1 @c Copyright (C) 1988-2014 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 * Initializers:: Non-constant initializers.
50 * Compound Literals:: Compound literals give structures, unions
52 * Designated Inits:: Labeling elements of initializers.
53 * Case Ranges:: `case 1 ... 9' and such.
54 * Cast to Union:: Casting to union type from any member of the union.
55 * Mixed Declarations:: Mixing declarations and code.
56 * Function Attributes:: Declaring that functions have no side effects,
57 or that they can never return.
58 * Attribute Syntax:: Formal syntax for attributes.
59 * Function Prototypes:: Prototype declarations and old-style definitions.
60 * C++ Comments:: C++ comments are recognized.
61 * Dollar Signs:: Dollar sign is allowed in identifiers.
62 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
63 * Variable Attributes:: Specifying attributes of variables.
64 * Type Attributes:: Specifying attributes of types.
65 * Alignment:: Inquiring about the alignment of a type or variable.
66 * Inline:: Defining inline functions (as fast as macros).
67 * Volatiles:: What constitutes an access to a volatile object.
68 * Extended Asm:: Assembler instructions with C expressions as operands.
69 (With them you can define ``built-in'' functions.)
70 * Constraints:: Constraints for asm operands
71 * Asm Labels:: Specifying the assembler name to use for a C symbol.
72 * Explicit Reg Vars:: Defining variables residing in specified registers.
73 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
74 * Incomplete Enums:: @code{enum foo;}, with details to follow.
75 * Function Names:: Printable strings which are the name of the current
77 * Return Address:: Getting the return or frame address of a function.
78 * Vector Extensions:: Using vector instructions through built-in functions.
79 * Offsetof:: Special syntax for implementing @code{offsetof}.
80 * __sync Builtins:: Legacy built-in functions for atomic memory access.
81 * __atomic Builtins:: Atomic built-in functions with memory model.
82 * x86 specific memory model extensions for transactional memory:: x86 memory models.
83 * Object Size Checking:: Built-in functions for limited buffer overflow
85 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
86 * Other Builtins:: Other built-in functions.
87 * Target Builtins:: Built-in functions specific to particular targets.
88 * Target Format Checks:: Format checks specific to particular targets.
89 * Pragmas:: Pragmas accepted by GCC.
90 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
91 * Thread-Local:: Per-thread variables.
92 * Binary constants:: Binary constants using the @samp{0b} prefix.
96 @section Statements and Declarations in Expressions
97 @cindex statements inside expressions
98 @cindex declarations inside expressions
99 @cindex expressions containing statements
100 @cindex macros, statements in expressions
102 @c the above section title wrapped and causes an underfull hbox.. i
103 @c changed it from "within" to "in". --mew 4feb93
104 A compound statement enclosed in parentheses may appear as an expression
105 in GNU C@. This allows you to use loops, switches, and local variables
106 within an expression.
108 Recall that a compound statement is a sequence of statements surrounded
109 by braces; in this construct, parentheses go around the braces. For
113 (@{ int y = foo (); int z;
120 is a valid (though slightly more complex than necessary) expression
121 for the absolute value of @code{foo ()}.
123 The last thing in the compound statement should be an expression
124 followed by a semicolon; the value of this subexpression serves as the
125 value of the entire construct. (If you use some other kind of statement
126 last within the braces, the construct has type @code{void}, and thus
127 effectively no value.)
129 This feature is especially useful in making macro definitions ``safe'' (so
130 that they evaluate each operand exactly once). For example, the
131 ``maximum'' function is commonly defined as a macro in standard C as
135 #define max(a,b) ((a) > (b) ? (a) : (b))
139 @cindex side effects, macro argument
140 But this definition computes either @var{a} or @var{b} twice, with bad
141 results if the operand has side effects. In GNU C, if you know the
142 type of the operands (here taken as @code{int}), you can define
143 the macro safely as follows:
146 #define maxint(a,b) \
147 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
150 Embedded statements are not allowed in constant expressions, such as
151 the value of an enumeration constant, the width of a bit-field, or
152 the initial value of a static variable.
154 If you don't know the type of the operand, you can still do this, but you
155 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
157 In G++, the result value of a statement expression undergoes array and
158 function pointer decay, and is returned by value to the enclosing
159 expression. For instance, if @code{A} is a class, then
168 constructs a temporary @code{A} object to hold the result of the
169 statement expression, and that is used to invoke @code{Foo}.
170 Therefore the @code{this} pointer observed by @code{Foo} is not the
173 In a statement expression, any temporaries created within a statement
174 are destroyed at that statement's end. This makes statement
175 expressions inside macros slightly different from function calls. In
176 the latter case temporaries introduced during argument evaluation are
177 destroyed at the end of the statement that includes the function
178 call. In the statement expression case they are destroyed during
179 the statement expression. For instance,
182 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
183 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
193 has different places where temporaries are destroyed. For the
194 @code{macro} case, the temporary @code{X} is destroyed just after
195 the initialization of @code{b}. In the @code{function} case that
196 temporary is destroyed when the function returns.
198 These considerations mean that it is probably a bad idea to use
199 statement expressions of this form in header files that are designed to
200 work with C++. (Note that some versions of the GNU C Library contained
201 header files using statement expressions that lead to precisely this
204 Jumping into a statement expression with @code{goto} or using a
205 @code{switch} statement outside the statement expression with a
206 @code{case} or @code{default} label inside the statement expression is
207 not permitted. Jumping into a statement expression with a computed
208 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
209 Jumping out of a statement expression is permitted, but if the
210 statement expression is part of a larger expression then it is
211 unspecified which other subexpressions of that expression have been
212 evaluated except where the language definition requires certain
213 subexpressions to be evaluated before or after the statement
214 expression. In any case, as with a function call, the evaluation of a
215 statement expression is not interleaved with the evaluation of other
216 parts of the containing expression. For example,
219 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
223 calls @code{foo} and @code{bar1} and does not call @code{baz} but
224 may or may not call @code{bar2}. If @code{bar2} is called, it is
225 called after @code{foo} and before @code{bar1}.
228 @section Locally Declared Labels
230 @cindex macros, local labels
232 GCC allows you to declare @dfn{local labels} in any nested block
233 scope. A local label is just like an ordinary label, but you can
234 only reference it (with a @code{goto} statement, or by taking its
235 address) within the block in which it is declared.
237 A local label declaration looks like this:
240 __label__ @var{label};
247 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
250 Local label declarations must come at the beginning of the block,
251 before any ordinary declarations or statements.
253 The label declaration defines the label @emph{name}, but does not define
254 the label itself. You must do this in the usual way, with
255 @code{@var{label}:}, within the statements of the statement expression.
257 The local label feature is useful for complex macros. If a macro
258 contains nested loops, a @code{goto} can be useful for breaking out of
259 them. However, an ordinary label whose scope is the whole function
260 cannot be used: if the macro can be expanded several times in one
261 function, the label is multiply defined in that function. A
262 local label avoids this problem. For example:
265 #define SEARCH(value, array, target) \
268 typeof (target) _SEARCH_target = (target); \
269 typeof (*(array)) *_SEARCH_array = (array); \
272 for (i = 0; i < max; i++) \
273 for (j = 0; j < max; j++) \
274 if (_SEARCH_array[i][j] == _SEARCH_target) \
275 @{ (value) = i; goto found; @} \
281 This could also be written using a statement expression:
284 #define SEARCH(array, target) \
287 typeof (target) _SEARCH_target = (target); \
288 typeof (*(array)) *_SEARCH_array = (array); \
291 for (i = 0; i < max; i++) \
292 for (j = 0; j < max; j++) \
293 if (_SEARCH_array[i][j] == _SEARCH_target) \
294 @{ value = i; goto found; @} \
301 Local label declarations also make the labels they declare visible to
302 nested functions, if there are any. @xref{Nested Functions}, for details.
304 @node Labels as Values
305 @section Labels as Values
306 @cindex labels as values
307 @cindex computed gotos
308 @cindex goto with computed label
309 @cindex address of a label
311 You can get the address of a label defined in the current function
312 (or a containing function) with the unary operator @samp{&&}. The
313 value has type @code{void *}. This value is a constant and can be used
314 wherever a constant of that type is valid. For example:
322 To use these values, you need to be able to jump to one. This is done
323 with the computed goto statement@footnote{The analogous feature in
324 Fortran is called an assigned goto, but that name seems inappropriate in
325 C, where one can do more than simply store label addresses in label
326 variables.}, @code{goto *@var{exp};}. For example,
333 Any expression of type @code{void *} is allowed.
335 One way of using these constants is in initializing a static array that
336 serves as a jump table:
339 static void *array[] = @{ &&foo, &&bar, &&hack @};
343 Then you can select a label with indexing, like this:
350 Note that this does not check whether the subscript is in bounds---array
351 indexing in C never does that.
353 Such an array of label values serves a purpose much like that of the
354 @code{switch} statement. The @code{switch} statement is cleaner, so
355 use that rather than an array unless the problem does not fit a
356 @code{switch} statement very well.
358 Another use of label values is in an interpreter for threaded code.
359 The labels within the interpreter function can be stored in the
360 threaded code for super-fast dispatching.
362 You may not use this mechanism to jump to code in a different function.
363 If you do that, totally unpredictable things happen. The best way to
364 avoid this is to store the label address only in automatic variables and
365 never pass it as an argument.
367 An alternate way to write the above example is
370 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
372 goto *(&&foo + array[i]);
376 This is more friendly to code living in shared libraries, as it reduces
377 the number of dynamic relocations that are needed, and by consequence,
378 allows the data to be read-only.
380 The @code{&&foo} expressions for the same label might have different
381 values if the containing function is inlined or cloned. If a program
382 relies on them being always the same,
383 @code{__attribute__((__noinline__,__noclone__))} should be used to
384 prevent inlining and cloning. If @code{&&foo} is used in a static
385 variable initializer, inlining and cloning is forbidden.
387 @node Nested Functions
388 @section Nested Functions
389 @cindex nested functions
390 @cindex downward funargs
393 A @dfn{nested function} is a function defined inside another function.
394 Nested functions are supported as an extension in GNU C, but are not
395 supported by GNU C++.
397 The nested function's name is local to the block where it is defined.
398 For example, here we define a nested function named @code{square}, and
403 foo (double a, double b)
405 double square (double z) @{ return z * z; @}
407 return square (a) + square (b);
412 The nested function can access all the variables of the containing
413 function that are visible at the point of its definition. This is
414 called @dfn{lexical scoping}. For example, here we show a nested
415 function which uses an inherited variable named @code{offset}:
419 bar (int *array, int offset, int size)
421 int access (int *array, int index)
422 @{ return array[index + offset]; @}
425 for (i = 0; i < size; i++)
426 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
431 Nested function definitions are permitted within functions in the places
432 where variable definitions are allowed; that is, in any block, mixed
433 with the other declarations and statements in the block.
435 It is possible to call the nested function from outside the scope of its
436 name by storing its address or passing the address to another function:
439 hack (int *array, int size)
441 void store (int index, int value)
442 @{ array[index] = value; @}
444 intermediate (store, size);
448 Here, the function @code{intermediate} receives the address of
449 @code{store} as an argument. If @code{intermediate} calls @code{store},
450 the arguments given to @code{store} are used to store into @code{array}.
451 But this technique works only so long as the containing function
452 (@code{hack}, in this example) does not exit.
454 If you try to call the nested function through its address after the
455 containing function exits, all hell breaks loose. If you try
456 to call it after a containing scope level exits, and if it refers
457 to some of the variables that are no longer in scope, you may be lucky,
458 but it's not wise to take the risk. If, however, the nested function
459 does not refer to anything that has gone out of scope, you should be
462 GCC implements taking the address of a nested function using a technique
463 called @dfn{trampolines}. This technique was described in
464 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
465 C++ Conference Proceedings, October 17-21, 1988).
467 A nested function can jump to a label inherited from a containing
468 function, provided the label is explicitly declared in the containing
469 function (@pxref{Local Labels}). Such a jump returns instantly to the
470 containing function, exiting the nested function that did the
471 @code{goto} and any intermediate functions as well. Here is an example:
475 bar (int *array, int offset, int size)
478 int access (int *array, int index)
482 return array[index + offset];
486 for (i = 0; i < size; i++)
487 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
491 /* @r{Control comes here from @code{access}
492 if it detects an error.} */
499 A nested function always has no linkage. Declaring one with
500 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
501 before its definition, use @code{auto} (which is otherwise meaningless
502 for function declarations).
505 bar (int *array, int offset, int size)
508 auto int access (int *, int);
510 int access (int *array, int index)
514 return array[index + offset];
520 @node Constructing Calls
521 @section Constructing Function Calls
522 @cindex constructing calls
523 @cindex forwarding calls
525 Using the built-in functions described below, you can record
526 the arguments a function received, and call another function
527 with the same arguments, without knowing the number or types
530 You can also record the return value of that function call,
531 and later return that value, without knowing what data type
532 the function tried to return (as long as your caller expects
535 However, these built-in functions may interact badly with some
536 sophisticated features or other extensions of the language. It
537 is, therefore, not recommended to use them outside very simple
538 functions acting as mere forwarders for their arguments.
540 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
541 This built-in function returns a pointer to data
542 describing how to perform a call with the same arguments as are passed
543 to the current function.
545 The function saves the arg pointer register, structure value address,
546 and all registers that might be used to pass arguments to a function
547 into a block of memory allocated on the stack. Then it returns the
548 address of that block.
551 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
552 This built-in function invokes @var{function}
553 with a copy of the parameters described by @var{arguments}
556 The value of @var{arguments} should be the value returned by
557 @code{__builtin_apply_args}. The argument @var{size} specifies the size
558 of the stack argument data, in bytes.
560 This function returns a pointer to data describing
561 how to return whatever value is returned by @var{function}. The data
562 is saved in a block of memory allocated on the stack.
564 It is not always simple to compute the proper value for @var{size}. The
565 value is used by @code{__builtin_apply} to compute the amount of data
566 that should be pushed on the stack and copied from the incoming argument
570 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
571 This built-in function returns the value described by @var{result} from
572 the containing function. You should specify, for @var{result}, a value
573 returned by @code{__builtin_apply}.
576 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
577 This built-in function represents all anonymous arguments of an inline
578 function. It can be used only in inline functions that are always
579 inlined, never compiled as a separate function, such as those using
580 @code{__attribute__ ((__always_inline__))} or
581 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
582 It must be only passed as last argument to some other function
583 with variable arguments. This is useful for writing small wrapper
584 inlines for variable argument functions, when using preprocessor
585 macros is undesirable. For example:
587 extern int myprintf (FILE *f, const char *format, ...);
588 extern inline __attribute__ ((__gnu_inline__)) int
589 myprintf (FILE *f, const char *format, ...)
591 int r = fprintf (f, "myprintf: ");
594 int s = fprintf (f, format, __builtin_va_arg_pack ());
602 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
603 This built-in function returns the number of anonymous arguments of
604 an inline function. It can be used only in inline functions that
605 are always inlined, never compiled as a separate function, such
606 as those using @code{__attribute__ ((__always_inline__))} or
607 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
608 For example following does link- or run-time checking of open
609 arguments for optimized code:
612 extern inline __attribute__((__gnu_inline__)) int
613 myopen (const char *path, int oflag, ...)
615 if (__builtin_va_arg_pack_len () > 1)
616 warn_open_too_many_arguments ();
618 if (__builtin_constant_p (oflag))
620 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
622 warn_open_missing_mode ();
623 return __open_2 (path, oflag);
625 return open (path, oflag, __builtin_va_arg_pack ());
628 if (__builtin_va_arg_pack_len () < 1)
629 return __open_2 (path, oflag);
631 return open (path, oflag, __builtin_va_arg_pack ());
638 @section Referring to a Type with @code{typeof}
641 @cindex macros, types of arguments
643 Another way to refer to the type of an expression is with @code{typeof}.
644 The syntax of using of this keyword looks like @code{sizeof}, but the
645 construct acts semantically like a type name defined with @code{typedef}.
647 There are two ways of writing the argument to @code{typeof}: with an
648 expression or with a type. Here is an example with an expression:
655 This assumes that @code{x} is an array of pointers to functions;
656 the type described is that of the values of the functions.
658 Here is an example with a typename as the argument:
665 Here the type described is that of pointers to @code{int}.
667 If you are writing a header file that must work when included in ISO C
668 programs, write @code{__typeof__} instead of @code{typeof}.
669 @xref{Alternate Keywords}.
671 A @code{typeof} construct can be used anywhere a typedef name can be
672 used. For example, you can use it in a declaration, in a cast, or inside
673 of @code{sizeof} or @code{typeof}.
675 The operand of @code{typeof} is evaluated for its side effects if and
676 only if it is an expression of variably modified type or the name of
679 @code{typeof} is often useful in conjunction with
680 statement expressions (@pxref{Statement Exprs}).
681 Here is how the two together can
682 be used to define a safe ``maximum'' macro which operates on any
683 arithmetic type and evaluates each of its arguments exactly once:
687 (@{ typeof (a) _a = (a); \
688 typeof (b) _b = (b); \
689 _a > _b ? _a : _b; @})
692 @cindex underscores in variables in macros
693 @cindex @samp{_} in variables in macros
694 @cindex local variables in macros
695 @cindex variables, local, in macros
696 @cindex macros, local variables in
698 The reason for using names that start with underscores for the local
699 variables is to avoid conflicts with variable names that occur within the
700 expressions that are substituted for @code{a} and @code{b}. Eventually we
701 hope to design a new form of declaration syntax that allows you to declare
702 variables whose scopes start only after their initializers; this will be a
703 more reliable way to prevent such conflicts.
706 Some more examples of the use of @code{typeof}:
710 This declares @code{y} with the type of what @code{x} points to.
717 This declares @code{y} as an array of such values.
724 This declares @code{y} as an array of pointers to characters:
727 typeof (typeof (char *)[4]) y;
731 It is equivalent to the following traditional C declaration:
737 To see the meaning of the declaration using @code{typeof}, and why it
738 might be a useful way to write, rewrite it with these macros:
741 #define pointer(T) typeof(T *)
742 #define array(T, N) typeof(T [N])
746 Now the declaration can be rewritten this way:
749 array (pointer (char), 4) y;
753 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
754 pointers to @code{char}.
757 In GNU C, but not GNU C++, you may also declare the type of a variable
758 as @code{__auto_type}. In that case, the declaration must declare
759 only one variable, whose declarator must just be an identifier, the
760 declaration must be initialized, and the type of the variable is
761 determined by the initializer; the name of the variable is not in
762 scope until after the initializer. (In C++, you should use C++11
763 @code{auto} for this purpose.) Using @code{__auto_type}, the
764 ``maximum'' macro above could be written as:
768 (@{ __auto_type _a = (a); \
769 __auto_type _b = (b); \
770 _a > _b ? _a : _b; @})
773 Using @code{__auto_type} instead of @code{typeof} has two advantages:
776 @item Each argument to the macro appears only once in the expansion of
777 the macro. This prevents the size of the macro expansion growing
778 exponentially when calls to such macros are nested inside arguments of
781 @item If the argument to the macro has variably modified type, it is
782 evaluated only once when using @code{__auto_type}, but twice if
783 @code{typeof} is used.
