector unsigned long long);
16337 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
16338 vector unsigned char,
16339 vector unsigned char);
16341 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
16342 vector unsigned short,
16343 vector unsigned short);
16345 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
16346 vector unsigned int,
16347 vector unsigned int);
16349 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
16350 vector unsigned long long,
16351 vector unsigned long long);
16353 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
16354 vector unsigned char);
16356 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
16357 vector unsigned short);
16359 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
16360 vector unsigned int);
16362 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
16363 vector unsigned long long);
16365 vector unsigned long long __builtin_crypto_vshasigmad
16366 (vector unsigned long long, int, int);
16368 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
16372 The second argument to the @var{__builtin_crypto_vshasigmad} and
16373 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
16374 integer that is 0 or 1. The third argument to these builtin functions
16375 must be a constant integer in the range of 0 to 15.
16377 @node PowerPC Hardware Transactional Memory Built-in Functions
16378 @subsection PowerPC Hardware Transactional Memory Built-in Functions
16379 GCC provides two interfaces for accessing the Hardware Transactional
16380 Memory (HTM) instructions available on some of the PowerPC family
16381 of processors (eg, POWER8). The two interfaces come in a low level
16382 interface, consisting of built-in functions specific to PowerPC and a
16383 higher level interface consisting of inline functions that are common
16384 between PowerPC and S/390.
16386 @subsubsection PowerPC HTM Low Level Built-in Functions
16388 The following low level built-in functions are available with
16389 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
16390 They all generate the machine instruction that is part of the name.
16392 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
16393 the full 4-bit condition register value set by their associated hardware
16394 instruction. The header file @code{htmintrin.h} defines some macros that can
16395 be used to decipher the return value. The @code{__builtin_tbegin} builtin
16396 returns a simple true or false value depending on whether a transaction was
16397 successfully started or not. The arguments of the builtins match exactly the
16398 type and order of the associated hardware instruction's operands, except for
16399 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
16400 Refer to the ISA manual for a description of each instruction's operands.
16403 unsigned int __builtin_tbegin (unsigned int)
16404 unsigned int __builtin_tend (unsigned int)
16406 unsigned int __builtin_tabort (unsigned int)
16407 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
16408 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
16409 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
16410 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
16412 unsigned int __builtin_tcheck (void)
16413 unsigned int __builtin_treclaim (unsigned int)
16414 unsigned int __builtin_trechkpt (void)
16415 unsigned int __builtin_tsr (unsigned int)
16418 In addition to the above HTM built-ins, we have added built-ins for
16419 some common extended mnemonics of the HTM instructions:
16422 unsigned int __builtin_tendall (void)
16423 unsigned int __builtin_tresume (void)
16424 unsigned int __builtin_tsuspend (void)
16427 Note that the semantics of the above HTM builtins are required to mimic
16428 the locking semantics used for critical sections. Builtins that are used
16429 to create a new transaction or restart a suspended transaction must have
16430 lock acquisition like semantics while those builtins that end or suspend a
16431 transaction must have lock release like semantics. Specifically, this must
16432 mimic lock semantics as specified by C++11, for example: Lock acquisition is
16433 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
16434 that returns 0, and lock release is as-if an execution of
16435 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
16436 implicit implementation-defined lock used for all transactions. The HTM
16437 instructions associated with with the builtins inherently provide the
16438 correct acquisition and release hardware barriers required. However,
16439 the compiler must also be prohibited from moving loads and stores across
16440 the builtins in a way that would violate their semantics. This has been
16441 accomplished by adding memory barriers to the associated HTM instructions
16442 (which is a conservative approach to provide acquire and release semantics).
16443 Earlier versions of the compiler did not treat the HTM instructions as
16444 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
16445 be used to determine whether the current compiler treats HTM instructions
16446 as memory barriers or not. This allows the user to explicitly add memory
16447 barriers to their code when using an older version of the compiler.
16449 The following set of built-in functions are available to gain access
16450 to the HTM specific special purpose registers.
16453 unsigned long __builtin_get_texasr (void)
16454 unsigned long __builtin_get_texasru (void)
16455 unsigned long __builtin_get_tfhar (void)
16456 unsigned long __builtin_get_tfiar (void)
16458 void __builtin_set_texasr (unsigned long);
16459 void __builtin_set_texasru (unsigned long);
16460 void __builtin_set_tfhar (unsigned long);
16461 void __builtin_set_tfiar (unsigned long);
16464 Example usage of these low level built-in functions may look like:
16467 #include <htmintrin.h>
16469 int num_retries = 10;
16473 if (__builtin_tbegin (0))
16475 /* Transaction State Initiated. */
16476 if (is_locked (lock))
16477 __builtin_tabort (0);
16478 ... transaction code...
16479 __builtin_tend (0);
16484 /* Transaction State Failed. Use locks if the transaction
16485 failure is "persistent" or we've tried too many times. */
16486 if (num_retries-- <= 0
16487 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
16489 acquire_lock (lock);
16490 ... non transactional fallback path...
16491 release_lock (lock);
16498 One final built-in function has been added that returns the value of
16499 the 2-bit Transaction State field of the Machine Status Register (MSR)
16500 as stored in @code{CR0}.
16503 unsigned long __builtin_ttest (void)
16506 This built-in can be used to determine the current transaction state
16507 using the following code example:
16510 #include <htmintrin.h>
16512 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
16514 if (tx_state == _HTM_TRANSACTIONAL)
16516 /* Code to use in transactional state. */
16518 else if (tx_state == _HTM_NONTRANSACTIONAL)
16520 /* Code to use in non-transactional state. */
16522 else if (tx_state == _HTM_SUSPENDED)
16524 /* Code to use in transaction suspended state. */
16528 @subsubsection PowerPC HTM High Level Inline Functions
16530 The following high level HTM interface is made available by including
16531 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
16532 where CPU is `power8' or later. This interface is common between PowerPC
16533 and S/390, allowing users to write one HTM source implementation that
16534 can be compiled and executed on either system.
16537 long __TM_simple_begin (void)
16538 long __TM_begin (void* const TM_buff)
16539 long __TM_end (void)
16540 void __TM_abort (void)
16541 void __TM_named_abort (unsigned char const code)
16542 void __TM_resume (void)
16543 void __TM_suspend (void)
16545 long __TM_is_user_abort (void* const TM_buff)
16546 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
16547 long __TM_is_illegal (void* const TM_buff)
16548 long __TM_is_footprint_exceeded (void* const TM_buff)
16549 long __TM_nesting_depth (void* const TM_buff)
16550 long __TM_is_nested_too_deep(void* const TM_buff)
16551 long __TM_is_conflict(void* const TM_buff)
16552 long __TM_is_failure_persistent(void* const TM_buff)
16553 long __TM_failure_address(void* const TM_buff)
16554 long long __TM_failure_code(void* const TM_buff)
16557 Using these common set of HTM inline functions, we can create
16558 a more portable version of the HTM example in the previous
16559 section that will work on either PowerPC or S/390:
16562 #include <htmxlintrin.h>
16564 int num_retries = 10;
16565 TM_buff_type TM_buff;
16569 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
16571 /* Transaction State Initiated. */
16572 if (is_locked (lock))
16574 ... transaction code...
16580 /* Transaction State Failed. Use locks if the transaction
16581 failure is "persistent" or we've tried too many times. */
16582 if (num_retries-- <= 0
16583 || __TM_is_failure_persistent (TM_buff))
16585 acquire_lock (lock);
16586 ... non transactional fallback path...
16587 release_lock (lock);
16594 @node RX Built-in Functions
16595 @subsection RX Built-in Functions
16596 GCC supports some of the RX instructions which cannot be expressed in
16597 the C programming language via the use of built-in functions. The
16598 following functions are supported:
16600 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
16601 Generates the @code{brk} machine instruction.
16604 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
16605 Generates the @code{clrpsw} machine instruction to clear the specified
16606 bit in the processor status word.
16609 @deftypefn {Built-in Function} void __builtin_rx_int (int)
16610 Generates the @code{int} machine instruction to generate an interrupt
16611 with the specified value.
16614 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
16615 Generates the @code{machi} machine instruction to add the result of
16616 multiplying the top 16 bits of the two arguments into the
16620 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
16621 Generates the @code{maclo} machine instruction to add the result of
16622 multiplying the bottom 16 bits of the two arguments into the
16626 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
16627 Generates the @code{mulhi} machine instruction to place the result of
16628 multiplying the top 16 bits of the two arguments into the
16632 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
16633 Generates the @code{mullo} machine instruction to place the result of
16634 multiplying the bottom 16 bits of the two arguments into the
16638 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
16639 Generates the @code{mvfachi} machine instruction to read the top
16640 32 bits of the accumulator.
16643 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
16644 Generates the @code{mvfacmi} machine instruction to read the middle
16645 32 bits of the accumulator.
16648 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
16649 Generates the @code{mvfc} machine instruction which reads the control
16650 register specified in its argument and returns its value.
16653 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
16654 Generates the @code{mvtachi} machine instruction to set the top
16655 32 bits of the accumulator.
16658 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
16659 Generates the @code{mvtaclo} machine instruction to set the bottom
16660 32 bits of the accumulator.
16663 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
16664 Generates the @code{mvtc} machine instruction which sets control
16665 register number @code{reg} to @code{val}.
16668 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
16669 Generates the @code{mvtipl} machine instruction set the interrupt
16673 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
16674 Generates the @code{racw} machine instruction to round the accumulator
16675 according to the specified mode.
16678 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
16679 Generates the @code{revw} machine instruction which swaps the bytes in
16680 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
16681 and also bits 16--23 occupy bits 24--31 and vice versa.
16684 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
16685 Generates the @code{rmpa} machine instruction which initiates a
16686 repeated multiply and accumulate sequence.
16689 @deftypefn {Built-in Function} void __builtin_rx_round (float)
16690 Generates the @code{round} machine instruction which returns the
16691 floating-point argument rounded according to the current rounding mode
16692 set in the floating-point status word register.
16695 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
16696 Generates the @code{sat} machine instruction which returns the
16697 saturated value of the argument.
16700 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
16701 Generates the @code{setpsw} machine instruction to set the specified
16702 bit in the processor status word.
16705 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
16706 Generates the @code{wait} machine instruction.
16709 @node S/390 System z Built-in Functions
16710 @subsection S/390 System z Built-in Functions
16711 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
16712 Generates the @code{tbegin} machine instruction starting a
16713 non-constrained hardware transaction. If the parameter is non-NULL the
16714 memory area is used to store the transaction diagnostic buffer and
16715 will be passed as first operand to @code{tbegin}. This buffer can be
16716 defined using the @code{struct __htm_tdb} C struct defined in
16717 @code{htmintrin.h} and must reside on a double-word boundary. The
16718 second tbegin operand is set to @code{0xff0c}. This enables
16719 save/restore of all GPRs and disables aborts for FPR and AR
16720 manipulations inside the transaction body. The condition code set by
16721 the tbegin instruction is returned as integer value. The tbegin
16722 instruction by definition overwrites the content of all FPRs. The
16723 compiler will generate code which saves and restores the FPRs. For
16724 soft-float code it is recommended to used the @code{*_nofloat}
16725 variant. In order to prevent a TDB from being written it is required
16726 to pass a constant zero value as parameter. Passing a zero value
16727 through a variable is not sufficient. Although modifications of
16728 access registers inside the transaction will not trigger an
16729 transaction abort it is not supported to actually modify them. Access
16730 registers do not get saved when entering a transaction. They will have
16731 undefined state when reaching the abort code.
16734 Macros for the possible return codes of tbegin are defined in the
16735 @code{htmintrin.h} header file:
16738 @item _HTM_TBEGIN_STARTED
16739 @code{tbegin} has been executed as part of normal processing. The
16740 transaction body is supposed to be executed.
16741 @item _HTM_TBEGIN_INDETERMINATE
16742 The transaction was aborted due to an indeterminate condition which
16743 might be persistent.
16744 @item _HTM_TBEGIN_TRANSIENT
16745 The transaction aborted due to a transient failure. The transaction
16746 should be re-executed in that case.
16747 @item _HTM_TBEGIN_PERSISTENT
16748 The transaction aborted due to a persistent failure. Re-execution
16749 under same circumstances will not be productive.
16752 @defmac _HTM_FIRST_USER_ABORT_CODE
16753 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
16754 specifies the first abort code which can be used for
16755 @code{__builtin_tabort}. Values below this threshold are reserved for
16759 @deftp {Data type} {struct __htm_tdb}
16760 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
16761 the structure of the transaction diagnostic block as specified in the
16762 Principles of Operation manual chapter 5-91.
16765 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
16766 Same as @code{__builtin_tbegin} but without FPR saves and restores.
16767 Using this variant in code making use of FPRs will leave the FPRs in
16768 undefined state when entering the transaction abort handler code.
16771 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
16772 In addition to @code{__builtin_tbegin} a loop for transient failures
16773 is generated. If tbegin returns a condition code of 2 the transaction
16774 will be retried as often as specified in the second argument. The
16775 perform processor assist instruction is used to tell the CPU about the
16776 number of fails so far.
16779 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
16780 Same as @code{__builtin_tbegin_retry} but without FPR saves and
16781 restores. Using this variant in code making use of FPRs will leave
16782 the FPRs in undefined state when entering the transaction abort
16786 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
16787 Generates the @code{tbeginc} machine instruction starting a constrained
16788 hardware transaction. The second operand is set to @code{0xff08}.
16791 @deftypefn {Built-in Function} int __builtin_tend (void)
16792 Generates the @code{tend} machine instruction finishing a transaction
16793 and making the changes visible to other threads. The condition code
16794 generated by tend is returned as integer value.
16797 @deftypefn {Built-in Function} void __builtin_tabort (int)
16798 Generates the @code{tabort} machine instruction with the specified
16799 abort code. Abort codes from 0 through 255 are reserved and will
16800 result in an error message.
16803 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
16804 Generates the @code{ppa rX,rY,1} machine instruction. Where the
16805 integer parameter is loaded into rX and a value of zero is loaded into
16806 rY. The integer parameter specifies the number of times the
16807 transaction repeatedly aborted.
16810 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
16811 Generates the @code{etnd} machine instruction. The current nesting
16812 depth is returned as integer value. For a nesting depth of 0 the code
16813 is not executed as part of an transaction.
16816 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
16818 Generates the @code{ntstg} machine instruction. The second argument
16819 is written to the first arguments location. The store operation will
16820 not be rolled-back in case of an transaction abort.
16823 @node SH Built-in Functions
16824 @subsection SH Built-in Functions
16825 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
16826 families of processors:
16828 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
16829 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
16830 used by system code that manages threads and execution contexts. The compiler
16831 normally does not generate code that modifies the contents of @samp{GBR} and
16832 thus the value is preserved across function calls. Changing the @samp{GBR}
16833 value in user code must be done with caution, since the compiler might use
16834 @samp{GBR} in order to access thread local variables.
16838 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
16839 Returns the value that is currently set in the @samp{GBR} register.
16840 Memory loads and stores that use the thread pointer as a base address are
16841 turned into @samp{GBR} based displacement loads and stores, if possible.
16849 int get_tcb_value (void)
16851 // Generate @samp{mov.l @@(8,gbr),r0} instruction
16852 return ((my_tcb*)__builtin_thread_pointer ())->c;
16858 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
16859 Returns the value that is currently set in the @samp{FPSCR} register.
16862 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
16863 Sets the @samp{FPSCR} register to the specified value @var{val}, while
16864 preserving the current values of the FR, SZ and PR bits.