786 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
787 a more limited extension that permitted one to write
790 typedef @var{T} = @var{expr};
794 with the effect of declaring @var{T} to have the type of the expression
795 @var{expr}. This extension does not work with GCC 3 (versions between
796 3.0 and 3.2 crash; 3.2.1 and later give an error). Code that
797 relies on it should be rewritten to use @code{typeof}:
800 typedef typeof(@var{expr}) @var{T};
804 This works with all versions of GCC@.
807 @section Conditionals with Omitted Operands
808 @cindex conditional expressions, extensions
809 @cindex omitted middle-operands
810 @cindex middle-operands, omitted
811 @cindex extensions, @code{?:}
812 @cindex @code{?:} extensions
814 The middle operand in a conditional expression may be omitted. Then
815 if the first operand is nonzero, its value is the value of the conditional
818 Therefore, the expression
825 has the value of @code{x} if that is nonzero; otherwise, the value of
828 This example is perfectly equivalent to
834 @cindex side effect in @code{?:}
835 @cindex @code{?:} side effect
837 In this simple case, the ability to omit the middle operand is not
838 especially useful. When it becomes useful is when the first operand does,
839 or may (if it is a macro argument), contain a side effect. Then repeating
840 the operand in the middle would perform the side effect twice. Omitting
841 the middle operand uses the value already computed without the undesirable
842 effects of recomputing it.
845 @section 128-bit integers
846 @cindex @code{__int128} data types
848 As an extension the integer scalar type @code{__int128} is supported for
849 targets which have an integer mode wide enough to hold 128 bits.
850 Simply write @code{__int128} for a signed 128-bit integer, or
851 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
852 support in GCC for expressing an integer constant of type @code{__int128}
853 for targets with @code{long long} integer less than 128 bits wide.
856 @section Double-Word Integers
857 @cindex @code{long long} data types
858 @cindex double-word arithmetic
859 @cindex multiprecision arithmetic
860 @cindex @code{LL} integer suffix
861 @cindex @code{ULL} integer suffix
863 ISO C99 supports data types for integers that are at least 64 bits wide,
864 and as an extension GCC supports them in C90 mode and in C++.
865 Simply write @code{long long int} for a signed integer, or
866 @code{unsigned long long int} for an unsigned integer. To make an
867 integer constant of type @code{long long int}, add the suffix @samp{LL}
868 to the integer. To make an integer constant of type @code{unsigned long
869 long int}, add the suffix @samp{ULL} to the integer.
871 You can use these types in arithmetic like any other integer types.
872 Addition, subtraction, and bitwise boolean operations on these types
873 are open-coded on all types of machines. Multiplication is open-coded
874 if the machine supports a fullword-to-doubleword widening multiply
875 instruction. Division and shifts are open-coded only on machines that
876 provide special support. The operations that are not open-coded use
877 special library routines that come with GCC@.
879 There may be pitfalls when you use @code{long long} types for function
880 arguments without function prototypes. If a function
881 expects type @code{int} for its argument, and you pass a value of type
882 @code{long long int}, confusion results because the caller and the
883 subroutine disagree about the number of bytes for the argument.
884 Likewise, if the function expects @code{long long int} and you pass
885 @code{int}. The best way to avoid such problems is to use prototypes.
888 @section Complex Numbers
889 @cindex complex numbers
890 @cindex @code{_Complex} keyword
891 @cindex @code{__complex__} keyword
893 ISO C99 supports complex floating data types, and as an extension GCC
894 supports them in C90 mode and in C++. GCC also supports complex integer data
895 types which are not part of ISO C99. You can declare complex types
896 using the keyword @code{_Complex}. As an extension, the older GNU
897 keyword @code{__complex__} is also supported.
899 For example, @samp{_Complex double x;} declares @code{x} as a
900 variable whose real part and imaginary part are both of type
901 @code{double}. @samp{_Complex short int y;} declares @code{y} to
902 have real and imaginary parts of type @code{short int}; this is not
903 likely to be useful, but it shows that the set of complex types is
906 To write a constant with a complex data type, use the suffix @samp{i} or
907 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
908 has type @code{_Complex float} and @code{3i} has type
909 @code{_Complex int}. Such a constant always has a pure imaginary
910 value, but you can form any complex value you like by adding one to a
911 real constant. This is a GNU extension; if you have an ISO C99
912 conforming C library (such as the GNU C Library), and want to construct complex
913 constants of floating type, you should include @code{<complex.h>} and
914 use the macros @code{I} or @code{_Complex_I} instead.
916 @cindex @code{__real__} keyword
917 @cindex @code{__imag__} keyword
918 To extract the real part of a complex-valued expression @var{exp}, write
919 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
920 extract the imaginary part. This is a GNU extension; for values of
921 floating type, you should use the ISO C99 functions @code{crealf},
922 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
923 @code{cimagl}, declared in @code{<complex.h>} and also provided as
924 built-in functions by GCC@.
926 @cindex complex conjugation
927 The operator @samp{~} performs complex conjugation when used on a value
928 with a complex type. This is a GNU extension; for values of
929 floating type, you should use the ISO C99 functions @code{conjf},
930 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
931 provided as built-in functions by GCC@.
933 GCC can allocate complex automatic variables in a noncontiguous
934 fashion; it's even possible for the real part to be in a register while
935 the imaginary part is on the stack (or vice versa). Only the DWARF 2
936 debug info format can represent this, so use of DWARF 2 is recommended.
937 If you are using the stabs debug info format, GCC describes a noncontiguous
938 complex variable as if it were two separate variables of noncomplex type.
939 If the variable's actual name is @code{foo}, the two fictitious
940 variables are named @code{foo$real} and @code{foo$imag}. You can
941 examine and set these two fictitious variables with your debugger.
944 @section Additional Floating Types
945 @cindex additional floating types
946 @cindex @code{__float80} data type
947 @cindex @code{__float128} data type
948 @cindex @code{w} floating point suffix
949 @cindex @code{q} floating point suffix
950 @cindex @code{W} floating point suffix
951 @cindex @code{Q} floating point suffix
953 As an extension, GNU C supports additional floating
954 types, @code{__float80} and @code{__float128} to support 80-bit
955 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
956 Support for additional types includes the arithmetic operators:
957 add, subtract, multiply, divide; unary arithmetic operators;
958 relational operators; equality operators; and conversions to and from
959 integer and other floating types. Use a suffix @samp{w} or @samp{W}
960 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
961 for @code{_float128}. You can declare complex types using the
962 corresponding internal complex type, @code{XCmode} for @code{__float80}
963 type and @code{TCmode} for @code{__float128} type:
966 typedef _Complex float __attribute__((mode(TC))) _Complex128;
967 typedef _Complex float __attribute__((mode(XC))) _Complex80;
970 Not all targets support additional floating-point types. @code{__float80}
971 and @code{__float128} types are supported on i386, x86_64 and IA-64 targets.
972 The @code{__float128} type is supported on hppa HP-UX targets.
975 @section Half-Precision Floating Point
976 @cindex half-precision floating point
977 @cindex @code{__fp16} data type
979 On ARM targets, GCC supports half-precision (16-bit) floating point via
980 the @code{__fp16} type. You must enable this type explicitly
981 with the @option{-mfp16-format} command-line option in order to use it.
983 ARM supports two incompatible representations for half-precision
984 floating-point values. You must choose one of the representations and
985 use it consistently in your program.
987 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
988 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
989 There are 11 bits of significand precision, approximately 3
992 Specifying @option{-mfp16-format=alternative} selects the ARM
993 alternative format. This representation is similar to the IEEE
994 format, but does not support infinities or NaNs. Instead, the range
995 of exponents is extended, so that this format can represent normalized
996 values in the range of @math{2^{-14}} to 131008.
998 The @code{__fp16} type is a storage format only. For purposes
999 of arithmetic and other operations, @code{__fp16} values in C or C++
1000 expressions are automatically promoted to @code{float}. In addition,
1001 you cannot declare a function with a return value or parameters
1002 of type @code{__fp16}.
1004 Note that conversions from @code{double} to @code{__fp16}
1005 involve an intermediate conversion to @code{float}. Because
1006 of rounding, this can sometimes produce a different result than a
1009 ARM provides hardware support for conversions between
1010 @code{__fp16} and @code{float} values
1011 as an extension to VFP and NEON (Advanced SIMD). GCC generates
1012 code using these hardware instructions if you compile with
1013 options to select an FPU that provides them;
1014 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1015 in addition to the @option{-mfp16-format} option to select
1016 a half-precision format.
1018 Language-level support for the @code{__fp16} data type is
1019 independent of whether GCC generates code using hardware floating-point
1020 instructions. In cases where hardware support is not specified, GCC
1021 implements conversions between @code{__fp16} and @code{float} values
1025 @section Decimal Floating Types
1026 @cindex decimal floating types
1027 @cindex @code{_Decimal32} data type
1028 @cindex @code{_Decimal64} data type
1029 @cindex @code{_Decimal128} data type
1030 @cindex @code{df} integer suffix
1031 @cindex @code{dd} integer suffix
1032 @cindex @code{dl} integer suffix
1033 @cindex @code{DF} integer suffix
1034 @cindex @code{DD} integer suffix
1035 @cindex @code{DL} integer suffix
1037 As an extension, GNU C supports decimal floating types as
1038 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1039 floating types in GCC will evolve as the draft technical report changes.
1040 Calling conventions for any target might also change. Not all targets
1041 support decimal floating types.
1043 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1044 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1045 @code{float}, @code{double}, and @code{long double} whose radix is not
1046 specified by the C standard but is usually two.
1048 Support for decimal floating types includes the arithmetic operators
1049 add, subtract, multiply, divide; unary arithmetic operators;
1050 relational operators; equality operators; and conversions to and from
1051 integer and other floating types. Use a suffix @samp{df} or
1052 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1053 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1056 GCC support of decimal float as specified by the draft technical report
1061 When the value of a decimal floating type cannot be represented in the
1062 integer type to which it is being converted, the result is undefined
1063 rather than the result value specified by the draft technical report.
1066 GCC does not provide the C library functionality associated with
1067 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1068 @file{wchar.h}, which must come from a separate C library implementation.
1069 Because of this the GNU C compiler does not define macro
1070 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1071 the technical report.
1074 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1075 are supported by the DWARF 2 debug information format.
1081 ISO C99 supports floating-point numbers written not only in the usual
1082 decimal notation, such as @code{1.55e1}, but also numbers such as
1083 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1084 supports this in C90 mode (except in some cases when strictly
1085 conforming) and in C++. In that format the
1086 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1087 mandatory. The exponent is a decimal number that indicates the power of
1088 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1095 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1096 is the same as @code{1.55e1}.
1098 Unlike for floating-point numbers in the decimal notation the exponent
1099 is always required in the hexadecimal notation. Otherwise the compiler
1100 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1101 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1102 extension for floating-point constants of type @code{float}.
1105 @section Fixed-Point Types
1106 @cindex fixed-point types
1107 @cindex @code{_Fract} data type
1108 @cindex @code{_Accum} data type
1109 @cindex @code{_Sat} data type
1110 @cindex @code{hr} fixed-suffix
1111 @cindex @code{r} fixed-suffix
1112 @cindex @code{lr} fixed-suffix
1113 @cindex @code{llr} fixed-suffix
1114 @cindex @code{uhr} fixed-suffix
1115 @cindex @code{ur} fixed-suffix
1116 @cindex @code{ulr} fixed-suffix
1117 @cindex @code{ullr} fixed-suffix
1118 @cindex @code{hk} fixed-suffix
1119 @cindex @code{k} fixed-suffix
1120 @cindex @code{lk} fixed-suffix
1121 @cindex @code{llk} fixed-suffix
1122 @cindex @code{uhk} fixed-suffix
1123 @cindex @code{uk} fixed-suffix
1124 @cindex @code{ulk} fixed-suffix
1125 @cindex @code{ullk} fixed-suffix
1126 @cindex @code{HR} fixed-suffix
1127 @cindex @code{R} fixed-suffix
1128 @cindex @code{LR} fixed-suffix
1129 @cindex @code{LLR} fixed-suffix
1130 @cindex @code{UHR} fixed-suffix
1131 @cindex @code{UR} fixed-suffix
1132 @cindex @code{ULR} fixed-suffix
1133 @cindex @code{ULLR} fixed-suffix
1134 @cindex @code{HK} fixed-suffix
1135 @cindex @code{K} fixed-suffix
1136 @cindex @code{LK} fixed-suffix
1137 @cindex @code{LLK} fixed-suffix
1138 @cindex @code{UHK} fixed-suffix
1139 @cindex @code{UK} fixed-suffix
1140 @cindex @code{ULK} fixed-suffix
1141 @cindex @code{ULLK} fixed-suffix
1143 As an extension, GNU C supports fixed-point types as
1144 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1145 types in GCC will evolve as the draft technical report changes.
1146 Calling conventions for any target might also change. Not all targets
1147 support fixed-point types.
1149 The fixed-point types are
1150 @code{short _Fract},
1153 @code{long long _Fract},
1154 @code{unsigned short _Fract},
1155 @code{unsigned _Fract},
1156 @code{unsigned long _Fract},
1157 @code{unsigned long long _Fract},
1158 @code{_Sat short _Fract},
1160 @code{_Sat long _Fract},
1161 @code{_Sat long long _Fract},
1162 @code{_Sat unsigned short _Fract},
1163 @code{_Sat unsigned _Fract},
1164 @code{_Sat unsigned long _Fract},
1165 @code{_Sat unsigned long long _Fract},
1166 @code{short _Accum},
1169 @code{long long _Accum},
1170 @code{unsigned short _Accum},
1171 @code{unsigned _Accum},
1172 @code{unsigned long _Accum},
1173 @code{unsigned long long _Accum},
1174 @code{_Sat short _Accum},
1176 @code{_Sat long _Accum},
1177 @code{_Sat long long _Accum},
1178 @code{_Sat unsigned short _Accum},
1179 @code{_Sat unsigned _Accum},
1180 @code{_Sat unsigned long _Accum},
1181 @code{_Sat unsigned long long _Accum}.
1183 Fixed-point data values contain fractional and optional integral parts.
1184 The format of fixed-point data varies and depends on the target machine.
1186 Support for fixed-point types includes:
1189 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1191 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1193 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1195 binary shift operators (@code{<<}, @code{>>})
1197 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1199 equality operators (@code{==}, @code{!=})
1201 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1202 @code{<<=}, @code{>>=})
1204 conversions to and from integer, floating-point, or fixed-point types
1207 Use a suffix in a fixed-point literal constant:
1209 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1210 @code{_Sat short _Fract}
1211 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1212 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1213 @code{_Sat long _Fract}
1214 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1215 @code{_Sat long long _Fract}
1216 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1217 @code{_Sat unsigned short _Fract}
1218 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1219 @code{_Sat unsigned _Fract}
1220 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1221 @code{_Sat unsigned long _Fract}
1222 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1223 and @code{_Sat unsigned long long _Fract}
1224 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1225 @code{_Sat short _Accum}
1226 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1227 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1228 @code{_Sat long _Accum}
1229 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1230 @code{_Sat long long _Accum}
1231 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1232 @code{_Sat unsigned short _Accum}
1233 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1234 @code{_Sat unsigned _Accum}
1235 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1236 @code{_Sat unsigned long _Accum}
1237 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1238 and @code{_Sat unsigned long long _Accum}
1241 GCC support of fixed-point types as specified by the draft technical report
1246 Pragmas to control overflow and rounding behaviors are not implemented.
1249 Fixed-point types are supported by the DWARF 2 debug information format.
1251 @node Named Address Spaces
1252 @section Named Address Spaces
1253 @cindex Named Address Spaces
1255 As an extension, GNU C supports named address spaces as
1256 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1257 address spaces in GCC will evolve as the draft technical report
1258 changes. Calling conventions for any target might also change. At
1259 present, only the AVR, SPU, M32C, and RL78 targets support address
1260 spaces other than the generic address space.
1262 Address space identifiers may be used exactly like any other C type
1263 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1264 document for more details.
1266 @anchor{AVR Named Address Spaces}
1267 @subsection AVR Named Address Spaces
1269 On the AVR target, there are several address spaces that can be used
1270 in order to put read-only data into the flash memory and access that
1271 data by means of the special instructions @code{LPM} or @code{ELPM}
1272 needed to read from flash.
1274 Per default, any data including read-only data is located in RAM
1275 (the generic address space) so that non-generic address spaces are
1276 needed to locate read-only data in flash memory
1277 @emph{and} to generate the right instructions to access this data
1278 without using (inline) assembler code.
1282 @cindex @code{__flash} AVR Named Address Spaces
1283 The @code{__flash} qualifier locates data in the
1284 @code{.progmem.data} section. Data is read using the @code{LPM}
1285 instruction. Pointers to this address space are 16 bits wide.
1292 @cindex @code{__flash1} AVR Named Address Spaces
1293 @cindex @code{__flash2} AVR Named Address Spaces
1294 @cindex @code{__flash3} AVR Named Address Spaces
1295 @cindex @code{__flash4} AVR Named Address Spaces
1296 @cindex @code{__flash5} AVR Named Address Spaces
1297 These are 16-bit address spaces locating data in section
1298 @code{.progmem@var{N}.data} where @var{N} refers to
1299 address space @code{__flash@var{N}}.
1300 The compiler sets the @code{RAMPZ} segment register appropriately
1301 before reading data by means of the @code{ELPM} instruction.
1304 @cindex @code{__memx} AVR Named Address Spaces
1305 This is a 24-bit address space that linearizes flash and RAM:
1306 If the high bit of the address is set, data is read from
1307 RAM using the lower two bytes as RAM address.
1308 If the high bit of the address is clear, data is read from flash
1309 with @code{RAMPZ} set according to the high byte of the address.
1310 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1312 Objects in this address space are located in @code{.progmemx.data}.
1318 char my_read (const __flash char ** p)
1320 /* p is a pointer to RAM that points to a pointer to flash.
1321 The first indirection of p reads that flash pointer
1322 from RAM and the second indirection reads a char from this
1328 /* Locate array[] in flash memory */
1329 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1335 /* Return 17 by reading from flash memory */
1336 return array[array[i]];
1341 For each named address space supported by avr-gcc there is an equally
1342 named but uppercase built-in macro defined.
1343 The purpose is to facilitate testing if respective address space
1344 support is available or not:
1348 const __flash int var = 1;
1355 #include <avr/pgmspace.h> /* From AVR-LibC */
1357 const int var PROGMEM = 1;
1361 return (int) pgm_read_word (&var);
1363 #endif /* __FLASH */
1367 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1368 locates data in flash but
1369 accesses to these data read from generic address space, i.e.@:
1371 so that you need special accessors like @code{pgm_read_byte}
1372 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1373 together with attribute @code{progmem}.
1376 @b{Limitations and caveats}
1380 Reading across the 64@tie{}KiB section boundary of
1381 the @code{__flash} or @code{__flash@var{N}} address spaces
1382 shows undefined behavior. The only address space that
1383 supports reading across the 64@tie{}KiB flash segment boundaries is
1387 If you use one of the @code{__flash@var{N}} address spaces
1388 you must arrange your linker script to locate the
1389 @code{.progmem@var{N}.data} sections according to your needs.
1392 Any data or pointers to the non-generic address spaces must
1393 be qualified as @code{const}, i.e.@: as read-only data.
1394 This still applies if the data in one of these address
1395 spaces like software version number or calibration lookup table are intended to
1396 be changed after load time by, say, a boot loader. In this case
1397 the right qualification is @code{const} @code{volatile} so that the compiler
1398 must not optimize away known values or insert them
1399 as immediates into operands of instructions.
1402 The following code initializes a variable @code{pfoo}
1403 located in static storage with a 24-bit address:
1405 extern const __memx char foo;
1406 const __memx void *pfoo = &foo;
1410 Such code requires at least binutils 2.23, see
1411 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1415 @subsection M32C Named Address Spaces
1416 @cindex @code{__far} M32C Named Address Spaces
1418 On the M32C target, with the R8C and M16C CPU variants, variables
1419 qualified with @code{__far} are accessed using 32-bit addresses in
1420 order to access memory beyond the first 64@tie{}Ki bytes. If
1421 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1424 @subsection RL78 Named Address Spaces
1425 @cindex @code{__far} RL78 Named Address Spaces
1427 On the RL78 target, variables qualified with @code{__far} are accessed
1428 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1429 addresses. Non-far variables are assumed to appear in the topmost
1430 64@tie{}KiB of the address space.
1432 @subsection SPU Named Address Spaces
1433 @cindex @code{__ea} SPU Named Address Spaces
1435 On the SPU target variables may be declared as
1436 belonging to another address space by qualifying the type with the
1437 @code{__ea} address space identifier:
1444 The compiler generates special code to access the variable @code{i}.
1445 It may use runtime library
1446 support, or generate special machine instructions to access that address
1450 @section Arrays of Length Zero
1451 @cindex arrays of length zero
1452 @cindex zero-length arrays
1453 @cindex length-zero arrays
1454 @cindex flexible array members
1456 Zero-length arrays are allowed in GNU C@. They are very useful as the
1457 last element of a structure that is really a header for a variable-length
1466 struct line *thisline = (struct line *)
1467 malloc (sizeof (struct line) + this_length);
1468 thisline->length = this_length;
1471 In ISO C90, you would have to give @code{contents} a length of 1, which
1472 means either you waste space or complicate the argument to @code{malloc}.
1474 In ISO C99, you would use a @dfn{flexible array member}, which is
1475 slightly different in syntax and semantics:
1479 Flexible array members are written as @code{contents[]} without
1483 Flexible array members have incomplete type, and so the @code{sizeof}
1484 operator may not be applied. As a quirk of the original implementation
1485 of zero-length arrays, @code{sizeof} evaluates to zero.
1488 Flexible array members may only appear as the last member of a
1489 @code{struct} that is otherwise non-empty.
1492 A structure containing a flexible array member, or a union containing
1493 such a structure (possibly recursively), may not be a member of a
1494 structure or an element of an array. (However, these uses are
1495 permitted by GCC as extensions.)
1498 GCC versions before 3.0 allowed zero-length arrays to be statically
1499 initialized, as if they were flexible arrays. In addition to those
1500 cases that were useful, it also allowed initializations in situations
1501 that would corrupt later data. Non-empty initialization of zero-length
1502 arrays is now treated like any case where there are more initializer
1503 elements than the array holds, in that a suitable warning about ``excess
1504 elements in array'' is given, and the excess elements (all of them, in
1505 this case) are ignored.
1507 Instead GCC allows static initialization of flexible array members.
1508 This is equivalent to defining a new structure containing the original
1509 structure followed by an array of sufficient size to contain the data.
1510 E.g.@: in the following, @code{f1} is constructed as if it were declared
1516 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1519 struct f1 f1; int data[3];
1520 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1524 The convenience of this extension is that @code{f1} has the desired
1525 type, eliminating the need to consistently refer to @code{f2.f1}.
1527 This has symmetry with normal static arrays, in that an array of
1528 unknown size is also written with @code{[]}.
1530 Of course, this extension only makes sense if the extra data comes at
1531 the end of a top-level object, as otherwise we would be overwriting
1532 data at subsequent offsets. To avoid undue complication and confusion
1533 with initialization of deeply nested arrays, we simply disallow any
1534 non-empty initialization except when the structure is the top-level
1535 object. For example:
1538 struct foo @{ int x; int y[]; @};
1539 struct bar @{ struct foo z; @};
1541 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1542 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1543 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1544 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1547 @node Empty Structures
1548 @section Structures With No Members
1549 @cindex empty structures
1550 @cindex zero-size structures
1552 GCC permits a C structure to have no members:
1559 The structure has size zero. In C++, empty structures are part
1560 of the language. G++ treats empty structures as if they had a single
1561 member of type @code{char}.
1563 @node Variable Length
1564 @section Arrays of Variable Length
1565 @cindex variable-length arrays
1566 @cindex arrays of variable length
1569 Variable-length automatic arrays are allowed in ISO C99, and as an
1570 extension GCC accepts them in C90 mode and in C++. These arrays are
1571 declared like any other automatic arrays, but with a length that is not
1572 a constant expression. The storage is allocated at the point of
1573 declaration and deallocated when the block scope containing the declaration
1579 concat_fopen (char *s1, char *s2, char *mode)
1581 char str[strlen (s1) + strlen (s2) + 1];
1584 return fopen (str, mode);
1588 @cindex scope of a variable length array
1589 @cindex variable-length array scope
1590 @cindex deallocating variable length arrays
1591 Jumping or breaking out of the scope of the array name deallocates the
1592 storage. Jumping into the scope is not allowed; you get an error
1595 @cindex @code{alloca} vs variable-length arrays
1596 You can use the function @code{alloca} to get an effect much like
1597 variable-length arrays. The function @code{alloca} is available in
1598 many other C implementations (but not in all). On the other hand,
1599 variable-length arrays are more elegant.
1601 There are other differences between these two methods. Space allocated
1602 with @code{alloca} exists until the containing @emph{function} returns.
1603 The space for a variable-length array is deallocated as soon as the array
1604 name's scope ends. (If you use both variable-length arrays and
1605 @code{alloca} in the same function, deallocation of a variable-length array
1606 also deallocates anything more recently allocated with @code{alloca}.)
1608 You can also use variable-length arrays as arguments to functions:
1612 tester (int len, char data[len][len])
1618 The length of an array is computed once when the storage is allocated
1619 and is remembered for the scope of the array in case you access it with
1622 If you want to pass the array first and the length afterward, you can
1623 use a forward declaration in the parameter list---another GNU extension.
1627 tester (int len; char data[len][len], int len)
1633 @cindex parameter forward declaration
1634 The @samp{int len} before the semicolon is a @dfn{parameter forward
1635 declaration}, and it serves the purpose of making the name @code{len}
1636 known when the declaration of @code{data} is parsed.
1638 You can write any number of such parameter forward declarations in the
1639 parameter list. They can be separated by commas or semicolons, but the
1640 last one must end with a semicolon, which is followed by the ``real''
1641 parameter declarations. Each forward declaration must match a ``real''
1642 declaration in parameter name and data type. ISO C99 does not support
1643 parameter forward declarations.
1645 @node Variadic Macros
1646 @section Macros with a Variable Number of Arguments.
1647 @cindex variable number of arguments
1648 @cindex macro with variable arguments
1649 @cindex rest argument (in macro)
1650 @cindex variadic macros
1652 In the ISO C standard of 1999, a macro can be declared to accept a
1653 variable number of arguments much as a function can. The syntax for
1654 defining the macro is similar to that of a function. Here is an
1658 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1662 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1663 such a macro, it represents the zero or more tokens until the closing
1664 parenthesis that ends the invocation, including any commas. This set of
1665 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1666 wherever it appears. See the CPP manual for more information.
1668 GCC has long supported variadic macros, and used a different syntax that
1669 allowed you to give a name to the variable arguments just like any other
1670 argument. Here is an example:
1673 #define debug(format, args...) fprintf (stderr, format, args)
1677 This is in all ways equivalent to the ISO C example above, but arguably
1678 more readable and descriptive.
1680 GNU CPP has two further variadic macro extensions, and permits them to
1681 be used with either of the above forms of macro definition.
1683 In standard C, you are not allowed to leave the variable argument out
1684 entirely; but you are allowed to pass an empty argument. For example,
1685 this invocation is invalid in ISO C, because there is no comma after
1692 GNU CPP permits you to completely omit the variable arguments in this
1693 way. In the above examples, the compiler would complain, though since
1694 the expansion of the macro still has the extra comma after the format
1697 To help solve this problem, CPP behaves specially for variable arguments
1698 used with the token paste operator, @samp{##}. If instead you write
1701 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1705 and if the variable arguments are omitted or empty, the @samp{##}
1706 operator causes the preprocessor to remove the comma before it. If you
1707 do provide some variable arguments in your macro invocation, GNU CPP
1708 does not complain about the paste operation and instead places the
1709 variable arguments after the comma. Just like any other pasted macro
1710 argument, these arguments are not macro expanded.
1712 @node Escaped Newlines
1713 @section Slightly Looser Rules for Escaped Newlines
1714 @cindex escaped newlines
1715 @cindex newlines (escaped)
1717 Recently, the preprocessor has relaxed its treatment of escaped
1718 newlines. Previously, the newline had to immediately follow a
1719 backslash. The current implementation allows whitespace in the form
1720 of spaces, horizontal and vertical tabs, and form feeds between the
1721 backslash and the subsequent newline. The preprocessor issues a
1722 warning, but treats it as a valid escaped newline and combines the two
1723 lines to form a single logical line. This works within comments and
1724 tokens, as well as between tokens. Comments are @emph{not} treated as
1725 whitespace for the purposes of this relaxation, since they have not
1726 yet been replaced with spaces.
1729 @section Non-Lvalue Arrays May Have Subscripts
1730 @cindex subscripting
1731 @cindex arrays, non-lvalue
1733 @cindex subscripting and function values
1734 In ISO C99, arrays that are not lvalues still decay to pointers, and
1735 may be subscripted, although they may not be modified or used after
1736 the next sequence point and the unary @samp{&} operator may not be
1737 applied to them. As an extension, GNU C allows such arrays to be
1738 subscripted in C90 mode, though otherwise they do not decay to
1739 pointers outside C99 mode. For example,
1740 this is valid in GNU C though not valid in C90:
1744 struct foo @{int a[4];@};
1750 return f().a[index];
1756 @section Arithmetic on @code{void}- and Function-Pointers
1757 @cindex void pointers, arithmetic
1758 @cindex void, size of pointer to
1759 @cindex function pointers, arithmetic
1760 @cindex function, size of pointer to
1762 In GNU C, addition and subtraction operations are supported on pointers to
1763 @code{void} and on pointers to functions. This is done by treating the
1764 size of a @code{void} or of a function as 1.
1766 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1767 and on function types, and returns 1.
1769 @opindex Wpointer-arith
1770 The option @option{-Wpointer-arith} requests a warning if these extensions
1774 @section Non-Constant Initializers
1775 @cindex initializers, non-constant
1776 @cindex non-constant initializers
1778 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1779 automatic variable are not required to be constant expressions in GNU C@.
1780 Here is an example of an initializer with run-time varying elements:
1783 foo (float f, float g)
1785 float beat_freqs[2] = @{ f-g, f+g @};
1790 @node Compound Literals
1791 @section Compound Literals
1792 @cindex constructor expressions
1793 @cindex initializations in expressions
1794 @cindex structures, constructor expression
1795 @cindex expressions, constructor
1796 @cindex compound literals
1797 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1799 ISO C99 supports compound literals. A compound literal looks like
1800 a cast containing an initializer. Its value is an object of the
1801 type specified in the cast, containing the elements specified in
1802 the initializer; it is an lvalue. As an extension, GCC supports
1803 compound literals in C90 mode and in C++, though the semantics are
1804 somewhat different in C++.
1806 Usually, the specified type is a structure. Assume that
1807 @code{struct foo} and @code{structure} are declared as shown:
1810 struct foo @{int a; char b[2];@} structure;
1814 Here is an example of constructing a @code{struct foo} with a compound literal:
1817 structure = ((struct foo) @{x + y, 'a', 0@});
1821 This is equivalent to writing the following:
1825 struct foo temp = @{x + y, 'a', 0@};
1830 You can also construct an array, though this is dangerous in C++, as
1831 explained below. If all the elements of the compound literal are
1832 (made up of) simple constant expressions, suitable for use in
1833 initializers of objects of static storage duration, then the compound
1834 literal can be coerced to a pointer to its first element and used in
1835 such an initializer, as shown here:
1838 char **foo = (char *[]) @{ "x", "y", "z" @};
1841 Compound literals for scalar types and union types are
1842 also allowed, but then the compound literal is equivalent
1845 As a GNU extension, GCC allows initialization of objects with static storage
1846 duration by compound literals (which is not possible in ISO C99, because
1847 the initializer is not a constant).
1848 It is handled as if the object is initialized only with the bracket
1849 enclosed list if the types of the compound literal and the object match.
1850 The initializer list of the compound literal must be constant.
1851 If the object being initialized has array type of unknown size, the size is
1852 determined by compound literal size.
1855 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1856 static int y[] = (int []) @{1, 2, 3@};
1857 static int z[] = (int [3]) @{1@};
1861 The above lines are equivalent to the following:
1863 static struct foo x = @{1, 'a', 'b'@};
1864 static int y[] = @{1, 2, 3@};
1865 static int z[] = @{1, 0, 0@};
1868 In C, a compound literal designates an unnamed object with static or
1869 automatic storage duration. In C++, a compound literal designates a
1870 temporary object, which only lives until the end of its
1871 full-expression. As a result, well-defined C code that takes the
1872 address of a subobject of a compound literal can be undefined in C++.
1873 For instance, if the array compound literal example above appeared
1874 inside a function, any subsequent use of @samp{foo} in C++ has
1875 undefined behavior because the lifetime of the array ends after the
1876 declaration of @samp{foo}. As a result, the C++ compiler now rejects
1877 the conversion of a temporary array to a pointer.
1879 As an optimization, the C++ compiler sometimes gives array compound
1880 literals longer lifetimes: when the array either appears outside a
1881 function or has const-qualified type. If @samp{foo} and its
1882 initializer had elements of @samp{char *const} type rather than
1883 @samp{char *}, or if @samp{foo} were a global variable, the array
1884 would have static storage duration. But it is probably safest just to
1885 avoid the use of array compound literals in code compiled as C++.
1887 @node Designated Inits
1888 @section Designated Initializers
1889 @cindex initializers with labeled elements
1890 @cindex labeled elements in initializers
1891 @cindex case labels in initializers
1892 @cindex designated initializers
1894 Standard C90 requires the elements of an initializer to appear in a fixed
1895 order, the same as the order of the elements in the array or structure
1898 In ISO C99 you can give the elements in any order, specifying the array
1899 indices or structure field names they apply to, and GNU C allows this as
1900 an extension in C90 mode as well. This extension is not
1901 implemented in GNU C++.
1903 To specify an array index, write
1904 @samp{[@var{index}] =} before the element value. For example,
1907 int a[6] = @{ [4] = 29, [2] = 15 @};
1914 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1918 The index values must be constant expressions, even if the array being
1919 initialized is automatic.
1921 An alternative syntax for this that has been obsolete since GCC 2.5 but
1922 GCC still accepts is to write @samp{[@var{index}]} before the element
1923 value, with no @samp{=}.
1925 To initialize a range of elements to the same value, write
1926 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1927 extension. For example,
1930 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1934 If the value in it has side-effects, the side-effects happen only once,
1935 not for each initialized field by the range initializer.
1938 Note that the length of the array is the highest value specified
1941 In a structure initializer, specify the name of a field to initialize
1942 with @samp{.@var{fieldname} =} before the element value. For example,
1943 given the following structure,
1946 struct point @{ int x, y; @};
1950 the following initialization
1953 struct point p = @{ .y = yvalue, .x = xvalue @};
1960 struct point p = @{ xvalue, yvalue @};
1963 Another syntax that has the same meaning, obsolete since GCC 2.5, is
1964 @samp{@var{fieldname}:}, as shown here:
1967 struct point p = @{ y: yvalue, x: xvalue @};
1971 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1972 @dfn{designator}. You can also use a designator (or the obsolete colon
1973 syntax) when initializing a union, to specify which element of the union
1974 should be used. For example,
1977 union foo @{ int i; double d; @};
1979 union foo f = @{ .d = 4 @};
1983 converts 4 to a @code{double} to store it in the union using
1984 the second element. By contrast, casting 4 to type @code{union foo}
1985 stores it into the union as the integer @code{i}, since it is
1986 an integer. (@xref{Cast to Union}.)
1988 You can combine this technique of naming elements with ordinary C
1989 initialization of successive elements. Each initializer element that
1990 does not have a designator applies to the next consecutive element of the
1991 array or structure. For example,
1994 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2001 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2004 Labeling the elements of an array initializer is especially useful
2005 when the indices are characters or belong to an @code{enum} type.
2010 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2011 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2014 @cindex designator lists
2015 You can also write a series of @samp{.@var{fieldname}} and
2016 @samp{[@var{index}]} designators before an @samp{=} to specify a
2017 nested subobject to initialize; the list is taken relative to the
2018 subobject corresponding to the closest surrounding brace pair. For
2019 example, with the @samp{struct point} declaration above:
2022 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2026 If the same field is initialized multiple times, it has the value from
2027 the last initialization. If any such overridden initialization has
2028 side-effect, it is unspecified whether the side-effect happens or not.
2029 Currently, GCC discards them and issues a warning.
2032 @section Case Ranges
2034 @cindex ranges in case statements
2036 You can specify a range of consecutive values in a single @code{case} label,
2040 case @var{low} ... @var{high}:
2044 This has the same effect as the proper number of individual @code{case}
2045 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2047 This feature is especially useful for ranges of ASCII character codes:
2053 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2054 it may be parsed wrong when you use it with integer values. For example,
2069 @section Cast to a Union Type
2070 @cindex cast to a union
2071 @cindex union, casting to a
2073 A cast to union type is similar to other casts, except that the type
2074 specified is a union type. You can specify the type either with
2075 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2076 a constructor, not a cast, and hence does not yield an lvalue like
2077 normal casts. (@xref{Compound Literals}.)
2079 The types that may be cast to the union type are those of the members
2080 of the union. Thus, given the following union and variables:
2083 union foo @{ int i; double d; @};
2089 both @code{x} and @code{y} can be cast to type @code{union foo}.