16867 @node SPARC VIS Built-in Functions
16868 @subsection SPARC VIS Built-in Functions
16870 GCC supports SIMD operations on the SPARC using both the generic vector
16871 extensions (@pxref{Vector Extensions}) as well as built-in functions for
16872 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
16873 switch, the VIS extension is exposed as the following built-in functions:
16876 typedef int v1si __attribute__ ((vector_size (4)));
16877 typedef int v2si __attribute__ ((vector_size (8)));
16878 typedef short v4hi __attribute__ ((vector_size (8)));
16879 typedef short v2hi __attribute__ ((vector_size (4)));
16880 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
16881 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
16883 void __builtin_vis_write_gsr (int64_t);
16884 int64_t __builtin_vis_read_gsr (void);
16886 void * __builtin_vis_alignaddr (void *, long);
16887 void * __builtin_vis_alignaddrl (void *, long);
16888 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
16889 v2si __builtin_vis_faligndatav2si (v2si, v2si);
16890 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
16891 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
16893 v4hi __builtin_vis_fexpand (v4qi);
16895 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
16896 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
16897 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
16898 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
16899 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
16900 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
16901 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
16903 v4qi __builtin_vis_fpack16 (v4hi);
16904 v8qi __builtin_vis_fpack32 (v2si, v8qi);
16905 v2hi __builtin_vis_fpackfix (v2si);
16906 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
16908 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
16910 long __builtin_vis_edge8 (void *, void *);
16911 long __builtin_vis_edge8l (void *, void *);
16912 long __builtin_vis_edge16 (void *, void *);
16913 long __builtin_vis_edge16l (void *, void *);
16914 long __builtin_vis_edge32 (void *, void *);
16915 long __builtin_vis_edge32l (void *, void *);
16917 long __builtin_vis_fcmple16 (v4hi, v4hi);
16918 long __builtin_vis_fcmple32 (v2si, v2si);
16919 long __builtin_vis_fcmpne16 (v4hi, v4hi);
16920 long __builtin_vis_fcmpne32 (v2si, v2si);
16921 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
16922 long __builtin_vis_fcmpgt32 (v2si, v2si);
16923 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
16924 long __builtin_vis_fcmpeq32 (v2si, v2si);
16926 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
16927 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
16928 v2si __builtin_vis_fpadd32 (v2si, v2si);
16929 v1si __builtin_vis_fpadd32s (v1si, v1si);
16930 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
16931 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
16932 v2si __builtin_vis_fpsub32 (v2si, v2si);
16933 v1si __builtin_vis_fpsub32s (v1si, v1si);
16935 long __builtin_vis_array8 (long, long);
16936 long __builtin_vis_array16 (long, long);
16937 long __builtin_vis_array32 (long, long);
16940 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
16941 functions also become available:
16944 long __builtin_vis_bmask (long, long);
16945 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
16946 v2si __builtin_vis_bshufflev2si (v2si, v2si);
16947 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
16948 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
16950 long __builtin_vis_edge8n (void *, void *);
16951 long __builtin_vis_edge8ln (void *, void *);
16952 long __builtin_vis_edge16n (void *, void *);
16953 long __builtin_vis_edge16ln (void *, void *);
16954 long __builtin_vis_edge32n (void *, void *);
16955 long __builtin_vis_edge32ln (void *, void *);
16958 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
16959 functions also become available:
16962 void __builtin_vis_cmask8 (long);
16963 void __builtin_vis_cmask16 (long);
16964 void __builtin_vis_cmask32 (long);
16966 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
16968 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
16969 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
16970 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
16971 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
16972 v2si __builtin_vis_fsll16 (v2si, v2si);
16973 v2si __builtin_vis_fslas16 (v2si, v2si);
16974 v2si __builtin_vis_fsrl16 (v2si, v2si);
16975 v2si __builtin_vis_fsra16 (v2si, v2si);
16977 long __builtin_vis_pdistn (v8qi, v8qi);
16979 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
16981 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
16982 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
16984 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
16985 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
16986 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
16987 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
16988 v2si __builtin_vis_fpadds32 (v2si, v2si);
16989 v1si __builtin_vis_fpadds32s (v1si, v1si);
16990 v2si __builtin_vis_fpsubs32 (v2si, v2si);
16991 v1si __builtin_vis_fpsubs32s (v1si, v1si);
16993 long __builtin_vis_fucmple8 (v8qi, v8qi);
16994 long __builtin_vis_fucmpne8 (v8qi, v8qi);
16995 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
16996 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
16998 float __builtin_vis_fhadds (float, float);
16999 double __builtin_vis_fhaddd (double, double);
17000 float __builtin_vis_fhsubs (float, float);
17001 double __builtin_vis_fhsubd (double, double);
17002 float __builtin_vis_fnhadds (float, float);
17003 double __builtin_vis_fnhaddd (double, double);
17005 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
17006 int64_t __builtin_vis_xmulx (int64_t, int64_t);
17007 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
17010 @node SPU Built-in Functions
17011 @subsection SPU Built-in Functions
17013 GCC provides extensions for the SPU processor as described in the
17014 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
17015 found at @uref{http://cell.scei.co.jp/} or
17016 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
17017 implementation differs in several ways.
17022 The optional extension of specifying vector constants in parentheses is
17026 A vector initializer requires no cast if the vector constant is of the
17027 same type as the variable it is initializing.
17030 If @code{signed} or @code{unsigned} is omitted, the signedness of the
17031 vector type is the default signedness of the base type. The default
17032 varies depending on the operating system, so a portable program should
17033 always specify the signedness.
17036 By default, the keyword @code{__vector} is added. The macro
17037 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
17041 GCC allows using a @code{typedef} name as the type specifier for a
17045 For C, overloaded functions are implemented with macros so the following
17049 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
17053 Since @code{spu_add} is a macro, the vector constant in the example
17054 is treated as four separate arguments. Wrap the entire argument in
17055 parentheses for this to work.
17058 The extended version of @code{__builtin_expect} is not supported.
17062 @emph{Note:} Only the interface described in the aforementioned
17063 specification is supported. Internally, GCC uses built-in functions to
17064 implement the required functionality, but these are not supported and
17065 are subject to change without notice.
17067 @node TI C6X Built-in Functions
17068 @subsection TI C6X Built-in Functions
17070 GCC provides intrinsics to access certain instructions of the TI C6X
17071 processors. These intrinsics, listed below, are available after
17072 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
17073 to C6X instructions.
17077 int _sadd (int, int)
17078 int _ssub (int, int)
17079 int _sadd2 (int, int)
17080 int _ssub2 (int, int)
17081 long long _mpy2 (int, int)
17082 long long _smpy2 (int, int)
17083 int _add4 (int, int)
17084 int _sub4 (int, int)
17085 int _saddu4 (int, int)
17087 int _smpy (int, int)
17088 int _smpyh (int, int)
17089 int _smpyhl (int, int)
17090 int _smpylh (int, int)
17092 int _sshl (int, int)
17093 int _subc (int, int)
17095 int _avg2 (int, int)
17096 int _avgu4 (int, int)
17098 int _clrr (int, int)
17099 int _extr (int, int)
17100 int _extru (int, int)
17106 @node TILE-Gx Built-in Functions
17107 @subsection TILE-Gx Built-in Functions
17109 GCC provides intrinsics to access every instruction of the TILE-Gx
17110 processor. The intrinsics are of the form:
17114 unsigned long long __insn_@var{op} (...)
17118 Where @var{op} is the name of the instruction. Refer to the ISA manual
17119 for the complete list of instructions.
17121 GCC also provides intrinsics to directly access the network registers.
17122 The intrinsics are:
17126 unsigned long long __tile_idn0_receive (void)
17127 unsigned long long __tile_idn1_receive (void)
17128 unsigned long long __tile_udn0_receive (void)
17129 unsigned long long __tile_udn1_receive (void)
17130 unsigned long long __tile_udn2_receive (void)
17131 unsigned long long __tile_udn3_receive (void)
17132 void __tile_idn_send (unsigned long long)
17133 void __tile_udn_send (unsigned long long)
17137 The intrinsic @code{void __tile_network_barrier (void)} is used to
17138 guarantee that no network operations before it are reordered with
17141 @node TILEPro Built-in Functions
17142 @subsection TILEPro Built-in Functions
17144 GCC provides intrinsics to access every instruction of the TILEPro
17145 processor. The intrinsics are of the form:
17149 unsigned __insn_@var{op} (...)
17154 where @var{op} is the name of the instruction. Refer to the ISA manual
17155 for the complete list of instructions.
17157 GCC also provides intrinsics to directly access the network registers.
17158 The intrinsics are:
17162 unsigned __tile_idn0_receive (void)
17163 unsigned __tile_idn1_receive (void)
17164 unsigned __tile_sn_receive (void)
17165 unsigned __tile_udn0_receive (void)
17166 unsigned __tile_udn1_receive (void)
17167 unsigned __tile_udn2_receive (void)
17168 unsigned __tile_udn3_receive (void)
17169 void __tile_idn_send (unsigned)
17170 void __tile_sn_send (unsigned)
17171 void __tile_udn_send (unsigned)
17175 The intrinsic @code{void __tile_network_barrier (void)} is used to
17176 guarantee that no network operations before it are reordered with
17179 @node x86 Built-in Functions
17180 @subsection x86 Built-in Functions
17182 These built-in functions are available for the x86-32 and x86-64 family
17183 of computers, depending on the command-line switches used.
17185 If you specify command-line switches such as @option{-msse},
17186 the compiler could use the extended instruction sets even if the built-ins
17187 are not used explicitly in the program. For this reason, applications
17188 that perform run-time CPU detection must compile separate files for each
17189 supported architecture, using the appropriate flags. In particular,
17190 the file containing the CPU detection code should be compiled without
17193 The following machine modes are available for use with MMX built-in functions
17194 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
17195 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
17196 vector of eight 8-bit integers. Some of the built-in functions operate on
17197 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
17199 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
17200 of two 32-bit floating-point values.
17202 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
17203 floating-point values. Some instructions use a vector of four 32-bit
17204 integers, these use @code{V4SI}. Finally, some instructions operate on an
17205 entire vector register, interpreting it as a 128-bit integer, these use mode
17208 In 64-bit mode, the x86-64 family of processors uses additional built-in
17209 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
17210 floating point and @code{TC} 128-bit complex floating-point values.
17212 The following floating-point built-in functions are available in 64-bit
17213 mode. All of them implement the function that is part of the name.
17216 __float128 __builtin_fabsq (__float128)
17217 __float128 __builtin_copysignq (__float128, __float128)
17220 The following built-in function is always available.
17223 @item void __builtin_ia32_pause (void)
17224 Generates the @code{pause} machine instruction with a compiler memory
17228 The following floating-point built-in functions are made available in the
17232 @item __float128 __builtin_infq (void)
17233 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
17234 @findex __builtin_infq
17236 @item __float128 __builtin_huge_valq (void)
17237 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
17238 @findex __builtin_huge_valq
17241 The following built-in functions are always available and can be used to
17242 check the target platform type.
17244 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
17245 This function runs the CPU detection code to check the type of CPU and the
17246 features supported. This built-in function needs to be invoked along with the built-in functions
17247 to check CPU type and features, @code{__builtin_cpu_is} and
17248 @code{__builtin_cpu_supports}, only when used in a function that is
17249 executed before any constructors are called. The CPU detection code is
17250 automatically executed in a very high priority constructor.
17252 For example, this function has to be used in @code{ifunc} resolvers that
17253 check for CPU type using the built-in functions @code{__builtin_cpu_is}
17254 and @code{__builtin_cpu_supports}, or in constructors on targets that
17255 don't support constructor priority.
17258 static void (*resolve_memcpy (void)) (void)
17260 // ifunc resolvers fire before constructors, explicitly call the init
17262 __builtin_cpu_init ();
17263 if (__builtin_cpu_supports ("ssse3"))
17264 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
17266 return default_memcpy;
17269 void *memcpy (void *, const void *, size_t)
17270 __attribute__ ((ifunc ("resolve_memcpy")));
17275 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
17276 This function returns a positive integer if the run-time CPU
17277 is of type @var{cpuname}
17278 and returns @code{0} otherwise. The following CPU names can be detected:
17294 Intel Core i7 Nehalem CPU.
17297 Intel Core i7 Westmere CPU.
17300 Intel Core i7 Sandy Bridge CPU.
17306 AMD Family 10h CPU.
17309 AMD Family 10h Barcelona CPU.
17312 AMD Family 10h Shanghai CPU.
17315 AMD Family 10h Istanbul CPU.
17318 AMD Family 14h CPU.
17321 AMD Family 15h CPU.
17324 AMD Family 15h Bulldozer version 1.
17327 AMD Family 15h Bulldozer version 2.
17330 AMD Family 15h Bulldozer version 3.
17333 AMD Family 15h Bulldozer version 4.
17336 AMD Family 16h CPU.
17339 AMD Family 17h CPU.
17342 Here is an example:
17344 if (__builtin_cpu_is ("corei7"))
17346 do_corei7 (); // Core i7 specific implementation.
17350 do_generic (); // Generic implementation.
17355 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
17356 This function returns a positive integer if the run-time CPU
17357 supports @var{feature}
17358 and returns @code{0} otherwise. The following features can be detected:
17366 POPCNT instruction.
17374 SSSE3 instructions.
17376 SSE4.1 instructions.
17378 SSE4.2 instructions.
17384 AVX512F instructions.
17387 Here is an example:
17389 if (__builtin_cpu_supports ("popcnt"))
17391 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
17395 count = generic_countbits (n); //generic implementation.
17401 The following built-in functions are made available by @option{-mmmx}.
17402 All of them generate the machine instruction that is part of the name.
17405 v8qi __builtin_ia32_paddb (v8qi, v8qi)
17406 v4hi __builtin_ia32_paddw (v4hi, v4hi)
17407 v2si __builtin_ia32_paddd (v2si, v2si)
17408 v8qi __builtin_ia32_psubb (v8qi, v8qi)
17409 v4hi __builtin_ia32_psubw (v4hi, v4hi)
17410 v2si __builtin_ia32_psubd (v2si, v2si)
17411 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
17412 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
17413 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
17414 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
17415 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
17416 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
17417 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
17418 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
17419 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
17420 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
17421 di __builtin_ia32_pand (di, di)
17422 di __builtin_ia32_pandn (di,di)
17423 di __builtin_ia32_por (di, di)
17424 di __builtin_ia32_pxor (di, di)
17425 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
17426 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
17427 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
17428 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
17429 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
17430 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
17431 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
17432 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
17433 v2si __builtin_ia32_punpckhdq (v2si, v2si)
17434 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
17435 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
17436 v2si __builtin_ia32_punpckldq (v2si, v2si)
17437 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
17438 v4hi __builtin_ia32_packssdw (v2si, v2si)
17439 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
17441 v4hi __builtin_ia32_psllw (v4hi, v4hi)
17442 v2si __builtin_ia32_pslld (v2si, v2si)
17443 v1di __builtin_ia32_psllq (v1di, v1di)
17444 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
17445 v2si __builtin_ia32_psrld (v2si, v2si)
17446 v1di __builtin_ia32_psrlq (v1di, v1di)
17447 v4hi __builtin_ia32_psraw (v4hi, v4hi)
17448 v2si __builtin_ia32_psrad (v2si, v2si)
17449 v4hi __builtin_ia32_psllwi (v4hi, int)
17450 v2si __builtin_ia32_pslldi (v2si, int)
17451 v1di __builtin_ia32_psllqi (v1di, int)
17452 v4hi __builtin_ia32_psrlwi (v4hi, int)
17453 v2si __builtin_ia32_psrldi (v2si, int)
17454 v1di __builtin_ia32_psrlqi (v1di, int)
17455 v4hi __builtin_ia32_psrawi (v4hi, int)
17456 v2si __builtin_ia32_psradi (v2si, int)
17460 The following built-in functions are made available either with
17461 @option{-msse}, or with a combination of @option{-m3dnow} and
17462 @option{-march=athlon}. All of them generate the machine
17463 instruction that is part of the name.
17466 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
17467 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
17468 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
17469 v1di __builtin_ia32_psadbw (v8qi, v8qi)
17470 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
17471 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
17472 v8qi __builtin_ia32_pminub (v8qi, v8qi)
17473 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
17474 int __builtin_ia32_pmovmskb (v8qi)
17475 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
17476 void __builtin_ia32_movntq (di *, di)
17477 void __builtin_ia32_sfence (void)
17480 The following built-in functions are available when @option{-msse} is used.
17481 All of them generate the machine instruction that is part of the name.
17484 int __builtin_ia32_comieq (v4sf, v4sf)
17485 int __builtin_ia32_comineq (v4sf, v4sf)
17486 int __builtin_ia32_comilt (v4sf, v4sf)
17487 int __builtin_ia32_comile (v4sf, v4sf)
17488 int __builtin_ia32_comigt (v4sf, v4sf)
17489 int __builtin_ia32_comige (v4sf, v4sf)
17490 int __builtin_ia32_ucomieq (v4sf, v4sf)
17491 int __builtin_ia32_ucomineq (v4sf, v4sf)
17492 int __builtin_ia32_ucomilt (v4sf, v4sf)
17493 int __builtin_ia32_ucomile (v4sf, v4sf)
17494 int __builtin_ia32_ucomigt (v4sf, v4sf)
17495 int __builtin_ia32_ucomige (v4sf, v4sf)
17496 v4sf __builtin_ia32_addps (v4sf, v4sf)
17497 v4sf __builtin_ia32_subps (v4sf, v4sf)
17498 v4sf __builtin_ia32_mulps (v4sf, v4sf)
17499 v4sf __builtin_ia32_divps (v4sf, v4sf)
17500 v4sf __builtin_ia32_addss (v4sf, v4sf)
17501 v4sf __builtin_ia32_subss (v4sf, v4sf)
17502 v4sf __builtin_ia32_mulss (v4sf, v4sf)
17503 v4sf __builtin_ia32_divss (v4sf, v4sf)
17504 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
17505 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
17506 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
17507 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
17508 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
17509 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
17510 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
17511 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
17512 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
17513 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
17514 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
17515 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
17516 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
17517 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
17518 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
17519 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
17520 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
17521 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
17522 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
17523 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
17524 v4sf __builtin_ia32_maxps (v4sf, v4sf)
17525 v4sf __builtin_ia32_maxss (v4sf, v4sf)
17526 v4sf __builtin_ia32_minps (v4sf, v4sf)
17527 v4sf __builtin_ia32_minss (v4sf, v4sf)
17528 v4sf __builtin_ia32_andps (v4sf, v4sf)
17529 v4sf __builtin_ia32_andnps (v4sf, v4sf)
17530 v4sf __builtin_ia32_orps (v4sf, v4sf)
17531 v4sf __builtin_ia32_xorps (v4sf, v4sf)
17532 v4sf __builtin_ia32_movss (v4sf, v4sf)
17533 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
17534 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
17535 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
17536 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
17537 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
17538 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
17539 v2si __builtin_ia32_cvtps2pi (v4sf)
17540 int __builtin_ia32_cvtss2si (v4sf)
17541 v2si __builtin_ia32_cvttps2pi (v4sf)
17542 int __builtin_ia32_cvttss2si (v4sf)
17543 v4sf __builtin_ia32_rcpps (v4sf)
17544 v4sf __builtin_ia32_rsqrtps (v4sf)
17545 v4sf __builtin_ia32_sqrtps (v4sf)
17546 v4sf __builtin_ia32_rcpss (v4sf)
17547 v4sf __builtin_ia32_rsqrtss (v4sf)
17548 v4sf __builtin_ia32_sqrtss (v4sf)
17549 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
17550 void __builtin_ia32_movntps (float *, v4sf)
17551 int __builtin_ia32_movmskps (v4sf)