2091 Using the cast as the right-hand side of an assignment to a variable of
2092 union type is equivalent to storing in a member of the union:
2097 u = (union foo) x @equiv{} u.i = x
2098 u = (union foo) y @equiv{} u.d = y
2101 You can also use the union cast as a function argument:
2104 void hack (union foo);
2106 hack ((union foo) x);
2109 @node Mixed Declarations
2110 @section Mixed Declarations and Code
2111 @cindex mixed declarations and code
2112 @cindex declarations, mixed with code
2113 @cindex code, mixed with declarations
2115 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2116 within compound statements. As an extension, GNU C also allows this in
2117 C90 mode. For example, you could do:
2126 Each identifier is visible from where it is declared until the end of
2127 the enclosing block.
2129 @node Function Attributes
2130 @section Declaring Attributes of Functions
2131 @cindex function attributes
2132 @cindex declaring attributes of functions
2133 @cindex functions that never return
2134 @cindex functions that return more than once
2135 @cindex functions that have no side effects
2136 @cindex functions in arbitrary sections
2137 @cindex functions that behave like malloc
2138 @cindex @code{volatile} applied to function
2139 @cindex @code{const} applied to function
2140 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2141 @cindex functions with non-null pointer arguments
2142 @cindex functions that are passed arguments in registers on the 386
2143 @cindex functions that pop the argument stack on the 386
2144 @cindex functions that do not pop the argument stack on the 386
2145 @cindex functions that have different compilation options on the 386
2146 @cindex functions that have different optimization options
2147 @cindex functions that are dynamically resolved
2149 In GNU C, you declare certain things about functions called in your program
2150 which help the compiler optimize function calls and check your code more
2153 The keyword @code{__attribute__} allows you to specify special
2154 attributes when making a declaration. This keyword is followed by an
2155 attribute specification inside double parentheses. The following
2156 attributes are currently defined for functions on all targets:
2157 @code{aligned}, @code{alloc_size}, @code{alloc_align}, @code{assume_aligned},
2158 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{noclone},
2159 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
2160 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
2161 @code{no_instrument_function}, @code{no_split_stack},
2162 @code{section}, @code{constructor},
2163 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
2164 @code{weak}, @code{malloc}, @code{alias}, @code{ifunc},
2165 @code{warn_unused_result}, @code{nonnull},
2166 @code{returns_nonnull}, @code{gnu_inline},
2167 @code{externally_visible}, @code{hot}, @code{cold}, @code{artificial},
2168 @code{no_sanitize_address}, @code{no_address_safety_analysis},
2169 @code{no_sanitize_undefined},
2170 @code{error} and @code{warning}.
2171 Several other attributes are defined for functions on particular
2172 target systems. Other attributes, including @code{section} are
2173 supported for variables declarations (@pxref{Variable Attributes})
2174 and for types (@pxref{Type Attributes}).
2176 GCC plugins may provide their own attributes.
2178 You may also specify attributes with @samp{__} preceding and following
2179 each keyword. This allows you to use them in header files without
2180 being concerned about a possible macro of the same name. For example,
2181 you may use @code{__noreturn__} instead of @code{noreturn}.
2183 @xref{Attribute Syntax}, for details of the exact syntax for using
2187 @c Keep this table alphabetized by attribute name. Treat _ as space.
2189 @item alias ("@var{target}")
2190 @cindex @code{alias} attribute
2191 The @code{alias} attribute causes the declaration to be emitted as an
2192 alias for another symbol, which must be specified. For instance,
2195 void __f () @{ /* @r{Do something.} */; @}
2196 void f () __attribute__ ((weak, alias ("__f")));
2200 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2201 mangled name for the target must be used. It is an error if @samp{__f}
2202 is not defined in the same translation unit.
2204 Not all target machines support this attribute.
2206 @item aligned (@var{alignment})
2207 @cindex @code{aligned} attribute
2208 This attribute specifies a minimum alignment for the function,
2211 You cannot use this attribute to decrease the alignment of a function,
2212 only to increase it. However, when you explicitly specify a function
2213 alignment this overrides the effect of the
2214 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2217 Note that the effectiveness of @code{aligned} attributes may be
2218 limited by inherent limitations in your linker. On many systems, the
2219 linker is only able to arrange for functions to be aligned up to a
2220 certain maximum alignment. (For some linkers, the maximum supported
2221 alignment may be very very small.) See your linker documentation for
2222 further information.
2224 The @code{aligned} attribute can also be used for variables and fields
2225 (@pxref{Variable Attributes}.)
2228 @cindex @code{alloc_size} attribute
2229 The @code{alloc_size} attribute is used to tell the compiler that the
2230 function return value points to memory, where the size is given by
2231 one or two of the functions parameters. GCC uses this
2232 information to improve the correctness of @code{__builtin_object_size}.
2234 The function parameter(s) denoting the allocated size are specified by
2235 one or two integer arguments supplied to the attribute. The allocated size
2236 is either the value of the single function argument specified or the product
2237 of the two function arguments specified. Argument numbering starts at
2243 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2244 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2248 declares that @code{my_calloc} returns memory of the size given by
2249 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2250 of the size given by parameter 2.
2253 @cindex @code{alloc_align} attribute
2254 The @code{alloc_align} attribute is used to tell the compiler that the
2255 function return value points to memory, where the returned pointer minimum
2256 alignment is given by one of the functions parameters. GCC uses this
2257 information to improve pointer alignment analysis.
2259 The function parameter denoting the allocated alignment is specified by
2260 one integer argument, whose number is the argument of the attribute.
2261 Argument numbering starts at one.
2266 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2270 declares that @code{my_memalign} returns memory with minimum alignment
2271 given by parameter 1.
2273 @item assume_aligned
2274 @cindex @code{assume_aligned} attribute
2275 The @code{assume_aligned} attribute is used to tell the compiler that the
2276 function return value points to memory, where the returned pointer minimum
2277 alignment is given by the first argument.
2278 If the attribute has two arguments, the second argument is misalignment offset.
2283 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2284 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2288 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2289 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2293 @cindex @code{always_inline} function attribute
2294 Generally, functions are not inlined unless optimization is specified.
2295 For functions declared inline, this attribute inlines the function even
2296 if no optimization level is specified.
2299 @cindex @code{gnu_inline} function attribute
2300 This attribute should be used with a function that is also declared
2301 with the @code{inline} keyword. It directs GCC to treat the function
2302 as if it were defined in gnu90 mode even when compiling in C99 or
2305 If the function is declared @code{extern}, then this definition of the
2306 function is used only for inlining. In no case is the function
2307 compiled as a standalone function, not even if you take its address
2308 explicitly. Such an address becomes an external reference, as if you
2309 had only declared the function, and had not defined it. This has
2310 almost the effect of a macro. The way to use this is to put a
2311 function definition in a header file with this attribute, and put
2312 another copy of the function, without @code{extern}, in a library
2313 file. The definition in the header file causes most calls to the
2314 function to be inlined. If any uses of the function remain, they
2315 refer to the single copy in the library. Note that the two
2316 definitions of the functions need not be precisely the same, although
2317 if they do not have the same effect your program may behave oddly.
2319 In C, if the function is neither @code{extern} nor @code{static}, then
2320 the function is compiled as a standalone function, as well as being
2321 inlined where possible.
2323 This is how GCC traditionally handled functions declared
2324 @code{inline}. Since ISO C99 specifies a different semantics for
2325 @code{inline}, this function attribute is provided as a transition
2326 measure and as a useful feature in its own right. This attribute is
2327 available in GCC 4.1.3 and later. It is available if either of the
2328 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2329 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2330 Function is As Fast As a Macro}.
2332 In C++, this attribute does not depend on @code{extern} in any way,
2333 but it still requires the @code{inline} keyword to enable its special
2337 @cindex @code{artificial} function attribute
2338 This attribute is useful for small inline wrappers that if possible
2339 should appear during debugging as a unit. Depending on the debug
2340 info format it either means marking the function as artificial
2341 or using the caller location for all instructions within the inlined
2345 @cindex interrupt handler functions
2346 When added to an interrupt handler with the M32C port, causes the
2347 prologue and epilogue to use bank switching to preserve the registers
2348 rather than saving them on the stack.
2351 @cindex @code{flatten} function attribute
2352 Generally, inlining into a function is limited. For a function marked with
2353 this attribute, every call inside this function is inlined, if possible.
2354 Whether the function itself is considered for inlining depends on its size and
2355 the current inlining parameters.
2357 @item error ("@var{message}")
2358 @cindex @code{error} function attribute
2359 If this attribute is used on a function declaration and a call to such a function
2360 is not eliminated through dead code elimination or other optimizations, an error
2361 that includes @var{message} is diagnosed. This is useful
2362 for compile-time checking, especially together with @code{__builtin_constant_p}
2363 and inline functions where checking the inline function arguments is not
2364 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2365 While it is possible to leave the function undefined and thus invoke
2366 a link failure, when using this attribute the problem is diagnosed
2367 earlier and with exact location of the call even in presence of inline
2368 functions or when not emitting debugging information.
2370 @item warning ("@var{message}")
2371 @cindex @code{warning} function attribute
2372 If this attribute is used on a function declaration and a call to such a function
2373 is not eliminated through dead code elimination or other optimizations, a warning
2374 that includes @var{message} is diagnosed. This is useful
2375 for compile-time checking, especially together with @code{__builtin_constant_p}
2376 and inline functions. While it is possible to define the function with
2377 a message in @code{.gnu.warning*} section, when using this attribute the problem
2378 is diagnosed earlier and with exact location of the call even in presence
2379 of inline functions or when not emitting debugging information.
2382 @cindex functions that do pop the argument stack on the 386
2384 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2385 assume that the calling function pops off the stack space used to
2386 pass arguments. This is
2387 useful to override the effects of the @option{-mrtd} switch.
2390 @cindex @code{const} function attribute
2391 Many functions do not examine any values except their arguments, and
2392 have no effects except the return value. Basically this is just slightly
2393 more strict class than the @code{pure} attribute below, since function is not
2394 allowed to read global memory.
2396 @cindex pointer arguments
2397 Note that a function that has pointer arguments and examines the data
2398 pointed to must @emph{not} be declared @code{const}. Likewise, a
2399 function that calls a non-@code{const} function usually must not be
2400 @code{const}. It does not make sense for a @code{const} function to
2403 The attribute @code{const} is not implemented in GCC versions earlier
2404 than 2.5. An alternative way to declare that a function has no side
2405 effects, which works in the current version and in some older versions,
2409 typedef int intfn ();
2411 extern const intfn square;
2415 This approach does not work in GNU C++ from 2.6.0 on, since the language
2416 specifies that the @samp{const} must be attached to the return value.
2420 @itemx constructor (@var{priority})
2421 @itemx destructor (@var{priority})
2422 @cindex @code{constructor} function attribute
2423 @cindex @code{destructor} function attribute
2424 The @code{constructor} attribute causes the function to be called
2425 automatically before execution enters @code{main ()}. Similarly, the
2426 @code{destructor} attribute causes the function to be called
2427 automatically after @code{main ()} completes or @code{exit ()} is
2428 called. Functions with these attributes are useful for
2429 initializing data that is used implicitly during the execution of
2432 You may provide an optional integer priority to control the order in
2433 which constructor and destructor functions are run. A constructor
2434 with a smaller priority number runs before a constructor with a larger
2435 priority number; the opposite relationship holds for destructors. So,
2436 if you have a constructor that allocates a resource and a destructor
2437 that deallocates the same resource, both functions typically have the
2438 same priority. The priorities for constructor and destructor
2439 functions are the same as those specified for namespace-scope C++
2440 objects (@pxref{C++ Attributes}).
2442 These attributes are not currently implemented for Objective-C@.
2445 @itemx deprecated (@var{msg})
2446 @cindex @code{deprecated} attribute.
2447 The @code{deprecated} attribute results in a warning if the function
2448 is used anywhere in the source file. This is useful when identifying
2449 functions that are expected to be removed in a future version of a
2450 program. The warning also includes the location of the declaration
2451 of the deprecated function, to enable users to easily find further
2452 information about why the function is deprecated, or what they should
2453 do instead. Note that the warnings only occurs for uses:
2456 int old_fn () __attribute__ ((deprecated));
2458 int (*fn_ptr)() = old_fn;
2462 results in a warning on line 3 but not line 2. The optional @var{msg}
2463 argument, which must be a string, is printed in the warning if
2466 The @code{deprecated} attribute can also be used for variables and
2467 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2470 @cindex @code{disinterrupt} attribute
2471 On Epiphany and MeP targets, this attribute causes the compiler to emit
2472 instructions to disable interrupts for the duration of the given
2476 @cindex @code{__declspec(dllexport)}
2477 On Microsoft Windows targets and Symbian OS targets the
2478 @code{dllexport} attribute causes the compiler to provide a global
2479 pointer to a pointer in a DLL, so that it can be referenced with the
2480 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2481 name is formed by combining @code{_imp__} and the function or variable
2484 You can use @code{__declspec(dllexport)} as a synonym for
2485 @code{__attribute__ ((dllexport))} for compatibility with other
2488 On systems that support the @code{visibility} attribute, this
2489 attribute also implies ``default'' visibility. It is an error to
2490 explicitly specify any other visibility.
2492 In previous versions of GCC, the @code{dllexport} attribute was ignored
2493 for inlined functions, unless the @option{-fkeep-inline-functions} flag
2494 had been used. The default behavior now is to emit all dllexported
2495 inline functions; however, this can cause object file-size bloat, in
2496 which case the old behavior can be restored by using
2497 @option{-fno-keep-inline-dllexport}.
2499 The attribute is also ignored for undefined symbols.
2501 When applied to C++ classes, the attribute marks defined non-inlined
2502 member functions and static data members as exports. Static consts
2503 initialized in-class are not marked unless they are also defined
2506 For Microsoft Windows targets there are alternative methods for
2507 including the symbol in the DLL's export table such as using a
2508 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2509 the @option{--export-all} linker flag.
2512 @cindex @code{__declspec(dllimport)}
2513 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2514 attribute causes the compiler to reference a function or variable via
2515 a global pointer to a pointer that is set up by the DLL exporting the
2516 symbol. The attribute implies @code{extern}. On Microsoft Windows
2517 targets, the pointer name is formed by combining @code{_imp__} and the
2518 function or variable name.
2520 You can use @code{__declspec(dllimport)} as a synonym for
2521 @code{__attribute__ ((dllimport))} for compatibility with other
2524 On systems that support the @code{visibility} attribute, this
2525 attribute also implies ``default'' visibility. It is an error to
2526 explicitly specify any other visibility.
2528 Currently, the attribute is ignored for inlined functions. If the
2529 attribute is applied to a symbol @emph{definition}, an error is reported.
2530 If a symbol previously declared @code{dllimport} is later defined, the
2531 attribute is ignored in subsequent references, and a warning is emitted.
2532 The attribute is also overridden by a subsequent declaration as
2535 When applied to C++ classes, the attribute marks non-inlined
2536 member functions and static data members as imports. However, the
2537 attribute is ignored for virtual methods to allow creation of vtables
2540 On the SH Symbian OS target the @code{dllimport} attribute also has
2541 another affect---it can cause the vtable and run-time type information
2542 for a class to be exported. This happens when the class has a
2543 dllimported constructor or a non-inline, non-pure virtual function
2544 and, for either of those two conditions, the class also has an inline
2545 constructor or destructor and has a key function that is defined in
2546 the current translation unit.
2548 For Microsoft Windows targets the use of the @code{dllimport}
2549 attribute on functions is not necessary, but provides a small
2550 performance benefit by eliminating a thunk in the DLL@. The use of the
2551 @code{dllimport} attribute on imported variables was required on older
2552 versions of the GNU linker, but can now be avoided by passing the
2553 @option{--enable-auto-import} switch to the GNU linker. As with
2554 functions, using the attribute for a variable eliminates a thunk in
2557 One drawback to using this attribute is that a pointer to a
2558 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2559 address. However, a pointer to a @emph{function} with the
2560 @code{dllimport} attribute can be used as a constant initializer; in
2561 this case, the address of a stub function in the import lib is
2562 referenced. On Microsoft Windows targets, the attribute can be disabled
2563 for functions by setting the @option{-mnop-fun-dllimport} flag.
2566 @cindex eight-bit data on the H8/300, H8/300H, and H8S
2567 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2568 variable should be placed into the eight-bit data section.
2569 The compiler generates more efficient code for certain operations
2570 on data in the eight-bit data area. Note the eight-bit data area is limited to
2573 You must use GAS and GLD from GNU binutils version 2.7 or later for
2574 this attribute to work correctly.
2577 @cindex exception handler functions
2578 Use this attribute on the NDS32 target to indicate that the specified function
2579 is an exception handler. The compiler will generate corresponding sections
2580 for use in an exception handler.
2582 @item exception_handler
2583 @cindex exception handler functions on the Blackfin processor
2584 Use this attribute on the Blackfin to indicate that the specified function
2585 is an exception handler. The compiler generates function entry and
2586 exit sequences suitable for use in an exception handler when this
2587 attribute is present.
2589 @item externally_visible
2590 @cindex @code{externally_visible} attribute.
2591 This attribute, attached to a global variable or function, nullifies
2592 the effect of the @option{-fwhole-program} command-line option, so the
2593 object remains visible outside the current compilation unit.
2595 If @option{-fwhole-program} is used together with @option{-flto} and
2596 @command{gold} is used as the linker plugin,
2597 @code{externally_visible} attributes are automatically added to functions
2598 (not variable yet due to a current @command{gold} issue)
2599 that are accessed outside of LTO objects according to resolution file
2600 produced by @command{gold}.
2601 For other linkers that cannot generate resolution file,
2602 explicit @code{externally_visible} attributes are still necessary.
2605 @cindex functions that handle memory bank switching
2606 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2607 use a calling convention that takes care of switching memory banks when
2608 entering and leaving a function. This calling convention is also the
2609 default when using the @option{-mlong-calls} option.
2611 On 68HC12 the compiler uses the @code{call} and @code{rtc} instructions
2612 to call and return from a function.
2614 On 68HC11 the compiler generates a sequence of instructions
2615 to invoke a board-specific routine to switch the memory bank and call the
2616 real function. The board-specific routine simulates a @code{call}.
2617 At the end of a function, it jumps to a board-specific routine
2618 instead of using @code{rts}. The board-specific return routine simulates
2621 On MeP targets this causes the compiler to use a calling convention
2622 that assumes the called function is too far away for the built-in
2625 @item fast_interrupt
2626 @cindex interrupt handler functions
2627 Use this attribute on the M32C and RX ports to indicate that the specified
2628 function is a fast interrupt handler. This is just like the
2629 @code{interrupt} attribute, except that @code{freit} is used to return
2630 instead of @code{reit}.
2633 @cindex functions that pop the argument stack on the 386
2634 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2635 pass the first argument (if of integral type) in the register ECX and
2636 the second argument (if of integral type) in the register EDX@. Subsequent
2637 and other typed arguments are passed on the stack. The called function
2638 pops the arguments off the stack. If the number of arguments is variable all
2639 arguments are pushed on the stack.
2642 @cindex functions that pop the argument stack on the 386
2643 On the Intel 386, the @code{thiscall} attribute causes the compiler to
2644 pass the first argument (if of integral type) in the register ECX.
2645 Subsequent and other typed arguments are passed on the stack. The called
2646 function pops the arguments off the stack.
2647 If the number of arguments is variable all arguments are pushed on the
2649 The @code{thiscall} attribute is intended for C++ non-static member functions.
2650 As a GCC extension, this calling convention can be used for C functions
2651 and for static member methods.
2653 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2654 @cindex @code{format} function attribute
2656 The @code{format} attribute specifies that a function takes @code{printf},
2657 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2658 should be type-checked against a format string. For example, the
2663 my_printf (void *my_object, const char *my_format, ...)
2664 __attribute__ ((format (printf, 2, 3)));
2668 causes the compiler to check the arguments in calls to @code{my_printf}
2669 for consistency with the @code{printf} style format string argument
2672 The parameter @var{archetype} determines how the format string is
2673 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2674 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2675 @code{strfmon}. (You can also use @code{__printf__},
2676 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2677 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2678 @code{ms_strftime} are also present.
2679 @var{archetype} values such as @code{printf} refer to the formats accepted
2680 by the system's C runtime library,
2681 while values prefixed with @samp{gnu_} always refer
2682 to the formats accepted by the GNU C Library. On Microsoft Windows
2683 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2684 @file{msvcrt.dll} library.
2685 The parameter @var{string-index}
2686 specifies which argument is the format string argument (starting
2687 from 1), while @var{first-to-check} is the number of the first
2688 argument to check against the format string. For functions
2689 where the arguments are not available to be checked (such as
2690 @code{vprintf}), specify the third parameter as zero. In this case the
2691 compiler only checks the format string for consistency. For
2692 @code{strftime} formats, the third parameter is required to be zero.
2693 Since non-static C++ methods have an implicit @code{this} argument, the
2694 arguments of such methods should be counted from two, not one, when
2695 giving values for @var{string-index} and @var{first-to-check}.
2697 In the example above, the format string (@code{my_format}) is the second
2698 argument of the function @code{my_print}, and the arguments to check
2699 start with the third argument, so the correct parameters for the format
2700 attribute are 2 and 3.
2702 @opindex ffreestanding
2703 @opindex fno-builtin
2704 The @code{format} attribute allows you to identify your own functions
2705 that take format strings as arguments, so that GCC can check the
2706 calls to these functions for errors. The compiler always (unless
2707 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2708 for the standard library functions @code{printf}, @code{fprintf},
2709 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2710 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2711 warnings are requested (using @option{-Wformat}), so there is no need to
2712 modify the header file @file{stdio.h}. In C99 mode, the functions
2713 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2714 @code{vsscanf} are also checked. Except in strictly conforming C
2715 standard modes, the X/Open function @code{strfmon} is also checked as
2716 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2717 @xref{C Dialect Options,,Options Controlling C Dialect}.
2719 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2720 recognized in the same context. Declarations including these format attributes
2721 are parsed for correct syntax, however the result of checking of such format
2722 strings is not yet defined, and is not carried out by this version of the
2725 The target may also provide additional types of format checks.
2726 @xref{Target Format Checks,,Format Checks Specific to Particular
2729 @item format_arg (@var{string-index})
2730 @cindex @code{format_arg} function attribute
2731 @opindex Wformat-nonliteral
2732 The @code{format_arg} attribute specifies that a function takes a format
2733 string for a @code{printf}, @code{scanf}, @code{strftime} or
2734 @code{strfmon} style function and modifies it (for example, to translate
2735 it into another language), so the result can be passed to a
2736 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2737 function (with the remaining arguments to the format function the same
2738 as they would have been for the unmodified string). For example, the
2743 my_dgettext (char *my_domain, const char *my_format)
2744 __attribute__ ((format_arg (2)));
2748 causes the compiler to check the arguments in calls to a @code{printf},
2749 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2750 format string argument is a call to the @code{my_dgettext} function, for
2751 consistency with the format string argument @code{my_format}. If the
2752 @code{format_arg} attribute had not been specified, all the compiler
2753 could tell in such calls to format functions would be that the format
2754 string argument is not constant; this would generate a warning when
2755 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2756 without the attribute.
2758 The parameter @var{string-index} specifies which argument is the format
2759 string argument (starting from one). Since non-static C++ methods have
2760 an implicit @code{this} argument, the arguments of such methods should
2761 be counted from two.
2763 The @code{format_arg} attribute allows you to identify your own
2764 functions that modify format strings, so that GCC can check the
2765 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2766 type function whose operands are a call to one of your own function.
2767 The compiler always treats @code{gettext}, @code{dgettext}, and
2768 @code{dcgettext} in this manner except when strict ISO C support is
2769 requested by @option{-ansi} or an appropriate @option{-std} option, or
2770 @option{-ffreestanding} or @option{-fno-builtin}
2771 is used. @xref{C Dialect Options,,Options
2772 Controlling C Dialect}.
2774 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2775 @code{NSString} reference for compatibility with the @code{format} attribute
2778 The target may also allow additional types in @code{format-arg} attributes.
2779 @xref{Target Format Checks,,Format Checks Specific to Particular
2782 @item function_vector
2783 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2784 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2785 function should be called through the function vector. Calling a
2786 function through the function vector reduces code size, however;
2787 the function vector has a limited size (maximum 128 entries on the H8/300
2788 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2790 On SH2A targets, this attribute declares a function to be called using the
2791 TBR relative addressing mode. The argument to this attribute is the entry
2792 number of the same function in a vector table containing all the TBR
2793 relative addressable functions. For correct operation the TBR must be setup
2794 accordingly to point to the start of the vector table before any functions with
2795 this attribute are invoked. Usually a good place to do the initialization is
2796 the startup routine. The TBR relative vector table can have at max 256 function
2797 entries. The jumps to these functions are generated using a SH2A specific,
2798 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
2799 from GNU binutils version 2.7 or later for this attribute to work correctly.
2801 Please refer the example of M16C target, to see the use of this
2802 attribute while declaring a function,
2804 In an application, for a function being called once, this attribute
2805 saves at least 8 bytes of code; and if other successive calls are being
2806 made to the same function, it saves 2 bytes of code per each of these
2809 On M16C/M32C targets, the @code{function_vector} attribute declares a
2810 special page subroutine call function. Use of this attribute reduces
2811 the code size by 2 bytes for each call generated to the
2812 subroutine. The argument to the attribute is the vector number entry
2813 from the special page vector table which contains the 16 low-order
2814 bits of the subroutine's entry address. Each vector table has special
2815 page number (18 to 255) that is used in @code{jsrs} instructions.
2816 Jump addresses of the routines are generated by adding 0x0F0000 (in
2817 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
2818 2-byte addresses set in the vector table. Therefore you need to ensure
2819 that all the special page vector routines should get mapped within the
2820 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2823 In the following example 2 bytes are saved for each call to
2824 function @code{foo}.
2827 void foo (void) __attribute__((function_vector(0x18)));
2838 If functions are defined in one file and are called in another file,
2839 then be sure to write this declaration in both files.
2841 This attribute is ignored for R8C target.
2843 @item ifunc ("@var{resolver}")
2844 @cindex @code{ifunc} attribute
2845 The @code{ifunc} attribute is used to mark a function as an indirect
2846 function using the STT_GNU_IFUNC symbol type extension to the ELF
2847 standard. This allows the resolution of the symbol value to be
2848 determined dynamically at load time, and an optimized version of the
2849 routine can be selected for the particular processor or other system
2850 characteristics determined then. To use this attribute, first define
2851 the implementation functions available, and a resolver function that
2852 returns a pointer to the selected implementation function. The
2853 implementation functions' declarations must match the API of the
2854 function being implemented, the resolver's declaration is be a
2855 function returning pointer to void function returning void:
2858 void *my_memcpy (void *dst, const void *src, size_t len)
2863 static void (*resolve_memcpy (void)) (void)
2865 return my_memcpy; // we'll just always select this routine
2870 The exported header file declaring the function the user calls would
2874 extern void *memcpy (void *, const void *, size_t);
2878 allowing the user to call this as a regular function, unaware of the
2879 implementation. Finally, the indirect function needs to be defined in
2880 the same translation unit as the resolver function:
2883 void *memcpy (void *, const void *, size_t)
2884 __attribute__ ((ifunc ("resolve_memcpy")));
2887 Indirect functions cannot be weak, and require a recent binutils (at
2888 least version 2.20.1), and GNU C library (at least version 2.11.1).
2891 @cindex interrupt handler functions
2892 Use this attribute on the ARC, ARM, AVR, CR16, Epiphany, M32C, M32R/D,
2893 m68k, MeP, MIPS, MSP430, RL78, RX and Xstormy16 ports to indicate that
2894 the specified function is an
2895 interrupt handler. The compiler generates function entry and exit
2896 sequences suitable for use in an interrupt handler when this attribute
2897 is present. With Epiphany targets it may also generate a special section with
2898 code to initialize the interrupt vector table.
2900 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, MicroBlaze,
2901 and SH processors can be specified via the @code{interrupt_handler} attribute.
2903 Note, on the ARC, you must specify the kind of interrupt to be handled
2904 in a parameter to the interrupt attribute like this:
2907 void f () __attribute__ ((interrupt ("ilink1")));
2910 Permissible values for this parameter are: @w{@code{ilink1}} and
2913 Note, on the AVR, the hardware globally disables interrupts when an
2914 interrupt is executed. The first instruction of an interrupt handler
2915 declared with this attribute is a @code{SEI} instruction to
2916 re-enable interrupts. See also the @code{signal} function attribute
2917 that does not insert a @code{SEI} instruction. If both @code{signal} and
2918 @code{interrupt} are specified for the same function, @code{signal}
2919 is silently ignored.
2921 Note, for the ARM, you can specify the kind of interrupt to be handled by
2922 adding an optional parameter to the interrupt attribute like this:
2925 void f () __attribute__ ((interrupt ("IRQ")));
2929 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
2930 @code{SWI}, @code{ABORT} and @code{UNDEF}.
2932 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2933 may be called with a word-aligned stack pointer.
2935 Note, for the MSP430 you can provide an argument to the interrupt
2936 attribute which specifies a name or number. If the argument is a
2937 number it indicates the slot in the interrupt vector table (0 - 31) to
2938 which this handler should be assigned. If the argument is a name it
2939 is treated as a symbolic name for the vector slot. These names should
2940 match up with appropriate entries in the linker script. By default
2941 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
2942 @code{reset} for vector 31 are recognised.
2944 You can also use the following function attributes to modify how
2945 normal functions interact with interrupt functions:
2949 @cindex @code{critical} attribute
2950 Critical functions disable interrupts upon entry and restore the
2951 previous interrupt state upon exit. Critical functions cannot also
2952 have the @code{naked} or @code{reentrant} attributes. They can have
2953 the @code{interrupt} attribute.
2956 @cindex @code{reentrant} attribute
2957 Reentrant functions disable interrupts upon entry and enable them
2958 upon exit. Reentrant functions cannot also have the @code{naked}
2959 or @code{critical} attributes. They can have the @code{interrupt}
2963 @cindex @code{wakeup} attribute
2964 This attribute only applies to interrupt functions. It is silently
2965 ignored if applied to a non-interrupt function. A wakeup interrupt
2966 function will rouse the processor from any low-power state that it
2967 might be in when the function exits.
2971 On Epiphany targets one or more optional parameters can be added like this:
2974 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
2977 Permissible values for these parameters are: @w{@code{reset}},
2978 @w{@code{software_exception}}, @w{@code{page_miss}},
2979 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
2980 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
2981 Multiple parameters indicate that multiple entries in the interrupt
2982 vector table should be initialized for this function, i.e.@: for each
2983 parameter @w{@var{name}}, a jump to the function is emitted in
2984 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
2985 entirely, in which case no interrupt vector table entry is provided.
2987 Note, on Epiphany targets, interrupts are enabled inside the function
2988 unless the @code{disinterrupt} attribute is also specified.
2990 On Epiphany targets, you can also use the following attribute to
2991 modify the behavior of an interrupt handler:
2993 @item forwarder_section
2994 @cindex @code{forwarder_section} attribute
2995 The interrupt handler may be in external memory which cannot be
2996 reached by a branch instruction, so generate a local memory trampoline
2997 to transfer control. The single parameter identifies the section where
2998 the trampoline is placed.
3001 The following examples are all valid uses of these attributes on
3004 void __attribute__ ((interrupt)) universal_handler ();
3005 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3006 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3007 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3008 fast_timer_handler ();
3009 void __attribute__ ((interrupt ("dma0, dma1"), forwarder_section ("tramp")))
3010 external_dma_handler ();
3013 On MIPS targets, you can use the following attributes to modify the behavior
3014 of an interrupt handler:
3016 @item use_shadow_register_set
3017 @cindex @code{use_shadow_register_set} attribute
3018 Assume that the handler uses a shadow register set, instead of
3019 the main general-purpose registers.
3021 @item keep_interrupts_masked
3022 @cindex @code{keep_interrupts_masked} attribute
3023 Keep interrupts masked for the whole function. Without this attribute,
3024 GCC tries to reenable interrupts for as much of the function as it can.
3026 @item use_debug_exception_return
3027 @cindex @code{use_debug_exception_return} attribute
3028 Return using the @code{deret} instruction. Interrupt handlers that don't
3029 have this attribute return using @code{eret} instead.
3032 You can use any combination of these attributes, as shown below:
3034 void __attribute__ ((interrupt)) v0 ();
3035 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
3036 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
3037 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
3038 void __attribute__ ((interrupt, use_shadow_register_set,
3039 keep_interrupts_masked)) v4 ();
3040 void __attribute__ ((interrupt, use_shadow_register_set,
3041 use_debug_exception_return)) v5 ();
3042 void __attribute__ ((interrupt, keep_interrupts_masked,
3043 use_debug_exception_return)) v6 ();
3044 void __attribute__ ((interrupt, use_shadow_register_set,
3045 keep_interrupts_masked,
3046 use_debug_exception_return)) v7 ();
3049 On NDS32 target, this attribute is to indicate that the specified function
3050 is an interrupt handler. The compiler will generate corresponding sections
3051 for use in an interrupt handler. You can use the following attributes
3052 to modify the behavior:
3055 @cindex @code{nested} attribute
3056 This interrupt service routine is interruptible.
3058 @cindex @code{not_nested} attribute
3059 This interrupt service routine is not interruptible.
3061 @cindex @code{nested_ready} attribute
3062 This interrupt service routine is interruptible after @code{PSW.GIE}
3063 (global interrupt enable) is set. This allows interrupt service routine to
3064 finish some short critical code before enabling interrupts.
3066 @cindex @code{save_all} attribute
3067 The system will help save all registers into stack before entering
3070 @cindex @code{partial_save} attribute
3071 The system will help save caller registers into stack before entering
3075 On RL78, use @code{brk_interrupt} instead of @code{interrupt} for
3076 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
3077 that must end with @code{RETB} instead of @code{RETI}).
3079 @item interrupt_handler
3080 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
3081 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
3082 indicate that the specified function is an interrupt handler. The compiler
3083 generates function entry and exit sequences suitable for use in an
3084 interrupt handler when this attribute is present.
3086 @item interrupt_thread
3087 @cindex interrupt thread functions on fido
3088 Use this attribute on fido, a subarchitecture of the m68k, to indicate
3089 that the specified function is an interrupt handler that is designed
3090 to run as a thread. The compiler omits generate prologue/epilogue
3091 sequences and replaces the return instruction with a @code{sleep}
3092 instruction. This attribute is available only on fido.
3095 @cindex interrupt service routines on ARM
3096 Use this attribute on ARM to write Interrupt Service Routines. This is an
3097 alias to the @code{interrupt} attribute above.
3100 @cindex User stack pointer in interrupts on the Blackfin
3101 When used together with @code{interrupt_handler}, @code{exception_handler}
3102 or @code{nmi_handler}, code is generated to load the stack pointer
3103 from the USP register in the function prologue.
3106 @cindex @code{l1_text} function attribute
3107 This attribute specifies a function to be placed into L1 Instruction
3108 SRAM@. The function is put into a specific section named @code{.l1.text}.
3109 With @option{-mfdpic}, function calls with a such function as the callee
3110 or caller uses inlined PLT.
3113 @cindex @code{l2} function attribute
3114 On the Blackfin, this attribute specifies a function to be placed into L2
3115 SRAM. The function is put into a specific section named
3116 @code{.l1.text}. With @option{-mfdpic}, callers of such functions use
3120 @cindex @code{leaf} function attribute
3121 Calls to external functions with this attribute must return to the current
3122 compilation unit only by return or by exception handling. In particular, leaf
3123 functions are not allowed to call callback function passed to it from the current
3124 compilation unit or directly call functions exported by the unit or longjmp
3125 into the unit. Leaf function might still call functions from other compilation
3126 units and thus they are not necessarily leaf in the sense that they contain no
3127 function calls at all.
3129 The attribute is intended for library functions to improve dataflow analysis.
3130 The compiler takes the hint that any data not escaping the current compilation unit can
3131 not be used or modified by the leaf function. For example, the @code{sin} function
3132 is a leaf function, but @code{qsort} is not.
3134 Note that leaf functions might invoke signals and signal handlers might be
3135 defined in the current compilation unit and use static variables. The only
3136 compliant way to write such a signal handler is to declare such variables
3139 The attribute has no effect on functions defined within the current compilation
3140 unit. This is to allow easy merging of multiple compilation units into one,
3141 for example, by using the link-time optimization. For this reason the
3142 attribute is not allowed on types to annotate indirect calls.
3144 @item long_call/medium_call/short_call
3145 @cindex indirect calls on ARC
3146 @cindex indirect calls on ARM
3147 @cindex indirect calls on Epiphany
3148 These attributes specify how a particular function is called on
3149 ARC, ARM and Epiphany - with @code{medium_call} being specific to ARC.
3150 These attributes override the
3151 @option{-mlong-calls} (@pxref{ARM Options} and @ref{ARC Options})
3152 and @option{-mmedium-calls} (@pxref{ARC Options})
3153 command-line switches and @code{#pragma long_calls} settings. For ARM, the
3154 @code{long_call} attribute indicates that the function might be far
3155 away from the call site and require a different (more expensive)
3156 calling sequence. The @code{short_call} attribute always places
3157 the offset to the function from the call site into the @samp{BL}
3158 instruction directly.
3160 For ARC, a function marked with the @code{long_call} attribute is
3161 always called using register-indirect jump-and-link instructions,
3162 thereby enabling the called function to be placed anywhere within the
3163 32-bit address space. A function marked with the @code{medium_call}
3164 attribute will always be close enough to be called with an unconditional
3165 branch-and-link instruction, which has a 25-bit offset from
3166 the call site. A function marked with the @code{short_call}
3167 attribute will always be close enough to be called with a conditional
3168 branch-and-link instruction, which has a 21-bit offset from
3171 @item longcall/shortcall
3172 @cindex functions called via pointer on the RS/6000 and PowerPC
3173 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
3174 indicates that the function might be far away from the call site and
3175 require a different (more expensive) calling sequence. The
3176 @code{shortcall} attribute indicates that the function is always close
3177 enough for the shorter calling sequence to be used. These attributes
3178 override both the @option{-mlongcall} switch and, on the RS/6000 and
3179 PowerPC, the @code{#pragma longcall} setting.
3181 @xref{RS/6000 and PowerPC Options}, for more information on whether long
3182 calls are necessary.