17554 The following built-in functions are available when @option{-msse} is used.
17557 @item v4sf __builtin_ia32_loadups (float *)
17558 Generates the @code{movups} machine instruction as a load from memory.
17559 @item void __builtin_ia32_storeups (float *, v4sf)
17560 Generates the @code{movups} machine instruction as a store to memory.
17561 @item v4sf __builtin_ia32_loadss (float *)
17562 Generates the @code{movss} machine instruction as a load from memory.
17563 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
17564 Generates the @code{movhps} machine instruction as a load from memory.
17565 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
17566 Generates the @code{movlps} machine instruction as a load from memory
17567 @item void __builtin_ia32_storehps (v2sf *, v4sf)
17568 Generates the @code{movhps} machine instruction as a store to memory.
17569 @item void __builtin_ia32_storelps (v2sf *, v4sf)
17570 Generates the @code{movlps} machine instruction as a store to memory.
17573 The following built-in functions are available when @option{-msse2} is used.
17574 All of them generate the machine instruction that is part of the name.
17577 int __builtin_ia32_comisdeq (v2df, v2df)
17578 int __builtin_ia32_comisdlt (v2df, v2df)
17579 int __builtin_ia32_comisdle (v2df, v2df)
17580 int __builtin_ia32_comisdgt (v2df, v2df)
17581 int __builtin_ia32_comisdge (v2df, v2df)
17582 int __builtin_ia32_comisdneq (v2df, v2df)
17583 int __builtin_ia32_ucomisdeq (v2df, v2df)
17584 int __builtin_ia32_ucomisdlt (v2df, v2df)
17585 int __builtin_ia32_ucomisdle (v2df, v2df)
17586 int __builtin_ia32_ucomisdgt (v2df, v2df)
17587 int __builtin_ia32_ucomisdge (v2df, v2df)
17588 int __builtin_ia32_ucomisdneq (v2df, v2df)
17589 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
17590 v2df __builtin_ia32_cmpltpd (v2df, v2df)
17591 v2df __builtin_ia32_cmplepd (v2df, v2df)
17592 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
17593 v2df __builtin_ia32_cmpgepd (v2df, v2df)
17594 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
17595 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
17596 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
17597 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
17598 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
17599 v2df __builtin_ia32_cmpngepd (v2df, v2df)
17600 v2df __builtin_ia32_cmpordpd (v2df, v2df)
17601 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
17602 v2df __builtin_ia32_cmpltsd (v2df, v2df)
17603 v2df __builtin_ia32_cmplesd (v2df, v2df)
17604 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
17605 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
17606 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
17607 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
17608 v2df __builtin_ia32_cmpordsd (v2df, v2df)
17609 v2di __builtin_ia32_paddq (v2di, v2di)
17610 v2di __builtin_ia32_psubq (v2di, v2di)
17611 v2df __builtin_ia32_addpd (v2df, v2df)
17612 v2df __builtin_ia32_subpd (v2df, v2df)
17613 v2df __builtin_ia32_mulpd (v2df, v2df)
17614 v2df __builtin_ia32_divpd (v2df, v2df)
17615 v2df __builtin_ia32_addsd (v2df, v2df)
17616 v2df __builtin_ia32_subsd (v2df, v2df)
17617 v2df __builtin_ia32_mulsd (v2df, v2df)
17618 v2df __builtin_ia32_divsd (v2df, v2df)
17619 v2df __builtin_ia32_minpd (v2df, v2df)
17620 v2df __builtin_ia32_maxpd (v2df, v2df)
17621 v2df __builtin_ia32_minsd (v2df, v2df)
17622 v2df __builtin_ia32_maxsd (v2df, v2df)
17623 v2df __builtin_ia32_andpd (v2df, v2df)
17624 v2df __builtin_ia32_andnpd (v2df, v2df)
17625 v2df __builtin_ia32_orpd (v2df, v2df)
17626 v2df __builtin_ia32_xorpd (v2df, v2df)
17627 v2df __builtin_ia32_movsd (v2df, v2df)
17628 v2df __builtin_ia32_unpckhpd (v2df, v2df)
17629 v2df __builtin_ia32_unpcklpd (v2df, v2df)
17630 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
17631 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
17632 v4si __builtin_ia32_paddd128 (v4si, v4si)
17633 v2di __builtin_ia32_paddq128 (v2di, v2di)
17634 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
17635 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
17636 v4si __builtin_ia32_psubd128 (v4si, v4si)
17637 v2di __builtin_ia32_psubq128 (v2di, v2di)
17638 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
17639 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
17640 v2di __builtin_ia32_pand128 (v2di, v2di)
17641 v2di __builtin_ia32_pandn128 (v2di, v2di)
17642 v2di __builtin_ia32_por128 (v2di, v2di)
17643 v2di __builtin_ia32_pxor128 (v2di, v2di)
17644 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
17645 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
17646 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
17647 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
17648 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
17649 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
17650 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
17651 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
17652 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
17653 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
17654 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
17655 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
17656 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
17657 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
17658 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
17659 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
17660 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
17661 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
17662 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
17663 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
17664 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
17665 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
17666 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
17667 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
17668 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
17669 v2df __builtin_ia32_loadupd (double *)
17670 void __builtin_ia32_storeupd (double *, v2df)
17671 v2df __builtin_ia32_loadhpd (v2df, double const *)
17672 v2df __builtin_ia32_loadlpd (v2df, double const *)
17673 int __builtin_ia32_movmskpd (v2df)
17674 int __builtin_ia32_pmovmskb128 (v16qi)
17675 void __builtin_ia32_movnti (int *, int)
17676 void __builtin_ia32_movnti64 (long long int *, long long int)
17677 void __builtin_ia32_movntpd (double *, v2df)
17678 void __builtin_ia32_movntdq (v2df *, v2df)
17679 v4si __builtin_ia32_pshufd (v4si, int)
17680 v8hi __builtin_ia32_pshuflw (v8hi, int)
17681 v8hi __builtin_ia32_pshufhw (v8hi, int)
17682 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
17683 v2df __builtin_ia32_sqrtpd (v2df)
17684 v2df __builtin_ia32_sqrtsd (v2df)
17685 v2df __builtin_ia32_shufpd (v2df, v2df, int)
17686 v2df __builtin_ia32_cvtdq2pd (v4si)
17687 v4sf __builtin_ia32_cvtdq2ps (v4si)
17688 v4si __builtin_ia32_cvtpd2dq (v2df)
17689 v2si __builtin_ia32_cvtpd2pi (v2df)
17690 v4sf __builtin_ia32_cvtpd2ps (v2df)
17691 v4si __builtin_ia32_cvttpd2dq (v2df)
17692 v2si __builtin_ia32_cvttpd2pi (v2df)
17693 v2df __builtin_ia32_cvtpi2pd (v2si)
17694 int __builtin_ia32_cvtsd2si (v2df)
17695 int __builtin_ia32_cvttsd2si (v2df)
17696 long long __builtin_ia32_cvtsd2si64 (v2df)
17697 long long __builtin_ia32_cvttsd2si64 (v2df)
17698 v4si __builtin_ia32_cvtps2dq (v4sf)
17699 v2df __builtin_ia32_cvtps2pd (v4sf)
17700 v4si __builtin_ia32_cvttps2dq (v4sf)
17701 v2df __builtin_ia32_cvtsi2sd (v2df, int)
17702 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
17703 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
17704 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
17705 void __builtin_ia32_clflush (const void *)
17706 void __builtin_ia32_lfence (void)
17707 void __builtin_ia32_mfence (void)
17708 v16qi __builtin_ia32_loaddqu (const char *)
17709 void __builtin_ia32_storedqu (char *, v16qi)
17710 v1di __builtin_ia32_pmuludq (v2si, v2si)
17711 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
17712 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
17713 v4si __builtin_ia32_pslld128 (v4si, v4si)
17714 v2di __builtin_ia32_psllq128 (v2di, v2di)
17715 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
17716 v4si __builtin_ia32_psrld128 (v4si, v4si)
17717 v2di __builtin_ia32_psrlq128 (v2di, v2di)
17718 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
17719 v4si __builtin_ia32_psrad128 (v4si, v4si)
17720 v2di __builtin_ia32_pslldqi128 (v2di, int)
17721 v8hi __builtin_ia32_psllwi128 (v8hi, int)
17722 v4si __builtin_ia32_pslldi128 (v4si, int)
17723 v2di __builtin_ia32_psllqi128 (v2di, int)
17724 v2di __builtin_ia32_psrldqi128 (v2di, int)
17725 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
17726 v4si __builtin_ia32_psrldi128 (v4si, int)
17727 v2di __builtin_ia32_psrlqi128 (v2di, int)
17728 v8hi __builtin_ia32_psrawi128 (v8hi, int)
17729 v4si __builtin_ia32_psradi128 (v4si, int)
17730 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
17731 v2di __builtin_ia32_movq128 (v2di)
17734 The following built-in functions are available when @option{-msse3} is used.
17735 All of them generate the machine instruction that is part of the name.
17738 v2df __builtin_ia32_addsubpd (v2df, v2df)
17739 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
17740 v2df __builtin_ia32_haddpd (v2df, v2df)
17741 v4sf __builtin_ia32_haddps (v4sf, v4sf)
17742 v2df __builtin_ia32_hsubpd (v2df, v2df)
17743 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
17744 v16qi __builtin_ia32_lddqu (char const *)
17745 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
17746 v4sf __builtin_ia32_movshdup (v4sf)
17747 v4sf __builtin_ia32_movsldup (v4sf)
17748 void __builtin_ia32_mwait (unsigned int, unsigned int)
17751 The following built-in functions are available when @option{-mssse3} is used.
17752 All of them generate the machine instruction that is part of the name.
17755 v2si __builtin_ia32_phaddd (v2si, v2si)
17756 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
17757 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
17758 v2si __builtin_ia32_phsubd (v2si, v2si)
17759 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
17760 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
17761 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
17762 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
17763 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
17764 v8qi __builtin_ia32_psignb (v8qi, v8qi)
17765 v2si __builtin_ia32_psignd (v2si, v2si)
17766 v4hi __builtin_ia32_psignw (v4hi, v4hi)
17767 v1di __builtin_ia32_palignr (v1di, v1di, int)
17768 v8qi __builtin_ia32_pabsb (v8qi)
17769 v2si __builtin_ia32_pabsd (v2si)
17770 v4hi __builtin_ia32_pabsw (v4hi)
17773 The following built-in functions are available when @option{-mssse3} is used.
17774 All of them generate the machine instruction that is part of the name.
17777 v4si __builtin_ia32_phaddd128 (v4si, v4si)
17778 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
17779 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
17780 v4si __builtin_ia32_phsubd128 (v4si, v4si)
17781 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
17782 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
17783 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
17784 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
17785 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
17786 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
17787 v4si __builtin_ia32_psignd128 (v4si, v4si)
17788 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
17789 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
17790 v16qi __builtin_ia32_pabsb128 (v16qi)
17791 v4si __builtin_ia32_pabsd128 (v4si)
17792 v8hi __builtin_ia32_pabsw128 (v8hi)
17795 The following built-in functions are available when @option{-msse4.1} is
17796 used. All of them generate the machine instruction that is part of the
17800 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
17801 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
17802 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
17803 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
17804 v2df __builtin_ia32_dppd (v2df, v2df, const int)
17805 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
17806 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
17807 v2di __builtin_ia32_movntdqa (v2di *);
17808 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
17809 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
17810 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
17811 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
17812 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
17813 v8hi __builtin_ia32_phminposuw128 (v8hi)
17814 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
17815 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
17816 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
17817 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
17818 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
17819 v4si __builtin_ia32_pminsd128 (v4si, v4si)
17820 v4si __builtin_ia32_pminud128 (v4si, v4si)
17821 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
17822 v4si __builtin_ia32_pmovsxbd128 (v16qi)
17823 v2di __builtin_ia32_pmovsxbq128 (v16qi)
17824 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
17825 v2di __builtin_ia32_pmovsxdq128 (v4si)
17826 v4si __builtin_ia32_pmovsxwd128 (v8hi)
17827 v2di __builtin_ia32_pmovsxwq128 (v8hi)
17828 v4si __builtin_ia32_pmovzxbd128 (v16qi)
17829 v2di __builtin_ia32_pmovzxbq128 (v16qi)
17830 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
17831 v2di __builtin_ia32_pmovzxdq128 (v4si)
17832 v4si __builtin_ia32_pmovzxwd128 (v8hi)
17833 v2di __builtin_ia32_pmovzxwq128 (v8hi)
17834 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
17835 v4si __builtin_ia32_pmulld128 (v4si, v4si)
17836 int __builtin_ia32_ptestc128 (v2di, v2di)
17837 int __builtin_ia32_ptestnzc128 (v2di, v2di)
17838 int __builtin_ia32_ptestz128 (v2di, v2di)
17839 v2df __builtin_ia32_roundpd (v2df, const int)
17840 v4sf __builtin_ia32_roundps (v4sf, const int)
17841 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
17842 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
17845 The following built-in functions are available when @option{-msse4.1} is
17849 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
17850 Generates the @code{insertps} machine instruction.
17851 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
17852 Generates the @code{pextrb} machine instruction.
17853 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
17854 Generates the @code{pinsrb} machine instruction.
17855 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
17856 Generates the @code{pinsrd} machine instruction.
17857 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
17858 Generates the @code{pinsrq} machine instruction in 64bit mode.
17861 The following built-in functions are changed to generate new SSE4.1
17862 instructions when @option{-msse4.1} is used.
17865 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
17866 Generates the @code{extractps} machine instruction.
17867 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
17868 Generates the @code{pextrd} machine instruction.
17869 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
17870 Generates the @code{pextrq} machine instruction in 64bit mode.
17873 The following built-in functions are available when @option{-msse4.2} is
17874 used. All of them generate the machine instruction that is part of the
17878 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
17879 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
17880 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
17881 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
17882 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
17883 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
17884 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
17885 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
17886 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
17887 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
17888 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
17889 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
17890 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
17891 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
17892 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
17895 The following built-in functions are available when @option{-msse4.2} is
17899 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
17900 Generates the @code{crc32b} machine instruction.
17901 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
17902 Generates the @code{crc32w} machine instruction.
17903 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
17904 Generates the @code{crc32l} machine instruction.
17905 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
17906 Generates the @code{crc32q} machine instruction.
17909 The following built-in functions are changed to generate new SSE4.2
17910 instructions when @option{-msse4.2} is used.
17913 @item int __builtin_popcount (unsigned int)
17914 Generates the @code{popcntl} machine instruction.
17915 @item int __builtin_popcountl (unsigned long)
17916 Generates the @code{popcntl} or @code{popcntq} machine instruction,
17917 depending on the size of @code{unsigned long}.
17918 @item int __builtin_popcountll (unsigned long long)
17919 Generates the @code{popcntq} machine instruction.
17922 The following built-in functions are available when @option{-mavx} is
17923 used. All of them generate the machine instruction that is part of the
17927 v4df __builtin_ia32_addpd256 (v4df,v4df)
17928 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
17929 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
17930 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
17931 v4df __builtin_ia32_andnpd256 (v4df,v4df)
17932 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
17933 v4df __builtin_ia32_andpd256 (v4df,v4df)
17934 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
17935 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
17936 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
17937 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
17938 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
17939 v2df __builtin_ia32_cmppd (v2df,v2df,int)
17940 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
17941 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
17942 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
17943 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
17944 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
17945 v4df __builtin_ia32_cvtdq2pd256 (v4si)
17946 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
17947 v4si __builtin_ia32_cvtpd2dq256 (v4df)
17948 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
17949 v8si __builtin_ia32_cvtps2dq256 (v8sf)
17950 v4df __builtin_ia32_cvtps2pd256 (v4sf)
17951 v4si __builtin_ia32_cvttpd2dq256 (v4df)
17952 v8si __builtin_ia32_cvttps2dq256 (v8sf)
17953 v4df __builtin_ia32_divpd256 (v4df,v4df)
17954 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
17955 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
17956 v4df __builtin_ia32_haddpd256 (v4df,v4df)
17957 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
17958 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
17959 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
17960 v32qi __builtin_ia32_lddqu256 (pcchar)
17961 v32qi __builtin_ia32_loaddqu256 (pcchar)
17962 v4df __builtin_ia32_loadupd256 (pcdouble)
17963 v8sf __builtin_ia32_loadups256 (pcfloat)
17964 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
17965 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
17966 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
17967 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
17968 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
17969 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
17970 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
17971 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
17972 v4df __builtin_ia32_maxpd256 (v4df,v4df)
17973 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
17974 v4df __builtin_ia32_minpd256 (v4df,v4df)
17975 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
17976 v4df __builtin_ia32_movddup256 (v4df)
17977 int __builtin_ia32_movmskpd256 (v4df)
17978 int __builtin_ia32_movmskps256 (v8sf)
17979 v8sf __builtin_ia32_movshdup256 (v8sf)
17980 v8sf __builtin_ia32_movsldup256 (v8sf)
17981 v4df __builtin_ia32_mulpd256 (v4df,v4df)
17982 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
17983 v4df __builtin_ia32_orpd256 (v4df,v4df)
17984 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
17985 v2df __builtin_ia32_pd_pd256 (v4df)
17986 v4df __builtin_ia32_pd256_pd (v2df)
17987 v4sf __builtin_ia32_ps_ps256 (v8sf)
17988 v8sf __builtin_ia32_ps256_ps (v4sf)
17989 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
17990 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
17991 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
17992 v8sf __builtin_ia32_rcpps256 (v8sf)
17993 v4df __builtin_ia32_roundpd256 (v4df,int)
17994 v8sf __builtin_ia32_roundps256 (v8sf,int)
17995 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
17996 v8sf __builtin_ia32_rsqrtps256 (v8sf)
17997 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
17998 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
17999 v4si __builtin_ia32_si_si256 (v8si)
18000 v8si __builtin_ia32_si256_si (v4si)
18001 v4df __builtin_ia32_sqrtpd256 (v4df)
18002 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
18003 v8sf __builtin_ia32_sqrtps256 (v8sf)
18004 void __builtin_ia32_storedqu256 (pchar,v32qi)
18005 void __builtin_ia32_storeupd256 (pdouble,v4df)
18006 void __builtin_ia32_storeups256 (pfloat,v8sf)
18007 v4df __builtin_ia32_subpd256 (v4df,v4df)
18008 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
18009 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
18010 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
18011 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
18012 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
18013 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
18014 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
18015 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
18016 v4sf __builtin_ia32_vbroadcastss (pcfloat)
18017 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
18018 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
18019 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
18020 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
18021 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
18022 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
18023 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
18024 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
18025 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
18026 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
18027 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
18028 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
18029 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
18030 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
18031 v2df __builtin_ia32_vpermilpd (v2df,int)
18032 v4df __builtin_ia32_vpermilpd256 (v4df,int)
18033 v4sf __builtin_ia32_vpermilps (v4sf,int)
18034 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
18035 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
18036 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
18037 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
18038 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
18039 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
18040 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
18041 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
18042 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
18043 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
18044 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
18045 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
18046 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
18047 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
18048 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
18049 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
18050 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
18051 void __builtin_ia32_vzeroall (void)
18052 void __builtin_ia32_vzeroupper (void)
18053 v4df __builtin_ia32_xorpd256 (v4df,v4df)
18054 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
18057 The following built-in functions are available when @option{-mavx2} is
18058 used. All of them generate the machine instruction that is part of the
18062 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
18063 v32qi __builtin_ia32_pabsb256 (v32qi)
18064 v16hi __builtin_ia32_pabsw256 (v16hi)
18065 v8si __builtin_ia32_pabsd256 (v8si)
18066 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
18067 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
18068 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
18069 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
18070 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
18071 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
18072 v8si __builtin_ia32_paddd256 (v8si,v8si)
18073 v4di __builtin_ia32_paddq256 (v4di,v4di)
18074 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
18075 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
18076 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
18077 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
18078 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
18079 v4di __builtin_ia32_andsi256 (v4di,v4di)
18080 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
18081 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
18082 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
18083 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
18084 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
18085 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
18086 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
18087 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
18088 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
18089 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
18090 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
18091 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
18092 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
18093 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
18094 v8si __builtin_ia32_phaddd256 (v8si,v8si)
18095 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
18096 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
18097 v8si __builtin_ia32_phsubd256 (v8si,v8si)
18098 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
18099 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
18100 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
18101 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
18102 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
18103 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
18104 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
18105 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
18106 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
18107 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
18108 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
18109 v8si __builtin_ia32_pminsd256 (v8si,v8si)
18110 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
18111 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
18112 v8si __builtin_ia32_pminud256 (v8si,v8si)
18113 int __builtin_ia32_pmovmskb256 (v32qi)
18114 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
18115 v8si __builtin_ia32_pmovsxbd256 (v16qi)
18116 v4di __builtin_ia32_pmovsxbq256 (v16qi)
18117 v8si __builtin_ia32_pmovsxwd256 (v8hi)
18118 v4di __builtin_ia32_pmovsxwq256 (v8hi)
18119 v4di __builtin_ia32_pmovsxdq256 (v4si)
18120 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
18121 v8si __builtin_ia32_pmovzxbd256 (v16qi)
18122 v4di __builtin_ia32_pmovzxbq256 (v16qi)
18123 v8si __builtin_ia32_pmovzxwd256 (v8hi)
18124 v4di __builtin_ia32_pmovzxwq256 (v8hi)
18125 v4di __builtin_ia32_pmovzxdq256 (v4si)
18126 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
18127 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
18128 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
18129 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
18130 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
18131 v8si __builtin_ia32_pmulld256 (v8si,v8si)
18132 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
18133 v4di __builtin_ia32_por256 (v4di,v4di)
18134 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
18135 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
18136 v8si __builtin_ia32_pshufd256 (v8si,int)
18137 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
18138 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
18139 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
18140 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
18141 v8si __builtin_ia32_psignd256 (v8si,v8si)
18142 v4di __builtin_ia32_pslldqi256 (v4di,int)
18143 v16hi __builtin_ia32_psllwi256 (16hi,int)
18144 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
18145 v8si __builtin_ia32_pslldi256 (v8si,int)
18146 v8si __builtin_ia32_pslld256(v8si,v4si)
18147 v4di __builtin_ia32_psllqi256 (v4di,int)
18148 v4di __builtin_ia32_psllq256(v4di,v2di)
18149 v16hi __builtin_ia32_psrawi256 (v16hi,int)
18150 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
18151 v8si __builtin_ia32_psradi256 (v8si,int)
18152 v8si __builtin_ia32_psrad256 (v8si,v4si)
18153 v4di __builtin_ia32_psrldqi256 (v4di, int)
18154 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
18155 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
18156 v8si __builtin_ia32_psrldi256 (v8si,int)
18157 v8si __builtin_ia32_psrld256 (v8si,v4si)
18158 v4di __builtin_ia32_psrlqi256 (v4di,int)
18159 v4di __builtin_ia32_psrlq256(v4di,v2di)
18160 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
18161 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
18162 v8si __builtin_ia32_psubd256 (v8si,v8si)
18163 v4di __builtin_ia32_psubq256 (v4di,v4di)
18164 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
18165 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
18166 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
18167 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
18168 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
18169 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
18170 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
18171 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
18172 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
18173 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
18174 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
18175 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
18176 v4di __builtin_ia32_pxor256 (v4di,v4di)
18177 v4di __builtin_ia32_movntdqa256 (pv4di)
18178 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
18179 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
18180 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
18181 v4di __builtin_ia32_vbroadcastsi256 (v2di)
18182 v4si __builtin_ia32_pblendd128 (v4si,v4si)
18183 v8si __builtin_ia32_pblendd256 (v8si,v8si)
18184 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
18185 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
18186 v8si __builtin_ia32_pbroadcastd256 (v4si)
18187 v4di __builtin_ia32_pbroadcastq256 (v2di)
18188 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
18189 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
18190 v4si __builtin_ia32_pbroadcastd128 (v4si)
18191 v2di __builtin_ia32_pbroadcastq128 (v2di)
18192 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
18193 v4df __builtin_ia32_permdf256 (v4df,int)
18194 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
18195 v4di __builtin_ia32_permdi256 (v4di,int)
18196 v4di __builtin_ia32_permti256 (v4di,v4di,int)
18197 v4di __builtin_ia32_extract128i256 (v4di,int)
18198 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
18199 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
18200 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
18201 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
18202 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
18203 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
18204 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
18205 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
18206 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
18207 v8si __builtin_ia32_psllv8si (v8si,v8si)
18208 v4si __builtin_ia32_psllv4si (v4si,v4si)
18209 v4di __builtin_ia32_psllv4di (v4di,v4di)
18210 v2di __builtin_ia32_psllv2di (v2di,v2di)
18211 v8si __builtin_ia32_psrav8si (v8si,v8si)
18212 v4si __builtin_ia32_psrav4si (v4si,v4si)
18213 v8si __builtin_ia32_psrlv8si (v8si,v8si)
18214 v4si __builtin_ia32_psrlv4si (v4si,v4si)
18215 v4di __builtin_ia32_psrlv4di (v4di,v4di)
18216 v2di __builtin_ia32_psrlv2di (v2di,v2di)
18217 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
18218 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
18219 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
18220 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
18221 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
18222 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
18223 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
18224 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
18225 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
18226 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
18227 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
18228 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
18229 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
18230 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
18231 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
18232 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
18235 The following built-in functions are available when @option{-maes} is
18236 used. All of them generate the machine instruction that is part of the
18240 v2di __builtin_ia32_aesenc128 (v2di, v2di)
18241 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
18242 v2di __builtin_ia32_aesdec128 (v2di, v2di)
18243 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
18244 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
18245 v2di __builtin_ia32_aesimc128 (v2di)
18248 The following built-in function is available when @option{-mpclmul} is
18252 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
18253 Generates the @code{pclmulqdq} machine instruction.
18256 The following built-in function is available when @option{-mfsgsbase} is
18257 used. All of them generate the machine instruction that is part of the
18261 unsigned int __builtin_ia32_rdfsbase32 (void)
18262 unsigned long long __builtin_ia32_rdfsbase64 (void)
18263 unsigned int __builtin_ia32_rdgsbase32 (void)
18264 unsigned long long __builtin_ia32_rdgsbase64 (void)
18265 void _writefsbase_u32 (unsigned int)
18266 void _writefsbase_u64 (unsigned long long)
18267 void _writegsbase_u32 (unsigned int)
18268 void _writegsbase_u64 (unsigned long long)
18271 The following built-in function is available when @option{-mrdrnd} is
18272 used. All of them generate the machine instruction that is part of the
18276 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
18277 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
18278 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
18281 The following built-in functions are available when @option{-msse4a} is used.
18282 All of them generate the machine instruction that is part of the name.
18285 void __builtin_ia32_movntsd (double *, v2df)
18286 void __builtin_ia32_movntss (float *, v4sf)
18287 v2di __builtin_ia32_extrq (v2di, v16qi)
18288 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
18289 v2di __builtin_ia32_insertq (v2di, v2di)
18290 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
18293 The following built-in functions are available when @option{-mxop} is used.
18295 v2df __builtin_ia32_vfrczpd (v2df)
18296 v4sf __builtin_ia32_vfrczps (v4sf)
18297 v2df __builtin_ia32_vfrczsd (v2df)
18298 v4sf __builtin_ia32_vfrczss (v4sf)
18299 v4df __builtin_ia32_vfrczpd256 (v4df)
18300 v8sf __builtin_ia32_vfrczps256 (v8sf)
18301 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
18302 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
18303 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
18304 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
18305 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
18306 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
18307 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
18308 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
18309 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
18310 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
18311 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
18312 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
18313 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
18314 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
18315 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18316 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
18317 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
18318 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
18319 v4si __builtin_ia32_vpcomequd (v4si, v4si)
18320 v2di __builtin_ia32_vpcomequq (v2di, v2di)
18321 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
18322 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
18323 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
18324 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
18325 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
18326 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
18327 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
18328 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
18329 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
18330 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
18331 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
18332 v4si __builtin_ia32_vpcomged (v4si, v4si)
18333 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
18334 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
18335 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
18336 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
18337 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
18338 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
18339 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
18340 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
18341 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
18342 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
18343 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
18344 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
18345 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
18346 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
18347 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
18348 v4si __builtin_ia32_vpcomled (v4si, v4si)
18349 v2di __builtin_ia32_vpcomleq (v2di, v2di)
18350 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
18351 v4si __builtin_ia32_vpcomleud (v4si, v4si)
18352 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
18353 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
18354 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
18355 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
18356 v4si __builtin_ia32_vpcomltd (v4si, v4si)
18357 v2di __builtin_ia32_vpcomltq (v2di, v2di)
18358 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
18359 v4si __builtin_ia32_vpcomltud (v4si, v4si)
18360 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
18361 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
18362 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
18363 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
18364 v4si __builtin_ia32_vpcomned (v4si, v4si)
18365 v2di __builtin_ia32_vpcomneq (v2di, v2di)
18366 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
18367 v4si __builtin_ia32_vpcomneud (v4si, v4si)
18368 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
18369 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
18370 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
18371 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
18372 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
18373 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
18374 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
18375 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
18376 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
18377 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
18378 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
18379 v4si __builtin_ia32_vphaddbd (v16qi)
18380 v2di __builtin_ia32_vphaddbq (v16qi)
18381 v8hi __builtin_ia32_vphaddbw (v16qi)
18382 v2di __builtin_ia32_vphadddq (v4si)
18383 v4si __builtin_ia32_vphaddubd (v16qi)
18384 v2di __builtin_ia32_vphaddubq (v16qi)
18385 v8hi __builtin_ia32_vphaddubw (v16qi)
18386 v2di __builtin_ia32_vphaddudq (v4si)
18387 v4si __builtin_ia32_vphadduwd (v8hi)
18388 v2di __builtin_ia32_vphadduwq (v8hi)
18389 v4si __builtin_ia32_vphaddwd (v8hi)
18390 v2di __builtin_ia32_vphaddwq (v8hi)
18391 v8hi __builtin_ia32_vphsubbw (v16qi)
18392 v2di __builtin_ia32_vphsubdq (v4si)
18393 v4si __builtin_ia32_vphsubwd (v8hi)
18394 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
18395 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
18396 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
18397 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
18398 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
18399 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
18400 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
18401 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
18402 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
18403 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
18404 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
18405 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
18406 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
18407 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
18408 v4si __builtin_ia32_vprotd (v4si, v4si)
18409 v2di __builtin_ia32_vprotq (v2di, v2di)
18410 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
18411 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
18412 v4si __builtin_ia32_vpshad (v4si, v4si)
18413 v2di __builtin_ia32_vpshaq (v2di, v2di)
18414 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
18415 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
18416 v4si __builtin_ia32_vpshld (v4si, v4si)
18417 v2di __builtin_ia32_vpshlq (v2di, v2di)
18418 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
18421 The following built-in functions are available when @option{-mfma4} is used.
18422 All of them generate the machine instruction that is part of the name.
18425 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
18426 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
18427 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
18428 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
18429 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
18430 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
18431 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
18432 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
18433 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
18434 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
18435 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
18436 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
18437 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
18438 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
18439 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
18440 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
18441 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
18442 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
18443 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
18444 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
18445 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
18446 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
18447 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
18448 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
18449 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
18450 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
18451 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
18452 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
18453 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
18454 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
18455 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
18456 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
18460 The following built-in functions are available when @option{-mlwp} is used.
18463 void __builtin_ia32_llwpcb16 (void *);
18464 void __builtin_ia32_llwpcb32 (void *);
18465 void __builtin_ia32_llwpcb64 (void *);
18466 void * __builtin_ia32_llwpcb16 (void);
18467 void * __builtin_ia32_llwpcb32 (void);
18468 void * __builtin_ia32_llwpcb64 (void);
18469 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
18470 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
18471 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
18472 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
18473 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
18474 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
18477 The following built-in functions are available when @option{-mbmi} is used.
18478 All of them generate the machine instruction that is part of the name.
18480 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
18481 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
18484 The following built-in functions are available when @option{-mbmi2} is used.
18485 All of them generate the machine instruction that is part of the name.
18487 unsigned int _bzhi_u32 (unsigned int, unsigned int)
18488 unsigned int _pdep_u32 (unsigned int, unsigned int)
18489 unsigned int _pext_u32 (unsigned int, unsigned int)
18490 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
18491 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
18492 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
18495 The following built-in functions are available when @option{-mlzcnt} is used.
18496 All of them generate the machine instruction that is part of the name.
18498 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
18499 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
18500 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
18503 The following built-in functions are available when @option{-mfxsr} is used.
18504 All of them generate the machine instruction that is part of the name.
18506 void __builtin_ia32_fxsave (void *)
18507 void __builtin_ia32_fxrstor (void *)
18508 void __builtin_ia32_fxsave64 (void *)
18509 void __builtin_ia32_fxrstor64 (void *)
18512 The following built-in functions are available when @option{-mxsave} is used.
18513 All of them generate the machine instruction that is part of the name.
18515 void __builtin_ia32_xsave (void *, long long)
18516 void __builtin_ia32_xrstor (void *, long long)
18517 void __builtin_ia32_xsave64 (void *, long long)
18518 void __builtin_ia32_xrstor64 (void *, long long)
18521 The following built-in functions are available when @option{-mxsaveopt} is used.
18522 All of them generate the machine instruction that is part of the name.
18524 void __builtin_ia32_xsaveopt (void *, long long)
18525 void __builtin_ia32_xsaveopt64 (void *, long long)
18528 The following built-in functions are available when @option{-mtbm} is used.
18529 Both of them generate the immediate form of the bextr machine instruction.
18531 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
18532 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
18536 The following built-in functions are available when @option{-m3dnow} is used.
18537 All of them generate the machine instruction that is part of the name.
18540 void __builtin_ia32_femms (void)
18541 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
18542 v2si __builtin_ia32_pf2id (v2sf)
18543 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
18544 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
18545 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
18546 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
18547 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
18548 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
18549 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
18550 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
18551 v2sf __builtin_ia32_pfrcp (v2sf)
18552 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
18553 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
18554 v2sf __builtin_ia32_pfrsqrt (v2sf)
18555 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
18556 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
18557 v2sf __builtin_ia32_pi2fd (v2si)
18558 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
18561 The following built-in functions are available when both @option{-m3dnow}
18562 and @option{-march=athlon} are used. All of them generate the machine
18563 instruction that is part of the name.