3184 @item long_call/near/far
3185 @cindex indirect calls on MIPS
3186 These attributes specify how a particular function is called on MIPS@.
3187 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
3188 command-line switch. The @code{long_call} and @code{far} attributes are
3189 synonyms, and cause the compiler to always call
3190 the function by first loading its address into a register, and then using
3191 the contents of that register. The @code{near} attribute has the opposite
3192 effect; it specifies that non-PIC calls should be made using the more
3193 efficient @code{jal} instruction.
3196 @cindex @code{malloc} attribute
3197 The @code{malloc} attribute is used to tell the compiler that a function
3198 may be treated as if any non-@code{NULL} pointer it returns cannot
3199 alias any other pointer valid when the function returns and that the memory
3200 has undefined content.
3201 This often improves optimization.
3202 Standard functions with this property include @code{malloc} and
3203 @code{calloc}. @code{realloc}-like functions do not have this
3204 property as the memory pointed to does not have undefined content.
3206 @item mips16/nomips16
3207 @cindex @code{mips16} attribute
3208 @cindex @code{nomips16} attribute
3210 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
3211 function attributes to locally select or turn off MIPS16 code generation.
3212 A function with the @code{mips16} attribute is emitted as MIPS16 code,
3213 while MIPS16 code generation is disabled for functions with the
3214 @code{nomips16} attribute. These attributes override the
3215 @option{-mips16} and @option{-mno-mips16} options on the command line
3216 (@pxref{MIPS Options}).
3218 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
3219 preprocessor symbol @code{__mips16} reflects the setting on the command line,
3220 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
3221 may interact badly with some GCC extensions such as @code{__builtin_apply}
3222 (@pxref{Constructing Calls}).
3224 @item micromips/nomicromips
3225 @cindex @code{micromips} attribute
3226 @cindex @code{nomicromips} attribute
3228 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
3229 function attributes to locally select or turn off microMIPS code generation.
3230 A function with the @code{micromips} attribute is emitted as microMIPS code,
3231 while microMIPS code generation is disabled for functions with the
3232 @code{nomicromips} attribute. These attributes override the
3233 @option{-mmicromips} and @option{-mno-micromips} options on the command line
3234 (@pxref{MIPS Options}).
3236 When compiling files containing mixed microMIPS and non-microMIPS code, the
3237 preprocessor symbol @code{__mips_micromips} reflects the setting on the
3239 not that within individual functions. Mixed microMIPS and non-microMIPS code
3240 may interact badly with some GCC extensions such as @code{__builtin_apply}
3241 (@pxref{Constructing Calls}).
3243 @item model (@var{model-name})
3244 @cindex function addressability on the M32R/D
3245 @cindex variable addressability on the IA-64
3247 On the M32R/D, use this attribute to set the addressability of an
3248 object, and of the code generated for a function. The identifier
3249 @var{model-name} is one of @code{small}, @code{medium}, or
3250 @code{large}, representing each of the code models.
3252 Small model objects live in the lower 16MB of memory (so that their
3253 addresses can be loaded with the @code{ld24} instruction), and are
3254 callable with the @code{bl} instruction.
3256 Medium model objects may live anywhere in the 32-bit address space (the
3257 compiler generates @code{seth/add3} instructions to load their addresses),
3258 and are callable with the @code{bl} instruction.
3260 Large model objects may live anywhere in the 32-bit address space (the
3261 compiler generates @code{seth/add3} instructions to load their addresses),
3262 and may not be reachable with the @code{bl} instruction (the compiler
3263 generates the much slower @code{seth/add3/jl} instruction sequence).
3265 On IA-64, use this attribute to set the addressability of an object.
3266 At present, the only supported identifier for @var{model-name} is
3267 @code{small}, indicating addressability via ``small'' (22-bit)
3268 addresses (so that their addresses can be loaded with the @code{addl}
3269 instruction). Caveat: such addressing is by definition not position
3270 independent and hence this attribute must not be used for objects
3271 defined by shared libraries.
3273 @item ms_abi/sysv_abi
3274 @cindex @code{ms_abi} attribute
3275 @cindex @code{sysv_abi} attribute
3277 On 32-bit and 64-bit (i?86|x86_64)-*-* targets, you can use an ABI attribute
3278 to indicate which calling convention should be used for a function. The
3279 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
3280 while the @code{sysv_abi} attribute tells the compiler to use the ABI
3281 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
3282 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
3284 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
3285 requires the @option{-maccumulate-outgoing-args} option.
3287 @item callee_pop_aggregate_return (@var{number})
3288 @cindex @code{callee_pop_aggregate_return} attribute
3290 On 32-bit i?86-*-* targets, you can use this attribute to control how
3291 aggregates are returned in memory. If the caller is responsible for
3292 popping the hidden pointer together with the rest of the arguments, specify
3293 @var{number} equal to zero. If callee is responsible for popping the
3294 hidden pointer, specify @var{number} equal to one.
3296 The default i386 ABI assumes that the callee pops the
3297 stack for hidden pointer. However, on 32-bit i386 Microsoft Windows targets,
3298 the compiler assumes that the
3299 caller pops the stack for hidden pointer.
3301 @item ms_hook_prologue
3302 @cindex @code{ms_hook_prologue} attribute
3304 On 32-bit i[34567]86-*-* targets and 64-bit x86_64-*-* targets, you can use
3305 this function attribute to make GCC generate the ``hot-patching'' function
3306 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
3309 @item hotpatch [(@var{prologue-halfwords})]
3310 @cindex @code{hotpatch} attribute
3312 On S/390 System z targets, you can use this function attribute to
3313 make GCC generate a ``hot-patching'' function prologue. The
3314 @code{hotpatch} has no effect on funtions that are explicitly
3315 inline. If the @option{-mhotpatch} or @option{-mno-hotpatch}
3316 command-line option is used at the same time, the @code{hotpatch}
3317 attribute takes precedence. If an argument is given, the maximum
3318 allowed value is 1000000.
3321 @cindex function without a prologue/epilogue code
3322 Use this attribute on the ARM, AVR, MCORE, MSP430, NDS32, RL78, RX and SPU
3323 ports to indicate that the specified function does not need prologue/epilogue
3324 sequences generated by the compiler.
3325 It is up to the programmer to provide these sequences. The
3326 only statements that can be safely included in naked functions are
3327 @code{asm} statements that do not have operands. All other statements,
3328 including declarations of local variables, @code{if} statements, and so
3329 forth, should be avoided. Naked functions should be used to implement the
3330 body of an assembly function, while allowing the compiler to construct
3331 the requisite function declaration for the assembler.
3334 @cindex functions that do not handle memory bank switching on 68HC11/68HC12
3335 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
3336 use the normal calling convention based on @code{jsr} and @code{rts}.
3337 This attribute can be used to cancel the effect of the @option{-mlong-calls}
3340 On MeP targets this attribute causes the compiler to assume the called
3341 function is close enough to use the normal calling convention,
3342 overriding the @option{-mtf} command-line option.
3345 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
3346 Use this attribute together with @code{interrupt_handler},
3347 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3348 entry code should enable nested interrupts or exceptions.
3351 @cindex NMI handler functions on the Blackfin processor
3352 Use this attribute on the Blackfin to indicate that the specified function
3353 is an NMI handler. The compiler generates function entry and
3354 exit sequences suitable for use in an NMI handler when this
3355 attribute is present.
3358 @cindex @code{nocompression} attribute
3359 On MIPS targets, you can use the @code{nocompression} function attribute
3360 to locally turn off MIPS16 and microMIPS code generation. This attribute
3361 overrides the @option{-mips16} and @option{-mmicromips} options on the
3362 command line (@pxref{MIPS Options}).
3364 @item no_instrument_function
3365 @cindex @code{no_instrument_function} function attribute
3366 @opindex finstrument-functions
3367 If @option{-finstrument-functions} is given, profiling function calls are
3368 generated at entry and exit of most user-compiled functions.
3369 Functions with this attribute are not so instrumented.
3371 @item no_split_stack
3372 @cindex @code{no_split_stack} function attribute
3373 @opindex fsplit-stack
3374 If @option{-fsplit-stack} is given, functions have a small
3375 prologue which decides whether to split the stack. Functions with the
3376 @code{no_split_stack} attribute do not have that prologue, and thus
3377 may run with only a small amount of stack space available.
3380 @cindex @code{noinline} function attribute
3381 This function attribute prevents a function from being considered for
3383 @c Don't enumerate the optimizations by name here; we try to be
3384 @c future-compatible with this mechanism.
3385 If the function does not have side-effects, there are optimizations
3386 other than inlining that cause function calls to be optimized away,
3387 although the function call is live. To keep such calls from being
3394 (@pxref{Extended Asm}) in the called function, to serve as a special
3398 @cindex @code{noclone} function attribute
3399 This function attribute prevents a function from being considered for
3400 cloning---a mechanism that produces specialized copies of functions
3401 and which is (currently) performed by interprocedural constant
3404 @item nonnull (@var{arg-index}, @dots{})
3405 @cindex @code{nonnull} function attribute
3406 The @code{nonnull} attribute specifies that some function parameters should
3407 be non-null pointers. For instance, the declaration:
3411 my_memcpy (void *dest, const void *src, size_t len)
3412 __attribute__((nonnull (1, 2)));
3416 causes the compiler to check that, in calls to @code{my_memcpy},
3417 arguments @var{dest} and @var{src} are non-null. If the compiler
3418 determines that a null pointer is passed in an argument slot marked
3419 as non-null, and the @option{-Wnonnull} option is enabled, a warning
3420 is issued. The compiler may also choose to make optimizations based
3421 on the knowledge that certain function arguments will never be null.
3423 If no argument index list is given to the @code{nonnull} attribute,
3424 all pointer arguments are marked as non-null. To illustrate, the
3425 following declaration is equivalent to the previous example:
3429 my_memcpy (void *dest, const void *src, size_t len)
3430 __attribute__((nonnull));
3433 @item returns_nonnull
3434 @cindex @code{returns_nonnull} function attribute
3435 The @code{returns_nonnull} attribute specifies that the function
3436 return value should be a non-null pointer. For instance, the declaration:
3440 mymalloc (size_t len) __attribute__((returns_nonnull));
3444 lets the compiler optimize callers based on the knowledge
3445 that the return value will never be null.
3448 @cindex @code{noreturn} function attribute
3449 A few standard library functions, such as @code{abort} and @code{exit},
3450 cannot return. GCC knows this automatically. Some programs define
3451 their own functions that never return. You can declare them
3452 @code{noreturn} to tell the compiler this fact. For example,
3456 void fatal () __attribute__ ((noreturn));
3459 fatal (/* @r{@dots{}} */)
3461 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3467 The @code{noreturn} keyword tells the compiler to assume that
3468 @code{fatal} cannot return. It can then optimize without regard to what
3469 would happen if @code{fatal} ever did return. This makes slightly
3470 better code. More importantly, it helps avoid spurious warnings of
3471 uninitialized variables.
3473 The @code{noreturn} keyword does not affect the exceptional path when that
3474 applies: a @code{noreturn}-marked function may still return to the caller
3475 by throwing an exception or calling @code{longjmp}.
3477 Do not assume that registers saved by the calling function are
3478 restored before calling the @code{noreturn} function.
3480 It does not make sense for a @code{noreturn} function to have a return
3481 type other than @code{void}.
3483 The attribute @code{noreturn} is not implemented in GCC versions
3484 earlier than 2.5. An alternative way to declare that a function does
3485 not return, which works in the current version and in some older
3486 versions, is as follows:
3489 typedef void voidfn ();
3491 volatile voidfn fatal;
3495 This approach does not work in GNU C++.
3498 @cindex @code{nothrow} function attribute
3499 The @code{nothrow} attribute is used to inform the compiler that a
3500 function cannot throw an exception. For example, most functions in
3501 the standard C library can be guaranteed not to throw an exception
3502 with the notable exceptions of @code{qsort} and @code{bsearch} that
3503 take function pointer arguments. The @code{nothrow} attribute is not
3504 implemented in GCC versions earlier than 3.3.
3506 @item nosave_low_regs
3507 @cindex @code{nosave_low_regs} attribute
3508 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
3509 function should not save and restore registers R0..R7. This can be used on SH3*
3510 and SH4* targets that have a second R0..R7 register bank for non-reentrant
3514 @cindex @code{optimize} function attribute
3515 The @code{optimize} attribute is used to specify that a function is to
3516 be compiled with different optimization options than specified on the
3517 command line. Arguments can either be numbers or strings. Numbers
3518 are assumed to be an optimization level. Strings that begin with
3519 @code{O} are assumed to be an optimization option, while other options
3520 are assumed to be used with a @code{-f} prefix. You can also use the
3521 @samp{#pragma GCC optimize} pragma to set the optimization options
3522 that affect more than one function.
3523 @xref{Function Specific Option Pragmas}, for details about the
3524 @samp{#pragma GCC optimize} pragma.
3526 This can be used for instance to have frequently-executed functions
3527 compiled with more aggressive optimization options that produce faster
3528 and larger code, while other functions can be compiled with less
3531 @item OS_main/OS_task
3532 @cindex @code{OS_main} AVR function attribute
3533 @cindex @code{OS_task} AVR function attribute
3534 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3535 do not save/restore any call-saved register in their prologue/epilogue.
3537 The @code{OS_main} attribute can be used when there @emph{is
3538 guarantee} that interrupts are disabled at the time when the function
3539 is entered. This saves resources when the stack pointer has to be
3540 changed to set up a frame for local variables.
3542 The @code{OS_task} attribute can be used when there is @emph{no
3543 guarantee} that interrupts are disabled at that time when the function
3544 is entered like for, e@.g@. task functions in a multi-threading operating
3545 system. In that case, changing the stack pointer register is
3546 guarded by save/clear/restore of the global interrupt enable flag.
3548 The differences to the @code{naked} function attribute are:
3550 @item @code{naked} functions do not have a return instruction whereas
3551 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3552 @code{RETI} return instruction.
3553 @item @code{naked} functions do not set up a frame for local variables
3554 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3559 @cindex @code{pcs} function attribute
3561 The @code{pcs} attribute can be used to control the calling convention
3562 used for a function on ARM. The attribute takes an argument that specifies
3563 the calling convention to use.
3565 When compiling using the AAPCS ABI (or a variant of it) then valid
3566 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3567 order to use a variant other than @code{"aapcs"} then the compiler must
3568 be permitted to use the appropriate co-processor registers (i.e., the
3569 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3573 /* Argument passed in r0, and result returned in r0+r1. */
3574 double f2d (float) __attribute__((pcs("aapcs")));
3577 Variadic functions always use the @code{"aapcs"} calling convention and
3578 the compiler rejects attempts to specify an alternative.
3581 @cindex @code{pure} function attribute
3582 Many functions have no effects except the return value and their
3583 return value depends only on the parameters and/or global variables.
3584 Such a function can be subject
3585 to common subexpression elimination and loop optimization just as an
3586 arithmetic operator would be. These functions should be declared
3587 with the attribute @code{pure}. For example,
3590 int square (int) __attribute__ ((pure));
3594 says that the hypothetical function @code{square} is safe to call
3595 fewer times than the program says.
3597 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3598 Interesting non-pure functions are functions with infinite loops or those
3599 depending on volatile memory or other system resource, that may change between
3600 two consecutive calls (such as @code{feof} in a multithreading environment).
3602 The attribute @code{pure} is not implemented in GCC versions earlier
3606 @cindex @code{hot} function attribute
3607 The @code{hot} attribute on a function is used to inform the compiler that
3608 the function is a hot spot of the compiled program. The function is
3609 optimized more aggressively and on many target it is placed into special
3610 subsection of the text section so all hot functions appears close together
3613 When profile feedback is available, via @option{-fprofile-use}, hot functions
3614 are automatically detected and this attribute is ignored.
3616 The @code{hot} attribute on functions is not implemented in GCC versions
3619 @cindex @code{hot} label attribute
3620 The @code{hot} attribute on a label is used to inform the compiler that
3621 path following the label are more likely than paths that are not so
3622 annotated. This attribute is used in cases where @code{__builtin_expect}
3623 cannot be used, for instance with computed goto or @code{asm goto}.
3625 The @code{hot} attribute on labels is not implemented in GCC versions
3629 @cindex @code{cold} function attribute
3630 The @code{cold} attribute on functions is used to inform the compiler that
3631 the function is unlikely to be executed. The function is optimized for
3632 size rather than speed and on many targets it is placed into special
3633 subsection of the text section so all cold functions appears close together
3634 improving code locality of non-cold parts of program. The paths leading
3635 to call of cold functions within code are marked as unlikely by the branch
3636 prediction mechanism. It is thus useful to mark functions used to handle
3637 unlikely conditions, such as @code{perror}, as cold to improve optimization
3638 of hot functions that do call marked functions in rare occasions.
3640 When profile feedback is available, via @option{-fprofile-use}, cold functions
3641 are automatically detected and this attribute is ignored.
3643 The @code{cold} attribute on functions is not implemented in GCC versions
3646 @cindex @code{cold} label attribute
3647 The @code{cold} attribute on labels is used to inform the compiler that
3648 the path following the label is unlikely to be executed. This attribute
3649 is used in cases where @code{__builtin_expect} cannot be used, for instance
3650 with computed goto or @code{asm goto}.
3652 The @code{cold} attribute on labels is not implemented in GCC versions
3655 @item no_sanitize_address
3656 @itemx no_address_safety_analysis
3657 @cindex @code{no_sanitize_address} function attribute
3658 The @code{no_sanitize_address} attribute on functions is used
3659 to inform the compiler that it should not instrument memory accesses
3660 in the function when compiling with the @option{-fsanitize=address} option.
3661 The @code{no_address_safety_analysis} is a deprecated alias of the
3662 @code{no_sanitize_address} attribute, new code should use
3663 @code{no_sanitize_address}.
3665 @item no_sanitize_undefined
3666 @cindex @code{no_sanitize_undefined} function attribute
3667 The @code{no_sanitize_undefined} attribute on functions is used
3668 to inform the compiler that it should not check for undefined behavior
3669 in the function when compiling with the @option{-fsanitize=undefined} option.
3671 @item regparm (@var{number})
3672 @cindex @code{regparm} attribute
3673 @cindex functions that are passed arguments in registers on the 386
3674 On the Intel 386, the @code{regparm} attribute causes the compiler to
3675 pass arguments number one to @var{number} if they are of integral type
3676 in registers EAX, EDX, and ECX instead of on the stack. Functions that
3677 take a variable number of arguments continue to be passed all of their
3678 arguments on the stack.
3680 Beware that on some ELF systems this attribute is unsuitable for
3681 global functions in shared libraries with lazy binding (which is the
3682 default). Lazy binding sends the first call via resolving code in
3683 the loader, which might assume EAX, EDX and ECX can be clobbered, as
3684 per the standard calling conventions. Solaris 8 is affected by this.