18566 v2si __builtin_ia32_pf2iw (v2sf)
18567 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
18568 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
18569 v2sf __builtin_ia32_pi2fw (v2si)
18570 v2sf __builtin_ia32_pswapdsf (v2sf)
18571 v2si __builtin_ia32_pswapdsi (v2si)
18574 The following built-in functions are available when @option{-mrtm} is used
18575 They are used for restricted transactional memory. These are the internal
18576 low level functions. Normally the functions in
18577 @ref{x86 transactional memory intrinsics} should be used instead.
18580 int __builtin_ia32_xbegin ()
18581 void __builtin_ia32_xend ()
18582 void __builtin_ia32_xabort (status)
18583 int __builtin_ia32_xtest ()
18586 The following built-in functions are available when @option{-mmwaitx} is used.
18587 All of them generate the machine instruction that is part of the name.
18589 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
18590 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
18593 The following built-in functions are available when @option{-mclzero} is used.
18594 All of them generate the machine instruction that is part of the name.
18596 void __builtin_i32_clzero (void *)
18599 The following built-in functions are available when @option{-mpku} is used.
18600 They generate reads and writes to PKRU.
18602 void __builtin_ia32_wrpkru (unsigned int)
18603 unsigned int __builtin_ia32_rdpkru ()
18606 @node x86 transactional memory intrinsics
18607 @subsection x86 Transactional Memory Intrinsics
18609 These hardware transactional memory intrinsics for x86 allow you to use
18610 memory transactions with RTM (Restricted Transactional Memory).
18611 This support is enabled with the @option{-mrtm} option.
18612 For using HLE (Hardware Lock Elision) see
18613 @ref{x86 specific memory model extensions for transactional memory} instead.
18615 A memory transaction commits all changes to memory in an atomic way,
18616 as visible to other threads. If the transaction fails it is rolled back
18617 and all side effects discarded.
18619 Generally there is no guarantee that a memory transaction ever succeeds
18620 and suitable fallback code always needs to be supplied.
18622 @deftypefn {RTM Function} {unsigned} _xbegin ()
18623 Start a RTM (Restricted Transactional Memory) transaction.
18624 Returns @code{_XBEGIN_STARTED} when the transaction
18625 started successfully (note this is not 0, so the constant has to be
18626 explicitly tested).
18628 If the transaction aborts, all side-effects
18629 are undone and an abort code encoded as a bit mask is returned.
18630 The following macros are defined:
18633 @item _XABORT_EXPLICIT
18634 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
18635 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
18636 @item _XABORT_RETRY
18637 Transaction retry is possible.
18638 @item _XABORT_CONFLICT
18639 Transaction abort due to a memory conflict with another thread.
18640 @item _XABORT_CAPACITY
18641 Transaction abort due to the transaction using too much memory.
18642 @item _XABORT_DEBUG
18643 Transaction abort due to a debug trap.
18644 @item _XABORT_NESTED
18645 Transaction abort in an inner nested transaction.
18648 There is no guarantee
18649 any transaction ever succeeds, so there always needs to be a valid
18653 @deftypefn {RTM Function} {void} _xend ()
18654 Commit the current transaction. When no transaction is active this faults.
18655 All memory side-effects of the transaction become visible
18656 to other threads in an atomic manner.
18659 @deftypefn {RTM Function} {int} _xtest ()
18660 Return a nonzero value if a transaction is currently active, otherwise 0.
18663 @deftypefn {RTM Function} {void} _xabort (status)
18664 Abort the current transaction. When no transaction is active this is a no-op.
18665 The @var{status} is an 8-bit constant; its value is encoded in the return
18666 value from @code{_xbegin}.
18669 Here is an example showing handling for @code{_XABORT_RETRY}
18670 and a fallback path for other failures:
18673 #include <immintrin.h>
18675 int n_tries, max_tries;
18676 unsigned status = _XABORT_EXPLICIT;
18679 for (n_tries = 0; n_tries < max_tries; n_tries++)
18681 status = _xbegin ();
18682 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
18685 if (status == _XBEGIN_STARTED)
18687 ... transaction code...
18692 ... non-transactional fallback path...
18697 Note that, in most cases, the transactional and non-transactional code
18698 must synchronize together to ensure consistency.
18700 @node Target Format Checks
18701 @section Format Checks Specific to Particular Target Machines
18703 For some target machines, GCC supports additional options to the
18705 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
18708 * Solaris Format Checks::
18709 * Darwin Format Checks::
18712 @node Solaris Format Checks
18713 @subsection Solaris Format Checks
18715 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
18716 check. @code{cmn_err} accepts a subset of the standard @code{printf}
18717 conversions, and the two-argument @code{%b} conversion for displaying
18718 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
18720 @node Darwin Format Checks
18721 @subsection Darwin Format Checks
18723 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
18724 attribute context. Declarations made with such attribution are parsed for correct syntax
18725 and format argument types. However, parsing of the format string itself is currently undefined
18726 and is not carried out by this version of the compiler.
18728 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
18729 also be used as format arguments. Note that the relevant headers are only likely to be
18730 available on Darwin (OSX) installations. On such installations, the XCode and system
18731 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
18732 associated functions.
18735 @section Pragmas Accepted by GCC
18737 @cindex @code{#pragma}
18739 GCC supports several types of pragmas, primarily in order to compile
18740 code originally written for other compilers. Note that in general
18741 we do not recommend the use of pragmas; @xref{Function Attributes},
18742 for further explanation.
18745 * AArch64 Pragmas::
18749 * RS/6000 and PowerPC Pragmas::
18752 * Solaris Pragmas::
18753 * Symbol-Renaming Pragmas::
18754 * Structure-Layout Pragmas::
18756 * Diagnostic Pragmas::
18757 * Visibility Pragmas::
18758 * Push/Pop Macro Pragmas::
18759 * Function Specific Option Pragmas::
18760 * Loop-Specific Pragmas::
18763 @node AArch64 Pragmas
18764 @subsection AArch64 Pragmas
18766 The pragmas defined by the AArch64 target correspond to the AArch64
18767 target function attributes. They can be specified as below:
18769 #pragma GCC target("string")
18772 where @code{@var{string}} can be any string accepted as an AArch64 target
18773 attribute. @xref{AArch64 Function Attributes}, for more details
18774 on the permissible values of @code{string}.
18777 @subsection ARM Pragmas
18779 The ARM target defines pragmas for controlling the default addition of
18780 @code{long_call} and @code{short_call} attributes to functions.
18781 @xref{Function Attributes}, for information about the effects of these
18786 @cindex pragma, long_calls
18787 Set all subsequent functions to have the @code{long_call} attribute.
18789 @item no_long_calls
18790 @cindex pragma, no_long_calls
18791 Set all subsequent functions to have the @code{short_call} attribute.
18793 @item long_calls_off
18794 @cindex pragma, long_calls_off
18795 Do not affect the @code{long_call} or @code{short_call} attributes of
18796 subsequent functions.
18800 @subsection M32C Pragmas
18803 @item GCC memregs @var{number}
18804 @cindex pragma, memregs
18805 Overrides the command-line option @code{-memregs=} for the current
18806 file. Use with care! This pragma must be before any function in the
18807 file, and mixing different memregs values in different objects may
18808 make them incompatible. This pragma is useful when a
18809 performance-critical function uses a memreg for temporary values,
18810 as it may allow you to reduce the number of memregs used.
18812 @item ADDRESS @var{name} @var{address}
18813 @cindex pragma, address
18814 For any declared symbols matching @var{name}, this does three things
18815 to that symbol: it forces the symbol to be located at the given
18816 address (a number), it forces the symbol to be volatile, and it
18817 changes the symbol's scope to be static. This pragma exists for
18818 compatibility with other compilers, but note that the common
18819 @code{1234H} numeric syntax is not supported (use @code{0x1234}
18823 #pragma ADDRESS port3 0x103
18830 @subsection MeP Pragmas
18834 @item custom io_volatile (on|off)
18835 @cindex pragma, custom io_volatile
18836 Overrides the command-line option @code{-mio-volatile} for the current
18837 file. Note that for compatibility with future GCC releases, this
18838 option should only be used once before any @code{io} variables in each
18841 @item GCC coprocessor available @var{registers}
18842 @cindex pragma, coprocessor available
18843 Specifies which coprocessor registers are available to the register
18844 allocator. @var{registers} may be a single register, register range
18845 separated by ellipses, or comma-separated list of those. Example:
18848 #pragma GCC coprocessor available $c0...$c10, $c28
18851 @item GCC coprocessor call_saved @var{registers}
18852 @cindex pragma, coprocessor call_saved
18853 Specifies which coprocessor registers are to be saved and restored by
18854 any function using them. @var{registers} may be a single register,
18855 register range separated by ellipses, or comma-separated list of
18859 #pragma GCC coprocessor call_saved $c4...$c6, $c31
18862 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
18863 @cindex pragma, coprocessor subclass
18864 Creates and defines a register class. These register classes can be
18865 used by inline @code{asm} constructs. @var{registers} may be a single
18866 register, register range separated by ellipses, or comma-separated
18867 list of those. Example:
18870 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
18872 asm ("cpfoo %0" : "=B" (x));
18875 @item GCC disinterrupt @var{name} , @var{name} @dots{}
18876 @cindex pragma, disinterrupt
18877 For the named functions, the compiler adds code to disable interrupts
18878 for the duration of those functions. If any functions so named
18879 are not encountered in the source, a warning is emitted that the pragma is
18880 not used. Examples:
18883 #pragma disinterrupt foo
18884 #pragma disinterrupt bar, grill
18885 int foo () @{ @dots{} @}
18888 @item GCC call @var{name} , @var{name} @dots{}
18889 @cindex pragma, call
18890 For the named functions, the compiler always uses a register-indirect
18891 call model when calling the named functions. Examples:
18900 @node RS/6000 and PowerPC Pragmas
18901 @subsection RS/6000 and PowerPC Pragmas
18903 The RS/6000 and PowerPC targets define one pragma for controlling
18904 whether or not the @code{longcall} attribute is added to function
18905 declarations by default. This pragma overrides the @option{-mlongcall}
18906 option, but not the @code{longcall} and @code{shortcall} attributes.
18907 @xref{RS/6000 and PowerPC Options}, for more information about when long
18908 calls are and are not necessary.
18912 @cindex pragma, longcall
18913 Apply the @code{longcall} attribute to all subsequent function
18917 Do not apply the @code{longcall} attribute to subsequent function
18921 @c Describe h8300 pragmas here.
18922 @c Describe sh pragmas here.
18923 @c Describe v850 pragmas here.
18925 @node S/390 Pragmas
18926 @subsection S/390 Pragmas
18928 The pragmas defined by the S/390 target correspond to the S/390
18929 target function attributes and some the additional options:
18936 Note that options of the pragma, unlike options of the target
18937 attribute, do change the value of preprocessor macros like
18938 @code{__VEC__}. They can be specified as below:
18941 #pragma GCC target("string[,string]...")
18942 #pragma GCC target("string"[,"string"]...)
18945 @node Darwin Pragmas
18946 @subsection Darwin Pragmas
18948 The following pragmas are available for all architectures running the
18949 Darwin operating system. These are useful for compatibility with other
18953 @item mark @var{tokens}@dots{}
18954 @cindex pragma, mark
18955 This pragma is accepted, but has no effect.
18957 @item options align=@var{alignment}
18958 @cindex pragma, options align
18959 This pragma sets the alignment of fields in structures. The values of
18960 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
18961 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
18962 properly; to restore the previous setting, use @code{reset} for the
18965 @item segment @var{tokens}@dots{}
18966 @cindex pragma, segment
18967 This pragma is accepted, but has no effect.
18969 @item unused (@var{var} [, @var{var}]@dots{})
18970 @cindex pragma, unused
18971 This pragma declares variables to be possibly unused. GCC does not
18972 produce warnings for the listed variables. The effect is similar to
18973 that of the @code{unused} attribute, except that this pragma may appear
18974 anywhere within the variables' scopes.
18977 @node Solaris Pragmas
18978 @subsection Solaris Pragmas
18980 The Solaris target supports @code{#pragma redefine_extname}
18981 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
18982 @code{#pragma} directives for compatibility with the system compiler.
18985 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
18986 @cindex pragma, align
18988 Increase the minimum alignment of each @var{variable} to @var{alignment}.
18989 This is the same as GCC's @code{aligned} attribute @pxref{Variable
18990 Attributes}). Macro expansion occurs on the arguments to this pragma
18991 when compiling C and Objective-C@. It does not currently occur when
18992 compiling C++, but this is a bug which may be fixed in a future
18995 @item fini (@var{function} [, @var{function}]...)
18996 @cindex pragma, fini
18998 This pragma causes each listed @var{function} to be called after
18999 main, or during shared module unloading, by adding a call to the
19000 @code{.fini} section.
19002 @item init (@var{function} [, @var{function}]...)
19003 @cindex pragma, init
19005 This pragma causes each listed @var{function} to be called during
19006 initialization (before @code{main}) or during shared module loading, by
19007 adding a call to the @code{.init} section.
19011 @node Symbol-Renaming Pragmas
19012 @subsection Symbol-Renaming Pragmas
19014 GCC supports a @code{#pragma} directive that changes the name used in
19015 assembly for a given declaration. While this pragma is supported on all
19016 platforms, it is intended primarily to provide compatibility with the
19017 Solaris system headers. This effect can also be achieved using the asm
19018 labels extension (@pxref{Asm Labels}).
19021 @item redefine_extname @var{oldname} @var{newname}
19022 @cindex pragma, redefine_extname
19024 This pragma gives the C function @var{oldname} the assembly symbol
19025 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
19026 is defined if this pragma is available (currently on all platforms).
19029 This pragma and the asm labels extension interact in a complicated
19030 manner. Here are some corner cases you may want to be aware of:
19033 @item This pragma silently applies only to declarations with external
19034 linkage. Asm labels do not have this restriction.
19036 @item In C++, this pragma silently applies only to declarations with
19037 ``C'' linkage. Again, asm labels do not have this restriction.
19039 @item If either of the ways of changing the assembly name of a
19040 declaration are applied to a declaration whose assembly name has
19041 already been determined (either by a previous use of one of these
19042 features, or because the compiler needed the assembly name in order to
19043 generate code), and the new name is different, a warning issues and
19044 the name does not change.
19046 @item The @var{oldname} used by @code{#pragma redefine_extname} is
19047 always the C-language name.
19050 @node Structure-Layout Pragmas
19051 @subsection Structure-Layout Pragmas
19053 For compatibility with Microsoft Windows compilers, GCC supports a
19054 set of @code{#pragma} directives that change the maximum alignment of
19055 members of structures (other than zero-width bit-fields), unions, and
19056 classes subsequently defined. The @var{n} value below always is required
19057 to be a small power of two and specifies the new alignment in bytes.
19060 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
19061 @item @code{#pragma pack()} sets the alignment to the one that was in
19062 effect when compilation started (see also command-line option
19063 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
19064 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
19065 setting on an internal stack and then optionally sets the new alignment.
19066 @item @code{#pragma pack(pop)} restores the alignment setting to the one
19067 saved at the top of the internal stack (and removes that stack entry).
19068 Note that @code{#pragma pack([@var{n}])} does not influence this internal
19069 stack; thus it is possible to have @code{#pragma pack(push)} followed by
19070 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
19071 @code{#pragma pack(pop)}.
19074 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
19075 directive which lays out structures and unions subsequently defined as the
19076 documented @code{__attribute__ ((ms_struct))}.
19079 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
19080 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
19081 @item @code{#pragma ms_struct reset} goes back to the default layout.
19084 Most targets also support the @code{#pragma scalar_storage_order} directive
19085 which lays out structures and unions subsequently defined as the documented
19086 @code{__attribute__ ((scalar_storage_order))}.
19089 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
19090 of the scalar fields to big-endian.
19091 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
19092 of the scalar fields to little-endian.
19093 @item @code{#pragma scalar_storage_order default} goes back to the endianness
19094 that was in effect when compilation started (see also command-line option
19095 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
19099 @subsection Weak Pragmas
19101 For compatibility with SVR4, GCC supports a set of @code{#pragma}
19102 directives for declaring symbols to be weak, and defining weak
19106 @item #pragma weak @var{symbol}
19107 @cindex pragma, weak
19108 This pragma declares @var{symbol} to be weak, as if the declaration
19109 had the attribute of the same name. The pragma may appear before
19110 or after the declaration of @var{symbol}. It is not an error for
19111 @var{symbol} to never be defined at all.
19113 @item #pragma weak @var{symbol1} = @var{symbol2}
19114 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
19115 It is an error if @var{symbol2} is not defined in the current
19119 @node Diagnostic Pragmas
19120 @subsection Diagnostic Pragmas
19122 GCC allows the user to selectively enable or disable certain types of
19123 diagnostics, and change the kind of the diagnostic. For example, a
19124 project's policy might require that all sources compile with
19125 @option{-Werror} but certain files might have exceptions allowing
19126 specific types of warnings. Or, a project might selectively enable
19127 diagnostics and treat them as errors depending on which preprocessor
19128 macros are defined.
19131 @item #pragma GCC diagnostic @var{kind} @var{option}
19132 @cindex pragma, diagnostic
19134 Modifies the disposition of a diagnostic. Note that not all
19135 diagnostics are modifiable; at the moment only warnings (normally
19136 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
19137 Use @option{-fdiagnostics-show-option} to determine which diagnostics
19138 are controllable and which option controls them.
19140 @var{kind} is @samp{error} to treat this diagnostic as an error,
19141 @samp{warning} to treat it like a warning (even if @option{-Werror} is
19142 in effect), or @samp{ignored} if the diagnostic is to be ignored.
19143 @var{option} is a double quoted string that matches the command-line
19147 #pragma GCC diagnostic warning "-Wformat"
19148 #pragma GCC diagnostic error "-Wformat"
19149 #pragma GCC diagnostic ignored "-Wformat"
19152 Note that these pragmas override any command-line options. GCC keeps
19153 track of the location of each pragma, and issues diagnostics according
19154 to the state as of that point in the source file. Thus, pragmas occurring
19155 after a line do not affect diagnostics caused by that line.
19157 @item #pragma GCC diagnostic push
19158 @itemx #pragma GCC diagnostic pop
19160 Causes GCC to remember the state of the diagnostics as of each
19161 @code{push}, and restore to that point at each @code{pop}. If a
19162 @code{pop} has no matching @code{push}, the command-line options are
19166 #pragma GCC diagnostic error "-Wuninitialized"
19167 foo(a); /* error is given for this one */
19168 #pragma GCC diagnostic push
19169 #pragma GCC diagnostic ignored "-Wuninitialized"
19170 foo(b); /* no diagnostic for this one */
19171 #pragma GCC diagnostic pop
19172 foo(c); /* error is given for this one */
19173 #pragma GCC diagnostic pop
19174 foo(d); /* depends on command-line options */
19179 GCC also offers a simple mechanism for printing messages during
19183 @item #pragma message @var{string}
19184 @cindex pragma, diagnostic
19186 Prints @var{string} as a compiler message on compilation. The message
19187 is informational only, and is neither a compilation warning nor an error.
19190 #pragma message "Compiling " __FILE__ "..."
19193 @var{string} may be parenthesized, and is printed with location
19194 information. For example,
19197 #define DO_PRAGMA(x) _Pragma (#x)
19198 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
19200 TODO(Remember to fix this)
19204 prints @samp{/tmp/file.c:4: note: #pragma message:
19205 TODO - Remember to fix this}.
19209 @node Visibility Pragmas
19210 @subsection Visibility Pragmas
19213 @item #pragma GCC visibility push(@var{visibility})
19214 @itemx #pragma GCC visibility pop
19215 @cindex pragma, visibility
19217 This pragma allows the user to set the visibility for multiple
19218 declarations without having to give each a visibility attribute
19219 (@pxref{Function Attributes}).
19221 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
19222 declarations. Class members and template specializations are not
19223 affected; if you want to override the visibility for a particular
19224 member or instantiation, you must use an attribute.
19229 @node Push/Pop Macro Pragmas
19230 @subsection Push/Pop Macro Pragmas
19232 For compatibility with Microsoft Windows compilers, GCC supports
19233 @samp{#pragma push_macro(@var{"macro_name"})}
19234 and @samp{#pragma pop_macro(@var{"macro_name"})}.
19237 @item #pragma push_macro(@var{"macro_name"})
19238 @cindex pragma, push_macro
19239 This pragma saves the value of the macro named as @var{macro_name} to
19240 the top of the stack for this macro.
19242 @item #pragma pop_macro(@var{"macro_name"})
19243 @cindex pragma, pop_macro
19244 This pragma sets the value of the macro named as @var{macro_name} to
19245 the value on top of the stack for this macro. If the stack for
19246 @var{macro_name} is empty, the value of the macro remains unchanged.
19253 #pragma push_macro("X")
19256 #pragma pop_macro("X")
19261 In this example, the definition of X as 1 is saved by @code{#pragma
19262 push_macro} and restored by @code{#pragma pop_macro}.
19264 @node Function Specific Option Pragmas
19265 @subsection Function Specific Option Pragmas
19268 @item #pragma GCC target (@var{"string"}...)
19269 @cindex pragma GCC target
19271 This pragma allows you to set target specific options for functions
19272 defined later in the source file. One or more strings can be
19273 specified. Each function that is defined after this point is as
19274 if @code{attribute((target("STRING")))} was specified for that
19275 function. The parenthesis around the options is optional.
19276 @xref{Function Attributes}, for more information about the
19277 @code{target} attribute and the attribute syntax.
19279 The @code{#pragma GCC target} pragma is presently implemented for
19280 x86, PowerPC, and Nios II targets only.
19284 @item #pragma GCC optimize (@var{"string"}...)
19285 @cindex pragma GCC optimize
19287 This pragma allows you to set global optimization options for functions
19288 defined later in the source file. One or more strings can be
19289 specified. Each function that is defined after this point is as
19290 if @code{attribute((optimize("STRING")))} was specified for that
19291 function. The parenthesis around the options is optional.
19292 @xref{Function Attributes}, for more information about the
19293 @code{optimize} attribute and the attribute syntax.
19297 @item #pragma GCC push_options
19298 @itemx #pragma GCC pop_options
19299 @cindex pragma GCC push_options
19300 @cindex pragma GCC pop_options
19302 These pragmas maintain a stack of the current target and optimization
19303 options. It is intended for include files where you temporarily want
19304 to switch to using a different @samp{#pragma GCC target} or
19305 @samp{#pragma GCC optimize} and then to pop back to the previous
19310 @item #pragma GCC reset_options
19311 @cindex pragma GCC reset_options
19313 This pragma clears the current @code{#pragma GCC target} and
19314 @code{#pragma GCC optimize} to use the default switches as specified
19315 on the command line.
19318 @node Loop-Specific Pragmas
19319 @subsection Loop-Specific Pragmas
19322 @item #pragma GCC ivdep
19323 @cindex pragma GCC ivdep
19326 With this pragma, the programmer asserts that there are no loop-carried
19327 dependencies which would prevent consecutive iterations of
19328 the following loop from executing concurrently with SIMD
19329 (single instruction multiple data) instructions.
19331 For example, the compiler can only unconditionally vectorize the following
19332 loop with the pragma:
19335 void foo (int n, int *a, int *b, int *c)
19339 for (i = 0; i < n; ++i)
19340 a[i] = b[i] + c[i];
19345 In this example, using the @code{restrict} qualifier had the same
19346 effect. In the following example, that would not be possible. Assume
19347 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
19348 that it can unconditionally vectorize the following loop:
19351 void ignore_vec_dep (int *a, int k, int c, int m)
19354 for (int i = 0; i < m; i++)
19355 a[i] = a[i + k] * c;
19360 @node Unnamed Fields
19361 @section Unnamed Structure and Union Fields
19362 @cindex @code{struct}
19363 @cindex @code{union}
19365 As permitted by ISO C11 and for compatibility with other compilers,
19366 GCC allows you to define
19367 a structure or union that contains, as fields, structures and unions
19368 without names. For example:
19382 In this example, you are able to access members of the unnamed
19383 union with code like @samp{foo.b}. Note that only unnamed structs and
19384 unions are allowed, you may not have, for example, an unnamed
19387 You must never create such structures that cause ambiguous field definitions.
19388 For example, in this structure:
19400 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
19401 The compiler gives errors for such constructs.
19403 @opindex fms-extensions
19404 Unless @option{-fms-extensions} is used, the unnamed field must be a
19405 structure or union definition without a tag (for example, @samp{struct
19406 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
19407 also be a definition with a tag such as @samp{struct foo @{ int a;
19408 @};}, a reference to a previously defined structure or union such as
19409 @samp{struct foo;}, or a reference to a @code{typedef} name for a
19410 previously defined structure or union type.
19412 @opindex fplan9-extensions
19413 The option @option{-fplan9-extensions} enables
19414 @option{-fms-extensions} as well as two other extensions. First, a
19415 pointer to a structure is automatically converted to a pointer to an
19416 anonymous field for assignments and function calls. For example:
19419 struct s1 @{ int a; @};
19420 struct s2 @{ struct s1; @};
19421 extern void f1 (struct s1 *);
19422 void f2 (struct s2 *p) @{ f1 (p); @}
19426 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
19427 converted into a pointer to the anonymous field.
19429 Second, when the type of an anonymous field is a @code{typedef} for a
19430 @code{struct} or @code{union}, code may refer to the field using the
19431 name of the @code{typedef}.
19434 typedef struct @{ int a; @} s1;
19435 struct s2 @{ s1; @};
19436 s1 f1 (struct s2 *p) @{ return p->s1; @}
19439 These usages are only permitted when they are not ambiguous.
19442 @section Thread-Local Storage
19443 @cindex Thread-Local Storage
19444 @cindex @acronym{TLS}
19445 @cindex @code{__thread}
19447 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
19448 are allocated such that there is one instance of the variable per extant
19449 thread. The runtime model GCC uses to implement this originates
19450 in the IA-64 processor-specific ABI, but has since been migrated
19451 to other processors as well. It requires significant support from
19452 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
19453 system libraries (@file{libc.so} and @file{libpthread.so}), so it
19454 is not available everywhere.
19456 At the user level, the extension is visible with a new storage
19457 class keyword: @code{__thread}. For example:
19461 extern __thread struct state s;
19462 static __thread char *p;
19465 The @code{__thread} specifier may be used alone, with the @code{extern}
19466 or @code{static} specifiers, but with no other storage class specifier.
19467 When used with @code{extern} or @code{static}, @code{__thread} must appear
19468 immediately after the other storage class specifier.
19470 The @code{__thread} specifier may be applied to any global, file-scoped
19471 static, function-scoped static, or static data member of a class. It may
19472 not be applied to block-scoped automatic or non-static data member.
19474 When the address-of operator is applied to a thread-local variable, it is
19475 evaluated at run time and returns the address of the current thread's
19476 instance of that variable. An address so obtained may be used by any
19477 thread. When a thread terminates, any pointers to thread-local variables
19478 in that thread become invalid.
19480 No static initialization may refer to the address of a thread-local variable.
19482 In C++, if an initializer is present for a thread-local variable, it must
19483 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
19486 See @uref{http://www.akkadia.org/drepper/tls.pdf,
19487 ELF Handling For Thread-Local Storage} for a detailed explanation of
19488 the four thread-local storage addressing models, and how the runtime
19489 is expected to function.
19492 * C99 Thread-Local Edits::
19493 * C++98 Thread-Local Edits::
19496 @node C99 Thread-Local Edits
19497 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
19499 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
19500 that document the exact semantics of the language extension.
19504 @cite{5.1.2 Execution environments}
19506 Add new text after paragraph 1
19509 Within either execution environment, a @dfn{thread} is a flow of
19510 control within a program. It is implementation defined whether
19511 or not there may be more than one thread associated with a program.
19512 It is implementation defined how threads beyond the first are
19513 created, the name and type of the function called at thread
19514 startup, and how threads may be terminated. However, objects
19515 with thread storage duration shall be initialized before thread
19520 @cite{6.2.4 Storage durations of objects}
19522 Add new text before paragraph 3
19525 An object whose identifier is declared with the storage-class
19526 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
19527 Its lifetime is the entire execution of the thread, and its
19528 stored value is initialized only once, prior to thread startup.
19532 @cite{6.4.1 Keywords}
19534 Add @code{__thread}.
19537 @cite{6.7.1 Storage-class specifiers}
19539 Add @code{__thread} to the list of storage class specifiers in
19542 Change paragraph 2 to
19545 With the exception of @code{__thread}, at most one storage-class
19546 specifier may be given [@dots{}]. The @code{__thread} specifier may
19547 be used alone, or immediately following @code{extern} or
19551 Add new text after paragraph 6
19554 The declaration of an identifier for a variable that has
19555 block scope that specifies @code{__thread} shall also
19556 specify either @code{extern} or @code{static}.
19558 The @code{__thread} specifier shall be used only with
19563 @node C++98 Thread-Local Edits
19564 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
19566 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
19567 that document the exact semantics of the language extension.
19571 @b{[intro.execution]}
19573 New text after paragraph 4
19576 A @dfn{thread} is a flow of control within the abstract machine.
19577 It is implementation defined whether or not there may be more than
19581 New text after paragraph 7
19584 It is unspecified whether additional action must be taken to
19585 ensure when and whether side effects are visible to other threads.
19591 Add @code{__thread}.
19594 @b{[basic.start.main]}
19596 Add after paragraph 5
19599 The thread that begins execution at the @code{main} function is called
19600 the @dfn{main thread}. It is implementation defined how functions
19601 beginning threads other than the main thread are designated or typed.
19602 A function so designated, as well as the @code{main} function, is called
19603 a @dfn{thread startup function}. It is implementation defined what
19604 happens if a thread startup function returns. It is implementation
19605 defined what happens to other threads when any thread calls @code{exit}.
19609 @b{[basic.start.init]}
19611 Add after paragraph 4
19614 The storage for an object of thread storage duration shall be
19615 statically initialized before the first statement of the thread startup
19616 function. An object of thread storage duration shall not require
19617 dynamic initialization.
19621 @b{[basic.start.term]}
19623 Add after paragraph 3
19626 The type of an object with thread storage duration shall not have a
19627 non-trivial destructor, nor shall it be an array type whose elements
19628 (directly or indirectly) have non-trivial destructors.
19634 Add ``thread storage duration'' to the list in paragraph 1.
19639 Thread, static, and automatic storage durations are associated with
19640 objects introduced by declarations [@dots{}].
19643 Add @code{__thread} to the list of specifiers in paragraph 3.
19646 @b{[basic.stc.thread]}
19648 New section before @b{[basic.stc.static]}
19651 The keyword @code{__thread} applied to a non-local object gives the
19652 object thread storage duration.
19654 A local variable or class data member declared both @code{static}
19655 and @code{__thread} gives the variable or member thread storage
19660 @b{[basic.stc.static]}
19665 All objects that have neither thread storage duration, dynamic
19666 storage duration nor are local [@dots{}].
19672 Add @code{__thread} to the list in paragraph 1.
19677 With the exception of @code{__thread}, at most one
19678 @var{storage-class-specifier} shall appear in a given
19679 @var{decl-specifier-seq}. The @code{__thread} specifier may
19680 be used alone, or immediately following the @code{extern} or
19681 @code{static} specifiers. [@dots{}]
19684 Add after paragraph 5
19687 The @code{__thread} specifier can be applied only to the names of objects
19688 and to anonymous unions.
19694 Add after paragraph 6
19697 Non-@code{static} members shall not be @code{__thread}.
19701 @node Binary constants
19702 @section Binary Constants using the @samp{0b} Prefix
19703 @cindex Binary constants using the @samp{0b} prefix
19705 Integer constants can be written as binary constants, consisting of a
19706 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
19707 @samp{0B}. This is particularly useful in environments that operate a
19708 lot on the bit level (like microcontrollers).
19710 The following statements are identical:
19719 The type of these constants follows the same rules as for octal or
19720 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
19723 @node C++ Extensions
19724 @chapter Extensions to the C++ Language
19725 @cindex extensions, C++ language
19726 @cindex C++ language extensions
19728 The GNU compiler provides these extensions to the C++ language (and you
19729 can also use most of the C language extensions in your C++ programs). If you
19730 want to write code that checks whether these features are available, you can
19731 test for the GNU compiler the same way as for C programs: check for a
19732 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
19733 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
19734 Predefined Macros,cpp,The GNU C Preprocessor}).
19737 * C++ Volatiles:: What constitutes an access to a volatile object.
19738 * Restricted Pointers:: C99 restricted pointers and references.
19739 * Vague Linkage:: Where G++ puts inlines, vtables and such.
19740 * C++ Interface:: You can use a single C++ header file for both
19741 declarations and definitions.
19742 * Template Instantiation:: Methods for ensuring that exactly one copy of
19743 each needed template instantiation is emitted.
19744 * Bound member functions:: You can extract a function pointer to the
19745 method denoted by a @samp{->*} or @samp{.*} expression.
19746 * C++ Attributes:: Variable, function, and type attributes for C++ only.
19747 * Function Multiversioning:: Declaring multiple function versions.
19748 * Namespace Association:: Strong using-directives for namespace association.
19749 * Type Traits:: Compiler support for type traits.