3685 Systems with the GNU C Library version 2.1 or higher
3686 and FreeBSD are believed to be
3687 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
3688 disabled with the linker or the loader if desired, to avoid the
3692 @cindex reset handler functions
3693 Use this attribute on the NDS32 target to indicate that the specified function
3694 is a reset handler. The compiler will generate corresponding sections
3695 for use in a reset handler. You can use the following attributes
3696 to provide extra exception handling:
3699 @cindex @code{nmi} attribute
3700 Provide a user-defined function to handle NMI exception.
3702 @cindex @code{warm} attribute
3703 Provide a user-defined function to handle warm reset exception.
3707 @cindex @code{sseregparm} attribute
3708 On the Intel 386 with SSE support, the @code{sseregparm} attribute
3709 causes the compiler to pass up to 3 floating-point arguments in
3710 SSE registers instead of on the stack. Functions that take a
3711 variable number of arguments continue to pass all of their
3712 floating-point arguments on the stack.
3714 @item force_align_arg_pointer
3715 @cindex @code{force_align_arg_pointer} attribute
3716 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
3717 applied to individual function definitions, generating an alternate
3718 prologue and epilogue that realigns the run-time stack if necessary.
3719 This supports mixing legacy codes that run with a 4-byte aligned stack
3720 with modern codes that keep a 16-byte stack for SSE compatibility.
3723 @cindex @code{renesas} attribute
3724 On SH targets this attribute specifies that the function or struct follows the
3728 @cindex @code{resbank} attribute
3729 On the SH2A target, this attribute enables the high-speed register
3730 saving and restoration using a register bank for @code{interrupt_handler}
3731 routines. Saving to the bank is performed automatically after the CPU
3732 accepts an interrupt that uses a register bank.
3734 The nineteen 32-bit registers comprising general register R0 to R14,
3735 control register GBR, and system registers MACH, MACL, and PR and the
3736 vector table address offset are saved into a register bank. Register
3737 banks are stacked in first-in last-out (FILO) sequence. Restoration
3738 from the bank is executed by issuing a RESBANK instruction.
3741 @cindex @code{returns_twice} attribute
3742 The @code{returns_twice} attribute tells the compiler that a function may
3743 return more than one time. The compiler ensures that all registers
3744 are dead before calling such a function and emits a warning about
3745 the variables that may be clobbered after the second return from the
3746 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3747 The @code{longjmp}-like counterpart of such function, if any, might need
3748 to be marked with the @code{noreturn} attribute.
3751 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
3752 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
3753 all registers except the stack pointer should be saved in the prologue
3754 regardless of whether they are used or not.
3756 @item save_volatiles
3757 @cindex save volatile registers on the MicroBlaze
3758 Use this attribute on the MicroBlaze to indicate that the function is
3759 an interrupt handler. All volatile registers (in addition to non-volatile
3760 registers) are saved in the function prologue. If the function is a leaf
3761 function, only volatiles used by the function are saved. A normal function
3762 return is generated instead of a return from interrupt.
3764 @item section ("@var{section-name}")
3765 @cindex @code{section} function attribute
3766 Normally, the compiler places the code it generates in the @code{text} section.
3767 Sometimes, however, you need additional sections, or you need certain
3768 particular functions to appear in special sections. The @code{section}
3769 attribute specifies that a function lives in a particular section.
3770 For example, the declaration:
3773 extern void foobar (void) __attribute__ ((section ("bar")));
3777 puts the function @code{foobar} in the @code{bar} section.
3779 Some file formats do not support arbitrary sections so the @code{section}
3780 attribute is not available on all platforms.
3781 If you need to map the entire contents of a module to a particular
3782 section, consider using the facilities of the linker instead.
3785 @cindex @code{sentinel} function attribute
3786 This function attribute ensures that a parameter in a function call is
3787 an explicit @code{NULL}. The attribute is only valid on variadic
3788 functions. By default, the sentinel is located at position zero, the
3789 last parameter of the function call. If an optional integer position
3790 argument P is supplied to the attribute, the sentinel must be located at
3791 position P counting backwards from the end of the argument list.
3794 __attribute__ ((sentinel))
3796 __attribute__ ((sentinel(0)))
3799 The attribute is automatically set with a position of 0 for the built-in
3800 functions @code{execl} and @code{execlp}. The built-in function
3801 @code{execle} has the attribute set with a position of 1.
3803 A valid @code{NULL} in this context is defined as zero with any pointer
3804 type. If your system defines the @code{NULL} macro with an integer type
3805 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3806 with a copy that redefines NULL appropriately.
3808 The warnings for missing or incorrect sentinels are enabled with
3812 See @code{long_call/short_call}.
3815 See @code{longcall/shortcall}.
3818 @cindex interrupt handler functions on the AVR processors
3819 Use this attribute on the AVR to indicate that the specified
3820 function is an interrupt handler. The compiler generates function
3821 entry and exit sequences suitable for use in an interrupt handler when this
3822 attribute is present.
3824 See also the @code{interrupt} function attribute.
3826 The AVR hardware globally disables interrupts when an interrupt is executed.
3827 Interrupt handler functions defined with the @code{signal} attribute
3828 do not re-enable interrupts. It is save to enable interrupts in a
3829 @code{signal} handler. This ``save'' only applies to the code
3830 generated by the compiler and not to the IRQ layout of the
3831 application which is responsibility of the application.
3833 If both @code{signal} and @code{interrupt} are specified for the same
3834 function, @code{signal} is silently ignored.
3837 @cindex @code{sp_switch} attribute
3838 Use this attribute on the SH to indicate an @code{interrupt_handler}
3839 function should switch to an alternate stack. It expects a string
3840 argument that names a global variable holding the address of the
3845 void f () __attribute__ ((interrupt_handler,
3846 sp_switch ("alt_stack")));
3850 @cindex functions that pop the argument stack on the 386
3851 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3852 assume that the called function pops off the stack space used to
3853 pass arguments, unless it takes a variable number of arguments.
3855 @item syscall_linkage
3856 @cindex @code{syscall_linkage} attribute
3857 This attribute is used to modify the IA-64 calling convention by marking
3858 all input registers as live at all function exits. This makes it possible
3859 to restart a system call after an interrupt without having to save/restore
3860 the input registers. This also prevents kernel data from leaking into
3864 @cindex @code{target} function attribute
3865 The @code{target} attribute is used to specify that a function is to
3866 be compiled with different target options than specified on the
3867 command line. This can be used for instance to have functions
3868 compiled with a different ISA (instruction set architecture) than the
3869 default. You can also use the @samp{#pragma GCC target} pragma to set
3870 more than one function to be compiled with specific target options.
3871 @xref{Function Specific Option Pragmas}, for details about the
3872 @samp{#pragma GCC target} pragma.
3874 For instance on a 386, you could compile one function with
3875 @code{target("sse4.1,arch=core2")} and another with
3876 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3877 compiling the first function with @option{-msse4.1} and
3878 @option{-march=core2} options, and the second function with
3879 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3880 user to make sure that a function is only invoked on a machine that
3881 supports the particular ISA it is compiled for (for example by using
3882 @code{cpuid} on 386 to determine what feature bits and architecture
3886 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3887 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3890 You can either use multiple
3891 strings to specify multiple options, or separate the options
3892 with a comma (@samp{,}).
3894 The @code{target} attribute is presently implemented for
3895 i386/x86_64, PowerPC, and Nios II targets only.
3896 The options supported are specific to each target.
3898 On the 386, the following options are allowed:
3903 @cindex @code{target("abm")} attribute
3904 Enable/disable the generation of the advanced bit instructions.
3908 @cindex @code{target("aes")} attribute
3909 Enable/disable the generation of the AES instructions.
3912 @cindex @code{target("default")} attribute
3913 @xref{Function Multiversioning}, where it is used to specify the
3914 default function version.
3918 @cindex @code{target("mmx")} attribute
3919 Enable/disable the generation of the MMX instructions.
3923 @cindex @code{target("pclmul")} attribute
3924 Enable/disable the generation of the PCLMUL instructions.
3928 @cindex @code{target("popcnt")} attribute
3929 Enable/disable the generation of the POPCNT instruction.
3933 @cindex @code{target("sse")} attribute
3934 Enable/disable the generation of the SSE instructions.
3938 @cindex @code{target("sse2")} attribute
3939 Enable/disable the generation of the SSE2 instructions.
3943 @cindex @code{target("sse3")} attribute
3944 Enable/disable the generation of the SSE3 instructions.
3948 @cindex @code{target("sse4")} attribute
3949 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3954 @cindex @code{target("sse4.1")} attribute
3955 Enable/disable the generation of the sse4.1 instructions.
3959 @cindex @code{target("sse4.2")} attribute
3960 Enable/disable the generation of the sse4.2 instructions.
3964 @cindex @code{target("sse4a")} attribute
3965 Enable/disable the generation of the SSE4A instructions.
3969 @cindex @code{target("fma4")} attribute
3970 Enable/disable the generation of the FMA4 instructions.
3974 @cindex @code{target("xop")} attribute
3975 Enable/disable the generation of the XOP instructions.
3979 @cindex @code{target("lwp")} attribute
3980 Enable/disable the generation of the LWP instructions.
3984 @cindex @code{target("ssse3")} attribute
3985 Enable/disable the generation of the SSSE3 instructions.
3989 @cindex @code{target("cld")} attribute
3990 Enable/disable the generation of the CLD before string moves.
3992 @item fancy-math-387
3993 @itemx no-fancy-math-387
3994 @cindex @code{target("fancy-math-387")} attribute
3995 Enable/disable the generation of the @code{sin}, @code{cos}, and
3996 @code{sqrt} instructions on the 387 floating-point unit.
3999 @itemx no-fused-madd
4000 @cindex @code{target("fused-madd")} attribute
4001 Enable/disable the generation of the fused multiply/add instructions.
4005 @cindex @code{target("ieee-fp")} attribute
4006 Enable/disable the generation of floating point that depends on IEEE arithmetic.
4008 @item inline-all-stringops
4009 @itemx no-inline-all-stringops
4010 @cindex @code{target("inline-all-stringops")} attribute
4011 Enable/disable inlining of string operations.
4013 @item inline-stringops-dynamically
4014 @itemx no-inline-stringops-dynamically
4015 @cindex @code{target("inline-stringops-dynamically")} attribute
4016 Enable/disable the generation of the inline code to do small string
4017 operations and calling the library routines for large operations.
4019 @item align-stringops
4020 @itemx no-align-stringops
4021 @cindex @code{target("align-stringops")} attribute
4022 Do/do not align destination of inlined string operations.
4026 @cindex @code{target("recip")} attribute
4027 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
4028 instructions followed an additional Newton-Raphson step instead of
4029 doing a floating-point division.
4031 @item arch=@var{ARCH}
4032 @cindex @code{target("arch=@var{ARCH}")} attribute
4033 Specify the architecture to generate code for in compiling the function.
4035 @item tune=@var{TUNE}
4036 @cindex @code{target("tune=@var{TUNE}")} attribute
4037 Specify the architecture to tune for in compiling the function.
4039 @item fpmath=@var{FPMATH}
4040 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
4041 Specify which floating-point unit to use. The
4042 @code{target("fpmath=sse,387")} option must be specified as
4043 @code{target("fpmath=sse+387")} because the comma would separate
4047 On the PowerPC, the following options are allowed:
4052 @cindex @code{target("altivec")} attribute
4053 Generate code that uses (does not use) AltiVec instructions. In
4054 32-bit code, you cannot enable AltiVec instructions unless
4055 @option{-mabi=altivec} is used on the command line.
4059 @cindex @code{target("cmpb")} attribute
4060 Generate code that uses (does not use) the compare bytes instruction
4061 implemented on the POWER6 processor and other processors that support
4062 the PowerPC V2.05 architecture.
4066 @cindex @code{target("dlmzb")} attribute
4067 Generate code that uses (does not use) the string-search @samp{dlmzb}
4068 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4069 generated by default when targeting those processors.
4073 @cindex @code{target("fprnd")} attribute
4074 Generate code that uses (does not use) the FP round to integer
4075 instructions implemented on the POWER5+ processor and other processors
4076 that support the PowerPC V2.03 architecture.
4080 @cindex @code{target("hard-dfp")} attribute
4081 Generate code that uses (does not use) the decimal floating-point
4082 instructions implemented on some POWER processors.
4086 @cindex @code{target("isel")} attribute
4087 Generate code that uses (does not use) ISEL instruction.
4091 @cindex @code{target("mfcrf")} attribute
4092 Generate code that uses (does not use) the move from condition
4093 register field instruction implemented on the POWER4 processor and
4094 other processors that support the PowerPC V2.01 architecture.
4098 @cindex @code{target("mfpgpr")} attribute
4099 Generate code that uses (does not use) the FP move to/from general
4100 purpose register instructions implemented on the POWER6X processor and
4101 other processors that support the extended PowerPC V2.05 architecture.
4105 @cindex @code{target("mulhw")} attribute
4106 Generate code that uses (does not use) the half-word multiply and
4107 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4108 These instructions are generated by default when targeting those
4113 @cindex @code{target("multiple")} attribute
4114 Generate code that uses (does not use) the load multiple word
4115 instructions and the store multiple word instructions.
4119 @cindex @code{target("update")} attribute
4120 Generate code that uses (does not use) the load or store instructions
4121 that update the base register to the address of the calculated memory
4126 @cindex @code{target("popcntb")} attribute
4127 Generate code that uses (does not use) the popcount and double-precision
4128 FP reciprocal estimate instruction implemented on the POWER5
4129 processor and other processors that support the PowerPC V2.02
4134 @cindex @code{target("popcntd")} attribute
4135 Generate code that uses (does not use) the popcount instruction
4136 implemented on the POWER7 processor and other processors that support
4137 the PowerPC V2.06 architecture.
4139 @item powerpc-gfxopt
4140 @itemx no-powerpc-gfxopt
4141 @cindex @code{target("powerpc-gfxopt")} attribute
4142 Generate code that uses (does not use) the optional PowerPC
4143 architecture instructions in the Graphics group, including
4144 floating-point select.
4147 @itemx no-powerpc-gpopt
4148 @cindex @code{target("powerpc-gpopt")} attribute
4149 Generate code that uses (does not use) the optional PowerPC
4150 architecture instructions in the General Purpose group, including
4151 floating-point square root.
4153 @item recip-precision
4154 @itemx no-recip-precision
4155 @cindex @code{target("recip-precision")} attribute
4156 Assume (do not assume) that the reciprocal estimate instructions
4157 provide higher-precision estimates than is mandated by the powerpc
4162 @cindex @code{target("string")} attribute
4163 Generate code that uses (does not use) the load string instructions
4164 and the store string word instructions to save multiple registers and
4165 do small block moves.
4169 @cindex @code{target("vsx")} attribute
4170 Generate code that uses (does not use) vector/scalar (VSX)
4171 instructions, and also enable the use of built-in functions that allow
4172 more direct access to the VSX instruction set. In 32-bit code, you
4173 cannot enable VSX or AltiVec instructions unless
4174 @option{-mabi=altivec} is used on the command line.
4178 @cindex @code{target("friz")} attribute
4179 Generate (do not generate) the @code{friz} instruction when the
4180 @option{-funsafe-math-optimizations} option is used to optimize
4181 rounding a floating-point value to 64-bit integer and back to floating
4182 point. The @code{friz} instruction does not return the same value if
4183 the floating-point number is too large to fit in an integer.
4185 @item avoid-indexed-addresses
4186 @itemx no-avoid-indexed-addresses
4187 @cindex @code{target("avoid-indexed-addresses")} attribute
4188 Generate code that tries to avoid (not avoid) the use of indexed load
4189 or store instructions.
4193 @cindex @code{target("paired")} attribute
4194 Generate code that uses (does not use) the generation of PAIRED simd
4199 @cindex @code{target("longcall")} attribute
4200 Generate code that assumes (does not assume) that all calls are far
4201 away so that a longer more expensive calling sequence is required.
4204 @cindex @code{target("cpu=@var{CPU}")} attribute
4205 Specify the architecture to generate code for when compiling the
4206 function. If you select the @code{target("cpu=power7")} attribute when
4207 generating 32-bit code, VSX and AltiVec instructions are not generated
4208 unless you use the @option{-mabi=altivec} option on the command line.
4210 @item tune=@var{TUNE}
4211 @cindex @code{target("tune=@var{TUNE}")} attribute
4212 Specify the architecture to tune for when compiling the function. If
4213 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4214 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4215 compilation tunes for the @var{CPU} architecture, and not the
4216 default tuning specified on the command line.
4219 When compiling for Nios II, the following options are allowed:
4222 @item custom-@var{insn}=@var{N}
4223 @itemx no-custom-@var{insn}
4224 @cindex @code{target("custom-@var{insn}=@var{N}")} attribute
4225 @cindex @code{target("no-custom-@var{insn}")} attribute
4226 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4227 custom instruction with encoding @var{N} when generating code that uses
4228 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4229 the custom instruction @var{insn}.
4230 These target attributes correspond to the
4231 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4232 command-line options, and support the same set of @var{insn} keywords.
4233 @xref{Nios II Options}, for more information.
4235 @item custom-fpu-cfg=@var{name}
4236 @cindex @code{target("custom-fpu-cfg=@var{name}")} attribute
4237 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4238 command-line option, to select a predefined set of custom instructions
4240 @xref{Nios II Options}, for more information.
4243 On the 386/x86_64 and PowerPC back ends, the inliner does not inline a
4244 function that has different target options than the caller, unless the
4245 callee has a subset of the target options of the caller. For example
4246 a function declared with @code{target("sse3")} can inline a function
4247 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
4250 @cindex tiny data section on the H8/300H and H8S
4251 Use this attribute on the H8/300H and H8S to indicate that the specified
4252 variable should be placed into the tiny data section.
4253 The compiler generates more efficient code for loads and stores
4254 on data in the tiny data section. Note the tiny data area is limited to
4255 slightly under 32KB of data.
4258 @cindex @code{trap_exit} attribute
4259 Use this attribute on the SH for an @code{interrupt_handler} to return using
4260 @code{trapa} instead of @code{rte}. This attribute expects an integer
4261 argument specifying the trap number to be used.
4264 @cindex @code{trapa_handler} attribute
4265 On SH targets this function attribute is similar to @code{interrupt_handler}
4266 but it does not save and restore all registers.
4269 @cindex @code{unused} attribute.
4270 This attribute, attached to a function, means that the function is meant
4271 to be possibly unused. GCC does not produce a warning for this
4275 @cindex @code{used} attribute.
4276 This attribute, attached to a function, means that code must be emitted
4277 for the function even if it appears that the function is not referenced.
4278 This is useful, for example, when the function is referenced only in
4281 When applied to a member function of a C++ class template, the
4282 attribute also means that the function is instantiated if the
4283 class itself is instantiated.
4286 @cindex @code{version_id} attribute
4287 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4288 symbol to contain a version string, thus allowing for function level
4289 versioning. HP-UX system header files may use function level versioning
4290 for some system calls.