19750 * C++ Concepts:: Improved support for generic programming.
19751 * Java Exceptions:: Tweaking exception handling to work with Java.
19752 * Deprecated Features:: Things will disappear from G++.
19753 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
19756 @node C++ Volatiles
19757 @section When is a Volatile C++ Object Accessed?
19758 @cindex accessing volatiles
19759 @cindex volatile read
19760 @cindex volatile write
19761 @cindex volatile access
19763 The C++ standard differs from the C standard in its treatment of
19764 volatile objects. It fails to specify what constitutes a volatile
19765 access, except to say that C++ should behave in a similar manner to C
19766 with respect to volatiles, where possible. However, the different
19767 lvalueness of expressions between C and C++ complicate the behavior.
19768 G++ behaves the same as GCC for volatile access, @xref{C
19769 Extensions,,Volatiles}, for a description of GCC's behavior.
19771 The C and C++ language specifications differ when an object is
19772 accessed in a void context:
19775 volatile int *src = @var{somevalue};
19779 The C++ standard specifies that such expressions do not undergo lvalue
19780 to rvalue conversion, and that the type of the dereferenced object may
19781 be incomplete. The C++ standard does not specify explicitly that it
19782 is lvalue to rvalue conversion that is responsible for causing an
19783 access. There is reason to believe that it is, because otherwise
19784 certain simple expressions become undefined. However, because it
19785 would surprise most programmers, G++ treats dereferencing a pointer to
19786 volatile object of complete type as GCC would do for an equivalent
19787 type in C@. When the object has incomplete type, G++ issues a
19788 warning; if you wish to force an error, you must force a conversion to
19789 rvalue with, for instance, a static cast.
19791 When using a reference to volatile, G++ does not treat equivalent
19792 expressions as accesses to volatiles, but instead issues a warning that
19793 no volatile is accessed. The rationale for this is that otherwise it
19794 becomes difficult to determine where volatile access occur, and not
19795 possible to ignore the return value from functions returning volatile
19796 references. Again, if you wish to force a read, cast the reference to
19799 G++ implements the same behavior as GCC does when assigning to a
19800 volatile object---there is no reread of the assigned-to object, the
19801 assigned rvalue is reused. Note that in C++ assignment expressions
19802 are lvalues, and if used as an lvalue, the volatile object is
19803 referred to. For instance, @var{vref} refers to @var{vobj}, as
19804 expected, in the following example:
19808 volatile int &vref = vobj = @var{something};
19811 @node Restricted Pointers
19812 @section Restricting Pointer Aliasing
19813 @cindex restricted pointers
19814 @cindex restricted references
19815 @cindex restricted this pointer
19817 As with the C front end, G++ understands the C99 feature of restricted pointers,
19818 specified with the @code{__restrict__}, or @code{__restrict} type
19819 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
19820 language flag, @code{restrict} is not a keyword in C++.
19822 In addition to allowing restricted pointers, you can specify restricted
19823 references, which indicate that the reference is not aliased in the local
19827 void fn (int *__restrict__ rptr, int &__restrict__ rref)
19834 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
19835 @var{rref} refers to a (different) unaliased integer.
19837 You may also specify whether a member function's @var{this} pointer is
19838 unaliased by using @code{__restrict__} as a member function qualifier.
19841 void T::fn () __restrict__
19848 Within the body of @code{T::fn}, @var{this} has the effective
19849 definition @code{T *__restrict__ const this}. Notice that the
19850 interpretation of a @code{__restrict__} member function qualifier is
19851 different to that of @code{const} or @code{volatile} qualifier, in that it
19852 is applied to the pointer rather than the object. This is consistent with
19853 other compilers that implement restricted pointers.
19855 As with all outermost parameter qualifiers, @code{__restrict__} is
19856 ignored in function definition matching. This means you only need to
19857 specify @code{__restrict__} in a function definition, rather than
19858 in a function prototype as well.
19860 @node Vague Linkage
19861 @section Vague Linkage
19862 @cindex vague linkage
19864 There are several constructs in C++ that require space in the object
19865 file but are not clearly tied to a single translation unit. We say that
19866 these constructs have ``vague linkage''. Typically such constructs are
19867 emitted wherever they are needed, though sometimes we can be more
19871 @item Inline Functions
19872 Inline functions are typically defined in a header file which can be
19873 included in many different compilations. Hopefully they can usually be
19874 inlined, but sometimes an out-of-line copy is necessary, if the address
19875 of the function is taken or if inlining fails. In general, we emit an
19876 out-of-line copy in all translation units where one is needed. As an
19877 exception, we only emit inline virtual functions with the vtable, since
19878 it always requires a copy.
19880 Local static variables and string constants used in an inline function
19881 are also considered to have vague linkage, since they must be shared
19882 between all inlined and out-of-line instances of the function.
19886 C++ virtual functions are implemented in most compilers using a lookup
19887 table, known as a vtable. The vtable contains pointers to the virtual
19888 functions provided by a class, and each object of the class contains a
19889 pointer to its vtable (or vtables, in some multiple-inheritance
19890 situations). If the class declares any non-inline, non-pure virtual
19891 functions, the first one is chosen as the ``key method'' for the class,
19892 and the vtable is only emitted in the translation unit where the key
19895 @emph{Note:} If the chosen key method is later defined as inline, the
19896 vtable is still emitted in every translation unit that defines it.
19897 Make sure that any inline virtuals are declared inline in the class
19898 body, even if they are not defined there.
19900 @item @code{type_info} objects
19901 @cindex @code{type_info}
19903 C++ requires information about types to be written out in order to
19904 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
19905 For polymorphic classes (classes with virtual functions), the @samp{type_info}
19906 object is written out along with the vtable so that @samp{dynamic_cast}
19907 can determine the dynamic type of a class object at run time. For all
19908 other types, we write out the @samp{type_info} object when it is used: when
19909 applying @samp{typeid} to an expression, throwing an object, or
19910 referring to a type in a catch clause or exception specification.
19912 @item Template Instantiations
19913 Most everything in this section also applies to template instantiations,
19914 but there are other options as well.
19915 @xref{Template Instantiation,,Where's the Template?}.
19919 When used with GNU ld version 2.8 or later on an ELF system such as
19920 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
19921 these constructs will be discarded at link time. This is known as
19924 On targets that don't support COMDAT, but do support weak symbols, GCC
19925 uses them. This way one copy overrides all the others, but
19926 the unused copies still take up space in the executable.
19928 For targets that do not support either COMDAT or weak symbols,
19929 most entities with vague linkage are emitted as local symbols to
19930 avoid duplicate definition errors from the linker. This does not happen
19931 for local statics in inlines, however, as having multiple copies
19932 almost certainly breaks things.
19934 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
19935 another way to control placement of these constructs.
19937 @node C++ Interface
19938 @section C++ Interface and Implementation Pragmas
19940 @cindex interface and implementation headers, C++
19941 @cindex C++ interface and implementation headers
19942 @cindex pragmas, interface and implementation
19944 @code{#pragma interface} and @code{#pragma implementation} provide the
19945 user with a way of explicitly directing the compiler to emit entities
19946 with vague linkage (and debugging information) in a particular
19949 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
19950 by COMDAT support and the ``key method'' heuristic
19951 mentioned in @ref{Vague Linkage}. Using them can actually cause your
19952 program to grow due to unnecessary out-of-line copies of inline
19956 @item #pragma interface
19957 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
19958 @kindex #pragma interface
19959 Use this directive in @emph{header files} that define object classes, to save
19960 space in most of the object files that use those classes. Normally,
19961 local copies of certain information (backup copies of inline member
19962 functions, debugging information, and the internal tables that implement
19963 virtual functions) must be kept in each object file that includes class
19964 definitions. You can use this pragma to avoid such duplication. When a
19965 header file containing @samp{#pragma interface} is included in a
19966 compilation, this auxiliary information is not generated (unless
19967 the main input source file itself uses @samp{#pragma implementation}).
19968 Instead, the object files contain references to be resolved at link
19971 The second form of this directive is useful for the case where you have
19972 multiple headers with the same name in different directories. If you
19973 use this form, you must specify the same string to @samp{#pragma
19976 @item #pragma implementation
19977 @itemx #pragma implementation "@var{objects}.h"
19978 @kindex #pragma implementation
19979 Use this pragma in a @emph{main input file}, when you want full output from
19980 included header files to be generated (and made globally visible). The
19981 included header file, in turn, should use @samp{#pragma interface}.
19982 Backup copies of inline member functions, debugging information, and the
19983 internal tables used to implement virtual functions are all generated in
19984 implementation files.
19986 @cindex implied @code{#pragma implementation}
19987 @cindex @code{#pragma implementation}, implied
19988 @cindex naming convention, implementation headers
19989 If you use @samp{#pragma implementation} with no argument, it applies to
19990 an include file with the same basename@footnote{A file's @dfn{basename}
19991 is the name stripped of all leading path information and of trailing
19992 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
19993 file. For example, in @file{allclass.cc}, giving just
19994 @samp{#pragma implementation}
19995 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
19997 Use the string argument if you want a single implementation file to
19998 include code from multiple header files. (You must also use
19999 @samp{#include} to include the header file; @samp{#pragma
20000 implementation} only specifies how to use the file---it doesn't actually
20003 There is no way to split up the contents of a single header file into
20004 multiple implementation files.
20007 @cindex inlining and C++ pragmas
20008 @cindex C++ pragmas, effect on inlining
20009 @cindex pragmas in C++, effect on inlining
20010 @samp{#pragma implementation} and @samp{#pragma interface} also have an
20011 effect on function inlining.
20013 If you define a class in a header file marked with @samp{#pragma
20014 interface}, the effect on an inline function defined in that class is
20015 similar to an explicit @code{extern} declaration---the compiler emits
20016 no code at all to define an independent version of the function. Its
20017 definition is used only for inlining with its callers.
20019 @opindex fno-implement-inlines
20020 Conversely, when you include the same header file in a main source file
20021 that declares it as @samp{#pragma implementation}, the compiler emits
20022 code for the function itself; this defines a version of the function
20023 that can be found via pointers (or by callers compiled without
20024 inlining). If all calls to the function can be inlined, you can avoid
20025 emitting the function by compiling with @option{-fno-implement-inlines}.
20026 If any calls are not inlined, you will get linker errors.
20028 @node Template Instantiation
20029 @section Where's the Template?
20030 @cindex template instantiation
20032 C++ templates were the first language feature to require more
20033 intelligence from the environment than was traditionally found on a UNIX
20034 system. Somehow the compiler and linker have to make sure that each
20035 template instance occurs exactly once in the executable if it is needed,
20036 and not at all otherwise. There are two basic approaches to this
20037 problem, which are referred to as the Borland model and the Cfront model.
20040 @item Borland model
20041 Borland C++ solved the template instantiation problem by adding the code
20042 equivalent of common blocks to their linker; the compiler emits template
20043 instances in each translation unit that uses them, and the linker
20044 collapses them together. The advantage of this model is that the linker
20045 only has to consider the object files themselves; there is no external
20046 complexity to worry about. The disadvantage is that compilation time
20047 is increased because the template code is being compiled repeatedly.
20048 Code written for this model tends to include definitions of all
20049 templates in the header file, since they must be seen to be
20053 The AT&T C++ translator, Cfront, solved the template instantiation
20054 problem by creating the notion of a template repository, an
20055 automatically maintained place where template instances are stored. A
20056 more modern version of the repository works as follows: As individual
20057 object files are built, the compiler places any template definitions and
20058 instantiations encountered in the repository. At link time, the link
20059 wrapper adds in the objects in the repository and compiles any needed
20060 instances that were not previously emitted. The advantages of this
20061 model are more optimal compilation speed and the ability to use the
20062 system linker; to implement the Borland model a compiler vendor also
20063 needs to replace the linker. The disadvantages are vastly increased
20064 complexity, and thus potential for error; for some code this can be
20065 just as transparent, but in practice it can been very difficult to build
20066 multiple programs in one directory and one program in multiple
20067 directories. Code written for this model tends to separate definitions
20068 of non-inline member templates into a separate file, which should be
20069 compiled separately.
20072 G++ implements the Borland model on targets where the linker supports it,
20073 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
20074 Otherwise G++ implements neither automatic model.
20076 You have the following options for dealing with template instantiations:
20080 Do nothing. Code written for the Borland model works fine, but
20081 each translation unit contains instances of each of the templates it
20082 uses. The duplicate instances will be discarded by the linker, but in
20083 a large program, this can lead to an unacceptable amount of code
20084 duplication in object files or shared libraries.
20086 Duplicate instances of a template can be avoided by defining an explicit
20087 instantiation in one object file, and preventing the compiler from doing
20088 implicit instantiations in any other object files by using an explicit
20089 instantiation declaration, using the @code{extern template} syntax:
20092 extern template int max (int, int);
20095 This syntax is defined in the C++ 2011 standard, but has been supported by
20096 G++ and other compilers since well before 2011.
20098 Explicit instantiations can be used for the largest or most frequently
20099 duplicated instances, without having to know exactly which other instances
20100 are used in the rest of the program. You can scatter the explicit
20101 instantiations throughout your program, perhaps putting them in the
20102 translation units where the instances are used or the translation units
20103 that define the templates themselves; you can put all of the explicit
20104 instantiations you need into one big file; or you can create small files
20111 template class Foo<int>;
20112 template ostream& operator <<
20113 (ostream&, const Foo<int>&);
20117 for each of the instances you need, and create a template instantiation
20118 library from those.
20120 This is the simplest option, but also offers flexibility and
20121 fine-grained control when necessary. It is also the most portable
20122 alternative and programs using this approach will work with most modern
20127 Compile your template-using code with @option{-frepo}. The compiler
20128 generates files with the extension @samp{.rpo} listing all of the
20129 template instantiations used in the corresponding object files that
20130 could be instantiated there; the link wrapper, @samp{collect2},
20131 then updates the @samp{.rpo} files to tell the compiler where to place
20132 those instantiations and rebuild any affected object files. The
20133 link-time overhead is negligible after the first pass, as the compiler
20134 continues to place the instantiations in the same files.
20136 This can be a suitable option for application code written for the Borland
20137 model, as it usually just works. Code written for the Cfront model
20138 needs to be modified so that the template definitions are available at
20139 one or more points of instantiation; usually this is as simple as adding
20140 @code{#include <tmethods.cc>} to the end of each template header.
20142 For library code, if you want the library to provide all of the template
20143 instantiations it needs, just try to link all of its object files
20144 together; the link will fail, but cause the instantiations to be
20145 generated as a side effect. Be warned, however, that this may cause
20146 conflicts if multiple libraries try to provide the same instantiations.
20147 For greater control, use explicit instantiation as described in the next
20151 @opindex fno-implicit-templates
20152 Compile your code with @option{-fno-implicit-templates} to disable the
20153 implicit generation of template instances, and explicitly instantiate
20154 all the ones you use. This approach requires more knowledge of exactly
20155 which instances you need than do the others, but it's less
20156 mysterious and allows greater control if you want to ensure that only
20157 the intended instances are used.
20159 If you are using Cfront-model code, you can probably get away with not
20160 using @option{-fno-implicit-templates} when compiling files that don't
20161 @samp{#include} the member template definitions.
20163 If you use one big file to do the instantiations, you may want to
20164 compile it without @option{-fno-implicit-templates} so you get all of the
20165 instances required by your explicit instantiations (but not by any
20166 other files) without having to specify them as well.
20168 In addition to forward declaration of explicit instantiations
20169 (with @code{extern}), G++ has extended the template instantiation
20170 syntax to support instantiation of the compiler support data for a
20171 template class (i.e.@: the vtable) without instantiating any of its
20172 members (with @code{inline}), and instantiation of only the static data
20173 members of a template class, without the support data or member
20174 functions (with @code{static}):
20177 inline template class Foo<int>;
20178 static template class Foo<int>;
20182 @node Bound member functions
20183 @section Extracting the Function Pointer from a Bound Pointer to Member Function
20185 @cindex pointer to member function
20186 @cindex bound pointer to member function
20188 In C++, pointer to member functions (PMFs) are implemented using a wide
20189 pointer of sorts to handle all the possible call mechanisms; the PMF
20190 needs to store information about how to adjust the @samp{this} pointer,
20191 and if the function pointed to is virtual, where to find the vtable, and
20192 where in the vtable to look for the member function. If you are using
20193 PMFs in an inner loop, you should really reconsider that decision. If
20194 that is not an option, you can extract the pointer to the function that
20195 would be called for a given object/PMF pair and call it directly inside
20196 the inner loop, to save a bit of time.
20198 Note that you still pay the penalty for the call through a
20199 function pointer; on most modern architectures, such a call defeats the
20200 branch prediction features of the CPU@. This is also true of normal
20201 virtual function calls.
20203 The syntax for this extension is
20207 extern int (A::*fp)();
20208 typedef int (*fptr)(A *);
20210 fptr p = (fptr)(a.*fp);
20213 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
20214 no object is needed to obtain the address of the function. They can be
20215 converted to function pointers directly:
20218 fptr p1 = (fptr)(&A::foo);
20221 @opindex Wno-pmf-conversions
20222 You must specify @option{-Wno-pmf-conversions} to use this extension.
20224 @node C++ Attributes
20225 @section C++-Specific Variable, Function, and Type Attributes
20227 Some attributes only make sense for C++ programs.
20230 @item abi_tag ("@var{tag}", ...)