4293 extern int foo () __attribute__((version_id ("20040821")));
4297 Calls to @var{foo} are mapped to calls to @var{foo@{20040821@}}.
4299 @item visibility ("@var{visibility_type}")
4300 @cindex @code{visibility} attribute
4301 This attribute affects the linkage of the declaration to which it is attached.
4302 There are four supported @var{visibility_type} values: default,
4303 hidden, protected or internal visibility.
4306 void __attribute__ ((visibility ("protected")))
4307 f () @{ /* @r{Do something.} */; @}
4308 int i __attribute__ ((visibility ("hidden")));
4311 The possible values of @var{visibility_type} correspond to the
4312 visibility settings in the ELF gABI.
4315 @c keep this list of visibilities in alphabetical order.
4318 Default visibility is the normal case for the object file format.
4319 This value is available for the visibility attribute to override other
4320 options that may change the assumed visibility of entities.
4322 On ELF, default visibility means that the declaration is visible to other
4323 modules and, in shared libraries, means that the declared entity may be
4326 On Darwin, default visibility means that the declaration is visible to
4329 Default visibility corresponds to ``external linkage'' in the language.
4332 Hidden visibility indicates that the entity declared has a new
4333 form of linkage, which we call ``hidden linkage''. Two
4334 declarations of an object with hidden linkage refer to the same object
4335 if they are in the same shared object.
4338 Internal visibility is like hidden visibility, but with additional
4339 processor specific semantics. Unless otherwise specified by the
4340 psABI, GCC defines internal visibility to mean that a function is
4341 @emph{never} called from another module. Compare this with hidden
4342 functions which, while they cannot be referenced directly by other
4343 modules, can be referenced indirectly via function pointers. By
4344 indicating that a function cannot be called from outside the module,
4345 GCC may for instance omit the load of a PIC register since it is known
4346 that the calling function loaded the correct value.
4349 Protected visibility is like default visibility except that it
4350 indicates that references within the defining module bind to the
4351 definition in that module. That is, the declared entity cannot be
4352 overridden by another module.
4356 All visibilities are supported on many, but not all, ELF targets
4357 (supported when the assembler supports the @samp{.visibility}
4358 pseudo-op). Default visibility is supported everywhere. Hidden
4359 visibility is supported on Darwin targets.
4361 The visibility attribute should be applied only to declarations that
4362 would otherwise have external linkage. The attribute should be applied
4363 consistently, so that the same entity should not be declared with
4364 different settings of the attribute.
4366 In C++, the visibility attribute applies to types as well as functions
4367 and objects, because in C++ types have linkage. A class must not have
4368 greater visibility than its non-static data member types and bases,
4369 and class members default to the visibility of their class. Also, a
4370 declaration without explicit visibility is limited to the visibility
4373 In C++, you can mark member functions and static member variables of a
4374 class with the visibility attribute. This is useful if you know a
4375 particular method or static member variable should only be used from
4376 one shared object; then you can mark it hidden while the rest of the
4377 class has default visibility. Care must be taken to avoid breaking
4378 the One Definition Rule; for example, it is usually not useful to mark
4379 an inline method as hidden without marking the whole class as hidden.
4381 A C++ namespace declaration can also have the visibility attribute.
4384 namespace nspace1 __attribute__ ((visibility ("protected")))
4385 @{ /* @r{Do something.} */; @}
4388 This attribute applies only to the particular namespace body, not to
4389 other definitions of the same namespace; it is equivalent to using
4390 @samp{#pragma GCC visibility} before and after the namespace
4391 definition (@pxref{Visibility Pragmas}).
4393 In C++, if a template argument has limited visibility, this
4394 restriction is implicitly propagated to the template instantiation.
4395 Otherwise, template instantiations and specializations default to the
4396 visibility of their template.
4398 If both the template and enclosing class have explicit visibility, the
4399 visibility from the template is used.
4402 @cindex @code{vliw} attribute
4403 On MeP, the @code{vliw} attribute tells the compiler to emit
4404 instructions in VLIW mode instead of core mode. Note that this
4405 attribute is not allowed unless a VLIW coprocessor has been configured
4406 and enabled through command-line options.
4408 @item warn_unused_result
4409 @cindex @code{warn_unused_result} attribute
4410 The @code{warn_unused_result} attribute causes a warning to be emitted
4411 if a caller of the function with this attribute does not use its
4412 return value. This is useful for functions where not checking
4413 the result is either a security problem or always a bug, such as
4417 int fn () __attribute__ ((warn_unused_result));
4420 if (fn () < 0) return -1;
4427 results in warning on line 5.
4430 @cindex @code{weak} attribute
4431 The @code{weak} attribute causes the declaration to be emitted as a weak
4432 symbol rather than a global. This is primarily useful in defining
4433 library functions that can be overridden in user code, though it can
4434 also be used with non-function declarations. Weak symbols are supported
4435 for ELF targets, and also for a.out targets when using the GNU assembler
4439 @itemx weakref ("@var{target}")
4440 @cindex @code{weakref} attribute
4441 The @code{weakref} attribute marks a declaration as a weak reference.
4442 Without arguments, it should be accompanied by an @code{alias} attribute
4443 naming the target symbol. Optionally, the @var{target} may be given as
4444 an argument to @code{weakref} itself. In either case, @code{weakref}
4445 implicitly marks the declaration as @code{weak}. Without a
4446 @var{target}, given as an argument to @code{weakref} or to @code{alias},
4447 @code{weakref} is equivalent to @code{weak}.
4450 static int x() __attribute__ ((weakref ("y")));
4451 /* is equivalent to... */
4452 static int x() __attribute__ ((weak, weakref, alias ("y")));
4454 static int x() __attribute__ ((weakref));
4455 static int x() __attribute__ ((alias ("y")));
4458 A weak reference is an alias that does not by itself require a
4459 definition to be given for the target symbol. If the target symbol is
4460 only referenced through weak references, then it becomes a @code{weak}
4461 undefined symbol. If it is directly referenced, however, then such
4462 strong references prevail, and a definition is required for the
4463 symbol, not necessarily in the same translation unit.
4465 The effect is equivalent to moving all references to the alias to a
4466 separate translation unit, renaming the alias to the aliased symbol,
4467 declaring it as weak, compiling the two separate translation units and
4468 performing a reloadable link on them.
4470 At present, a declaration to which @code{weakref} is attached can
4471 only be @code{static}.
4475 You can specify multiple attributes in a declaration by separating them
4476 by commas within the double parentheses or by immediately following an
4477 attribute declaration with another attribute declaration.
4479 @cindex @code{#pragma}, reason for not using
4480 @cindex pragma, reason for not using
4481 Some people object to the @code{__attribute__} feature, suggesting that
4482 ISO C's @code{#pragma} should be used instead. At the time
4483 @code{__attribute__} was designed, there were two reasons for not doing
4488 It is impossible to generate @code{#pragma} commands from a macro.
4491 There is no telling what the same @code{#pragma} might mean in another
4495 These two reasons applied to almost any application that might have been
4496 proposed for @code{#pragma}. It was basically a mistake to use
4497 @code{#pragma} for @emph{anything}.
4499 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
4500 to be generated from macros. In addition, a @code{#pragma GCC}
4501 namespace is now in use for GCC-specific pragmas. However, it has been
4502 found convenient to use @code{__attribute__} to achieve a natural
4503 attachment of attributes to their corresponding declarations, whereas
4504 @code{#pragma GCC} is of use for constructs that do not naturally form
4505 part of the grammar. @xref{Pragmas,,Pragmas Accepted by GCC}.
4507 @node Attribute Syntax
4508 @section Attribute Syntax
4509 @cindex attribute syntax
4511 This section describes the syntax with which @code{__attribute__} may be
4512 used, and the constructs to which attribute specifiers bind, for the C
4513 language. Some details may vary for C++ and Objective-C@. Because of
4514 infelicities in the grammar for attributes, some forms described here
4515 may not be successfully parsed in all cases.
4517 There are some problems with the semantics of attributes in C++. For
4518 example, there are no manglings for attributes, although they may affect
4519 code generation, so problems may arise when attributed types are used in
4520 conjunction with templates or overloading. Similarly, @code{typeid}
4521 does not distinguish between types with different attributes. Support
4522 for attributes in C++ may be restricted in future to attributes on
4523 declarations only, but not on nested declarators.
4525 @xref{Function Attributes}, for details of the semantics of attributes
4526 applying to functions. @xref{Variable Attributes}, for details of the
4527 semantics of attributes applying to variables. @xref{Type Attributes},
4528 for details of the semantics of attributes applying to structure, union
4529 and enumerated types.
4531 An @dfn{attribute specifier} is of the form
4532 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
4533 is a possibly empty comma-separated sequence of @dfn{attributes}, where
4534 each attribute is one of the following:
4538 Empty. Empty attributes are ignored.
4541 A word (which may be an identifier such as @code{unused}, or a reserved
4542 word such as @code{const}).
4545 A word, followed by, in parentheses, parameters for the attribute.
4546 These parameters take one of the following forms:
4550 An identifier. For example, @code{mode} attributes use this form.
4553 An identifier followed by a comma and a non-empty comma-separated list
4554 of expressions. For example, @code{format} attributes use this form.
4557 A possibly empty comma-separated list of expressions. For example,
4558 @code{format_arg} attributes use this form with the list being a single
4559 integer constant expression, and @code{alias} attributes use this form
4560 with the list being a single string constant.
4564 An @dfn{attribute specifier list} is a sequence of one or more attribute
4565 specifiers, not separated by any other tokens.
4567 In GNU C, an attribute specifier list may appear after the colon following a
4568 label, other than a @code{case} or @code{default} label. The only
4569 attribute it makes sense to use after a label is @code{unused}. This
4570 feature is intended for program-generated code that may contain unused labels,
4571 but which is compiled with @option{-Wall}. It is
4572 not normally appropriate to use in it human-written code, though it
4573 could be useful in cases where the code that jumps to the label is
4574 contained within an @code{#ifdef} conditional. GNU C++ only permits
4575 attributes on labels if the attribute specifier is immediately
4576 followed by a semicolon (i.e., the label applies to an empty
4577 statement). If the semicolon is missing, C++ label attributes are
4578 ambiguous, as it is permissible for a declaration, which could begin
4579 with an attribute list, to be labelled in C++. Declarations cannot be
4580 labelled in C90 or C99, so the ambiguity does not arise there.
4582 An attribute specifier list may appear as part of a @code{struct},
4583 @code{union} or @code{enum} specifier. It may go either immediately
4584 after the @code{struct}, @code{union} or @code{enum} keyword, or after
4585 the closing brace. The former syntax is preferred.
4586 Where attribute specifiers follow the closing brace, they are considered
4587 to relate to the structure, union or enumerated type defined, not to any
4588 enclosing declaration the type specifier appears in, and the type
4589 defined is not complete until after the attribute specifiers.
4590 @c Otherwise, there would be the following problems: a shift/reduce
4591 @c conflict between attributes binding the struct/union/enum and
4592 @c binding to the list of specifiers/qualifiers; and "aligned"
4593 @c attributes could use sizeof for the structure, but the size could be
4594 @c changed later by "packed" attributes.
4596 Otherwise, an attribute specifier appears as part of a declaration,
4597 counting declarations of unnamed parameters and type names, and relates
4598 to that declaration (which may be nested in another declaration, for
4599 example in the case of a parameter declaration), or to a particular declarator
4600 within a declaration. Where an
4601 attribute specifier is applied to a parameter declared as a function or
4602 an array, it should apply to the function or array rather than the
4603 pointer to which the parameter is implicitly converted, but this is not
4604 yet correctly implemented.
4606 Any list of specifiers and qualifiers at the start of a declaration may
4607 contain attribute specifiers, whether or not such a list may in that
4608 context contain storage class specifiers. (Some attributes, however,
4609 are essentially in the nature of storage class specifiers, and only make
4610 sense where storage class specifiers may be used; for example,
4611 @code{section}.) There is one necessary limitation to this syntax: the
4612 first old-style parameter declaration in a function definition cannot
4613 begin with an attribute specifier, because such an attribute applies to
4614 the function instead by syntax described below (which, however, is not
4615 yet implemented in this case). In some other cases, attribute
4616 specifiers are permitted by this grammar but not yet supported by the
4617 compiler. All attribute specifiers in this place relate to the
4618 declaration as a whole. In the obsolescent usage where a type of
4619 @code{int} is implied by the absence of type specifiers, such a list of
4620 specifiers and qualifiers may be an attribute specifier list with no
4621 other specifiers or qualifiers.
4623 At present, the first parameter in a function prototype must have some
4624 type specifier that is not an attribute specifier; this resolves an
4625 ambiguity in the interpretation of @code{void f(int
4626 (__attribute__((foo)) x))}, but is subject to change. At present, if
4627 the parentheses of a function declarator contain only attributes then
4628 those attributes are ignored, rather than yielding an error or warning
4629 or implying a single parameter of type int, but this is subject to
4632 An attribute specifier list may appear immediately before a declarator
4633 (other than the first) in a comma-separated list of declarators in a
4634 declaration of more than one identifier using a single list of
4635 specifiers and qualifiers. Such attribute specifiers apply
4636 only to the identifier before whose declarator they appear. For
4640 __attribute__((noreturn)) void d0 (void),
4641 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
4646 the @code{noreturn} attribute applies to all the functions
4647 declared; the @code{format} attribute only applies to @code{d1}.
4649 An attribute specifier list may appear immediately before the comma,
4650 @code{=} or semicolon terminating the declaration of an identifier other
4651 than a function definition. Such attribute specifiers apply
4652 to the declared object or function. Where an
4653 assembler name for an object or function is specified (@pxref{Asm
4654 Labels}), the attribute must follow the @code{asm}
4657 An attribute specifier list may, in future, be permitted to appear after
4658 the declarator in a function definition (before any old-style parameter
4659 declarations or the function body).
4661 Attribute specifiers may be mixed with type qualifiers appearing inside
4662 the @code{[]} of a parameter array declarator, in the C99 construct by
4663 which such qualifiers are applied to the pointer to which the array is
4664 implicitly converted. Such attribute specifiers apply to the pointer,
4665 not to the array, but at present this is not implemented and they are
4668 An attribute specifier list may appear at the start of a nested
4669 declarator. At present, there are some limitations in this usage: the
4670 attributes correctly apply to the declarator, but for most individual
4671 attributes the semantics this implies are not implemented.
4672 When attribute specifiers follow the @code{*} of a pointer
4673 declarator, they may be mixed with any type qualifiers present.
4674 The following describes the formal semantics of this syntax. It makes the
4675 most sense if you are familiar with the formal specification of
4676 declarators in the ISO C standard.
4678 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
4679 D1}, where @code{T} contains declaration specifiers that specify a type
4680 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
4681 contains an identifier @var{ident}. The type specified for @var{ident}
4682 for derived declarators whose type does not include an attribute
4683 specifier is as in the ISO C standard.
4685 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
4686 and the declaration @code{T D} specifies the type
4687 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4688 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4689 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
4691 If @code{D1} has the form @code{*
4692 @var{type-qualifier-and-attribute-specifier-list} D}, and the
4693 declaration @code{T D} specifies the type
4694 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4695 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4696 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
4702 void (__attribute__((noreturn)) ****f) (void);
4706 specifies the type ``pointer to pointer to pointer to pointer to
4707 non-returning function returning @code{void}''. As another example,
4710 char *__attribute__((aligned(8))) *f;
4714 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
4715 Note again that this does not work with most attributes; for example,
4716 the usage of @samp{aligned} and @samp{noreturn} attributes given above
4717 is not yet supported.
4719 For compatibility with existing code written for compiler versions that
4720 did not implement attributes on nested declarators, some laxity is
4721 allowed in the placing of attributes. If an attribute that only applies
4722 to types is applied to a declaration, it is treated as applying to
4723 the type of that declaration. If an attribute that only applies to
4724 declarations is applied to the type of a declaration, it is treated
4725 as applying to that declaration; and, for compatibility with code
4726 placing the attributes immediately before the identifier declared, such
4727 an attribute applied to a function return type is treated as
4728 applying to the function type, and such an attribute applied to an array
4729 element type is treated as applying to the array type. If an
4730 attribute that only applies to function types is applied to a
4731 pointer-to-function type, it is treated as applying to the pointer
4732 target type; if such an attribute is applied to a function return type
4733 that is not a pointer-to-function type, it is treated as applying
4734 to the function type.
4736 @node Function Prototypes
4737 @section Prototypes and Old-Style Function Definitions
4738 @cindex function prototype declarations
4739 @cindex old-style function definitions
4740 @cindex promotion of formal parameters
4742 GNU C extends ISO C to allow a function prototype to override a later
4743 old-style non-prototype definition. Consider the following example:
4746 /* @r{Use prototypes unless the compiler is old-fashioned.} */
4753 /* @r{Prototype function declaration.} */
4754 int isroot P((uid_t));
4756 /* @r{Old-style function definition.} */
4758 isroot (x) /* @r{??? lossage here ???} */
4765 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
4766 not allow this example, because subword arguments in old-style
4767 non-prototype definitions are promoted. Therefore in this example the
4768 function definition's argument is really an @code{int}, which does not
4769 match the prototype argument type of @code{short}.
4771 This restriction of ISO C makes it hard to write code that is portable
4772 to traditional C compilers, because the programmer does not know
4773 whether the @code{uid_t} type is @code{short}, @code{int}, or
4774 @code{long}. Therefore, in cases like these GNU C allows a prototype
4775 to override a later old-style definition. More precisely, in GNU C, a
4776 function prototype argument type overrides the argument type specified
4777 by a later old-style definition if the former type is the same as the
4778 latter type before promotion. Thus in GNU C the above example is
4779 equivalent to the following:
4792 GNU C++ does not support old-style function definitions, so this
4793 extension is irrelevant.
4796 @section C++ Style Comments
4798 @cindex C++ comments
4799 @cindex comments, C++ style
4801 In GNU C, you may use C++ style comments, which start with @samp{//} and
4802 continue until the end of the line. Many other C implementations allow
4803 such comments, and they are included in the 1999 C standard. However,
4804 C++ style comments are not recognized if you specify an @option{-std}
4805 option specifying a version of ISO C before C99, or @option{-ansi}
4806 (equivalent to @option{-std=c90}).
4809 @section Dollar Signs in Identifier Names
4811 @cindex dollar signs in identifier names
4812 @cindex identifier names, dollar signs in
4814 In GNU C, you may normally use dollar signs in identifier names.
4815 This is because many traditional C implementations allow such identifiers.
4816 However, dollar signs in identifiers are not supported on a few target
4817 machines, typically because the target assembler does not allow them.
4819 @node Character Escapes
4820 @section The Character @key{ESC} in Constants
4822 You can use the sequence @samp{\e} in a string or character constant to
4823 stand for the ASCII character @key{ESC}.
4825 @node Variable Attributes
4826 @section Specifying Attributes of Variables
4827 @cindex attribute of variables
4828 @cindex variable attributes
4830 The keyword @code{__attribute__} allows you to specify special
4831 attributes of variables or structure fields. This keyword is followed
4832 by an attribute specification inside double parentheses. Some
4833 attributes are currently defined generically for variables.
4834 Other attributes are defined for variables on particular target
4835 systems. Other attributes are available for