20231 @cindex @code{abi_tag} function attribute
20232 @cindex @code{abi_tag} variable attribute
20233 @cindex @code{abi_tag} type attribute
20234 The @code{abi_tag} attribute can be applied to a function, variable, or class
20235 declaration. It modifies the mangled name of the entity to
20236 incorporate the tag name, in order to distinguish the function or
20237 class from an earlier version with a different ABI; perhaps the class
20238 has changed size, or the function has a different return type that is
20239 not encoded in the mangled name.
20241 The attribute can also be applied to an inline namespace, but does not
20242 affect the mangled name of the namespace; in this case it is only used
20243 for @option{-Wabi-tag} warnings and automatic tagging of functions and
20244 variables. Tagging inline namespaces is generally preferable to
20245 tagging individual declarations, but the latter is sometimes
20246 necessary, such as when only certain members of a class need to be
20249 The argument can be a list of strings of arbitrary length. The
20250 strings are sorted on output, so the order of the list is
20253 A redeclaration of an entity must not add new ABI tags,
20254 since doing so would change the mangled name.
20256 The ABI tags apply to a name, so all instantiations and
20257 specializations of a template have the same tags. The attribute will
20258 be ignored if applied to an explicit specialization or instantiation.
20260 The @option{-Wabi-tag} flag enables a warning about a class which does
20261 not have all the ABI tags used by its subobjects and virtual functions; for users with code
20262 that needs to coexist with an earlier ABI, using this option can help
20263 to find all affected types that need to be tagged.
20265 When a type involving an ABI tag is used as the type of a variable or
20266 return type of a function where that tag is not already present in the
20267 signature of the function, the tag is automatically applied to the
20268 variable or function. @option{-Wabi-tag} also warns about this
20269 situation; this warning can be avoided by explicitly tagging the
20270 variable or function or moving it into a tagged inline namespace.
20272 @item init_priority (@var{priority})
20273 @cindex @code{init_priority} variable attribute
20275 In Standard C++, objects defined at namespace scope are guaranteed to be
20276 initialized in an order in strict accordance with that of their definitions
20277 @emph{in a given translation unit}. No guarantee is made for initializations
20278 across translation units. However, GNU C++ allows users to control the
20279 order of initialization of objects defined at namespace scope with the
20280 @code{init_priority} attribute by specifying a relative @var{priority},
20281 a constant integral expression currently bounded between 101 and 65535
20282 inclusive. Lower numbers indicate a higher priority.
20284 In the following example, @code{A} would normally be created before
20285 @code{B}, but the @code{init_priority} attribute reverses that order:
20288 Some_Class A __attribute__ ((init_priority (2000)));
20289 Some_Class B __attribute__ ((init_priority (543)));
20293 Note that the particular values of @var{priority} do not matter; only their
20296 @item java_interface
20297 @cindex @code{java_interface} type attribute
20299 This type attribute informs C++ that the class is a Java interface. It may
20300 only be applied to classes declared within an @code{extern "Java"} block.
20301 Calls to methods declared in this interface are dispatched using GCJ's
20302 interface table mechanism, instead of regular virtual table dispatch.
20305 @cindex @code{warn_unused} type attribute
20307 For C++ types with non-trivial constructors and/or destructors it is
20308 impossible for the compiler to determine whether a variable of this
20309 type is truly unused if it is not referenced. This type attribute
20310 informs the compiler that variables of this type should be warned
20311 about if they appear to be unused, just like variables of fundamental
20314 This attribute is appropriate for types which just represent a value,
20315 such as @code{std::string}; it is not appropriate for types which
20316 control a resource, such as @code{std::mutex}.
20318 This attribute is also accepted in C, but it is unnecessary because C
20319 does not have constructors or destructors.
20323 See also @ref{Namespace Association}.
20325 @node Function Multiversioning
20326 @section Function Multiversioning
20327 @cindex function versions
20329 With the GNU C++ front end, for x86 targets, you may specify multiple
20330 versions of a function, where each function is specialized for a
20331 specific target feature. At runtime, the appropriate version of the
20332 function is automatically executed depending on the characteristics of
20333 the execution platform. Here is an example.
20336 __attribute__ ((target ("default")))
20339 // The default version of foo.
20343 __attribute__ ((target ("sse4.2")))
20346 // foo version for SSE4.2
20350 __attribute__ ((target ("arch=atom")))
20353 // foo version for the Intel ATOM processor
20357 __attribute__ ((target ("arch=amdfam10")))
20360 // foo version for the AMD Family 0x10 processors.
20367 assert ((*p) () == foo ());
20372 In the above example, four versions of function foo are created. The
20373 first version of foo with the target attribute "default" is the default
20374 version. This version gets executed when no other target specific
20375 version qualifies for execution on a particular platform. A new version
20376 of foo is created by using the same function signature but with a
20377 different target string. Function foo is called or a pointer to it is
20378 taken just like a regular function. GCC takes care of doing the
20379 dispatching to call the right version at runtime. Refer to the
20380 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
20381 Function Multiversioning} for more details.
20383 @node Namespace Association
20384 @section Namespace Association
20386 @strong{Caution:} The semantics of this extension are equivalent
20387 to C++ 2011 inline namespaces. Users should use inline namespaces
20388 instead as this extension will be removed in future versions of G++.
20390 A using-directive with @code{__attribute ((strong))} is stronger
20391 than a normal using-directive in two ways:
20395 Templates from the used namespace can be specialized and explicitly
20396 instantiated as though they were members of the using namespace.
20399 The using namespace is considered an associated namespace of all
20400 templates in the used namespace for purposes of argument-dependent
20404 The used namespace must be nested within the using namespace so that
20405 normal unqualified lookup works properly.
20407 This is useful for composing a namespace transparently from
20408 implementation namespaces. For example:
20413 template <class T> struct A @{ @};
20415 using namespace debug __attribute ((__strong__));
20416 template <> struct A<int> @{ @}; // @r{OK to specialize}
20418 template <class T> void f (A<T>);
20423 f (std::A<float>()); // @r{lookup finds} std::f
20429 @section Type Traits
20431 The C++ front end implements syntactic extensions that allow
20432 compile-time determination of
20433 various characteristics of a type (or of a
20437 @item __has_nothrow_assign (type)
20438 If @code{type} is const qualified or is a reference type then the trait is
20439 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
20440 is true, else if @code{type} is a cv class or union type with copy assignment
20441 operators that are known not to throw an exception then the trait is true,
20442 else it is false. Requires: @code{type} shall be a complete type,
20443 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20445 @item __has_nothrow_copy (type)
20446 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
20447 @code{type} is a cv class or union type with copy constructors that
20448 are known not to throw an exception then the trait is true, else it is false.
20449 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
20450 @code{void}, or an array of unknown bound.
20452 @item __has_nothrow_constructor (type)
20453 If @code{__has_trivial_constructor (type)} is true then the trait is
20454 true, else if @code{type} is a cv class or union type (or array
20455 thereof) with a default constructor that is known not to throw an
20456 exception then the trait is true, else it is false. Requires:
20457 @code{type} shall be a complete type, (possibly cv-qualified)
20458 @code{void}, or an array of unknown bound.
20460 @item __has_trivial_assign (type)
20461 If @code{type} is const qualified or is a reference type then the trait is
20462 false. Otherwise if @code{__is_pod (type)} is true then the trait is
20463 true, else if @code{type} is a cv class or union type with a trivial
20464 copy assignment ([class.copy]) then the trait is true, else it is
20465 false. Requires: @code{type} shall be a complete type, (possibly
20466 cv-qualified) @code{void}, or an array of unknown bound.
20468 @item __has_trivial_copy (type)
20469 If @code{__is_pod (type)} is true or @code{type} is a reference type
20470 then the trait is true, else if @code{type} is a cv class or union type
20471 with a trivial copy constructor ([class.copy]) then the trait
20472 is true, else it is false. Requires: @code{type} shall be a complete
20473 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20475 @item __has_trivial_constructor (type)
20476 If @code{__is_pod (type)} is true then the trait is true, else if
20477 @code{type} is a cv class or union type (or array thereof) with a
20478 trivial default constructor ([class.ctor]) then the trait is true,
20479 else it is false. Requires: @code{type} shall be a complete
20480 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20482 @item __has_trivial_destructor (type)
20483 If @code{__is_pod (type)} is true or @code{type} is a reference type then
20484 the trait is true, else if @code{type} is a cv class or union type (or
20485 array thereof) with a trivial destructor ([class.dtor]) then the trait
20486 is true, else it is false. Requires: @code{type} shall be a complete
20487 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20489 @item __has_virtual_destructor (type)
20490 If @code{type} is a class type with a virtual destructor
20491 ([class.dtor]) then the trait is true, else it is false. Requires:
20492 @code{type} shall be a complete type, (possibly cv-qualified)
20493 @code{void}, or an array of unknown bound.
20495 @item __is_abstract (type)
20496 If @code{type} is an abstract class ([class.abstract]) then the trait
20497 is true, else it is false. Requires: @code{type} shall be a complete
20498 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20500 @item __is_base_of (base_type, derived_type)
20501 If @code{base_type} is a base class of @code{derived_type}
20502 ([class.derived]) then the trait is true, otherwise it is false.
20503 Top-level cv qualifications of @code{base_type} and
20504 @code{derived_type} are ignored. For the purposes of this trait, a
20505 class type is considered is own base. Requires: if @code{__is_class
20506 (base_type)} and @code{__is_class (derived_type)} are true and
20507 @code{base_type} and @code{derived_type} are not the same type
20508 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
20509 type. A diagnostic is produced if this requirement is not met.
20511 @item __is_class (type)
20512 If @code{type} is a cv class type, and not a union type
20513 ([basic.compound]) the trait is true, else it is false.
20515 @item __is_empty (type)
20516 If @code{__is_class (type)} is false then the trait is false.
20517 Otherwise @code{type} is considered empty if and only if: @code{type}
20518 has no non-static data members, or all non-static data members, if
20519 any, are bit-fields of length 0, and @code{type} has no virtual
20520 members, and @code{type} has no virtual base classes, and @code{type}
20521 has no base classes @code{base_type} for which
20522 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
20523 be a complete type, (possibly cv-qualified) @code{void}, or an array
20526 @item __is_enum (type)
20527 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
20528 true, else it is false.
20530 @item __is_literal_type (type)
20531 If @code{type} is a literal type ([basic.types]) the trait is
20532 true, else it is false. Requires: @code{type} shall be a complete type,
20533 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20535 @item __is_pod (type)
20536 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
20537 else it is false. Requires: @code{type} shall be a complete type,
20538 (possibly cv-qualified) @code{void}, or an array of unknown bound.
20540 @item __is_polymorphic (type)
20541 If @code{type} is a polymorphic class ([class.virtual]) then the trait
20542 is true, else it is false. Requires: @code{type} shall be a complete
20543 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20545 @item __is_standard_layout (type)
20546 If @code{type} is a standard-layout type ([basic.types]) the trait is
20547 true, else it is false. Requires: @code{type} shall be a complete
20548 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20550 @item __is_trivial (type)
20551 If @code{type} is a trivial type ([basic.types]) the trait is
20552 true, else it is false. Requires: @code{type} shall be a complete
20553 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
20555 @item __is_union (type)
20556 If @code{type} is a cv union type ([basic.compound]) the trait is
20557 true, else it is false.
20559 @item __underlying_type (type)
20560 The underlying type of @code{type}. Requires: @code{type} shall be
20561 an enumeration type ([dcl.enum]).
20567 @section C++ Concepts
20569 C++ concepts provide much-improved support for generic programming. In
20570 particular, they allow the specification of constraints on template arguments.
20571 The constraints are used to extend the usual overloading and partial
20572 specialization capabilities of the language, allowing generic data structures
20573 and algorithms to be ``refined'' based on their properties rather than their
20576 The following keywords are reserved for concepts.
20580 States an expression as an assumption, and if possible, verifies that the
20581 assumption is valid. For example, @code{assume(n > 0)}.
20584 Introduces an axiom definition. Axioms introduce requirements on values.
20587 Introduces a universally quantified object in an axiom. For example,
20588 @code{forall (int n) n + 0 == n}).
20591 Introduces a concept definition. Concepts are sets of syntactic and semantic
20592 requirements on types and their values.
20595 Introduces constraints on template arguments or requirements for a member
20596 function of a class template.
20600 The front end also exposes a number of internal mechanism that can be used
20601 to simplify the writing of type traits. Note that some of these traits are
20602 likely to be removed in the future.
20605 @item __is_same (type1, type2)
20606 A binary type trait: true whenever the type arguments are the same.
20611 @node Java Exceptions
20612 @section Java Exceptions
20614 The Java language uses a slightly different exception handling model
20615 from C++. Normally, GNU C++ automatically detects when you are
20616 writing C++ code that uses Java exceptions, and handle them
20617 appropriately. However, if C++ code only needs to execute destructors
20618 when Java exceptions are thrown through it, GCC guesses incorrectly.
20619 Sample problematic code is:
20622 struct S @{ ~S(); @};
20623 extern void bar(); // @r{is written in Java, and may throw exceptions}
20632 The usual effect of an incorrect guess is a link failure, complaining of
20633 a missing routine called @samp{__gxx_personality_v0}.
20635 You can inform the compiler that Java exceptions are to be used in a
20636 translation unit, irrespective of what it might think, by writing
20637 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
20638 @samp{#pragma} must appear before any functions that throw or catch
20639 exceptions, or run destructors when exceptions are thrown through them.
20641 You cannot mix Java and C++ exceptions in the same translation unit. It
20642 is believed to be safe to throw a C++ exception from one file through
20643 another file compiled for the Java exception model, or vice versa, but
20644 there may be bugs in this area.
20646 @node Deprecated Features
20647 @section Deprecated Features
20649 In the past, the GNU C++ compiler was extended to experiment with new
20650 features, at a time when the C++ language was still evolving. Now that
20651 the C++ standard is complete, some of those features are superseded by
20652 superior alternatives. Using the old features might cause a warning in
20653 some cases that the feature will be dropped in the future. In other
20654 cases, the feature might be gone already.
20656 While the list below is not exhaustive, it documents some of the options
20657 that are now deprecated:
20660 @item -fexternal-templates
20661 @itemx -falt-external-templates
20662 These are two of the many ways for G++ to implement template
20663 instantiation. @xref{Template Instantiation}. The C++ standard clearly
20664 defines how template definitions have to be organized across
20665 implementation units. G++ has an implicit instantiation mechanism that
20666 should work just fine for standard-conforming code.
20668 @item -fstrict-prototype
20669 @itemx -fno-strict-prototype
20670 Previously it was possible to use an empty prototype parameter list to
20671 indicate an unspecified number of parameters (like C), rather than no
20672 parameters, as C++ demands. This feature has been removed, except where
20673 it is required for backwards compatibility. @xref{Backwards Compatibility}.
20676 G++ allows a virtual function returning @samp{void *} to be overridden
20677 by one returning a different pointer type. This extension to the
20678 covariant return type rules is now deprecated and will be removed from a
20681 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
20682 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
20683 and are now removed from G++. Code using these operators should be
20684 modified to use @code{std::min} and @code{std::max} instead.
20686 The named return value extension has been deprecated, and is now
20689 The use of initializer lists with new expressions has been deprecated,
20690 and is now removed from G++.
20692 Floating and complex non-type template parameters have been deprecated,
20693 and are now removed from G++.
20695 The implicit typename extension has been deprecated and is now
20698 The use of default arguments in function pointers, function typedefs
20699 and other places where they are not permitted by the standard is
20700 deprecated and will be removed from a future version of G++.
20702 G++ allows floating-point literals to appear in integral constant expressions,
20703 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
20704 This extension is deprecated and will be removed from a future version.
20706 G++ allows static data members of const floating-point type to be declared
20707 with an initializer in a class definition. The standard only allows
20708 initializers for static members of const integral types and const
20709 enumeration types so this extension has been deprecated and will be removed
20710 from a future version.
20712 @node Backwards Compatibility
20713 @section Backwards Compatibility
20714 @cindex Backwards Compatibility
20715 @cindex ARM [Annotated C++ Reference Manual]
20717 Now that there is a definitive ISO standard C++, G++ has a specification
20718 to adhere to. The C++ language evolved over time, and features that
20719 used to be acceptable in previous drafts of the standard, such as the ARM
20720 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
20721 compilation of C++ written to such drafts, G++ contains some backwards
20722 compatibilities. @emph{All such backwards compatibility features are
20723 liable to disappear in future versions of G++.} They should be considered
20724 deprecated. @xref{Deprecated Features}.
20728 If a variable is declared at for scope, it used to remain in scope until
20729 the end of the scope that contained the for statement (rather than just
20730 within the for scope). G++ retains this, but issues a warning, if such a
20731 variable is accessed outside the for scope.
20733 @item Implicit C language
20734 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
20735 scope to set the language. On such systems, all header files are
20736 implicitly scoped inside a C language scope. Also, an empty prototype
20737 @code{()} is treated as an unspecified number of arguments, rather
20738 than no arguments, as C++ demands.
20741 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
20742 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr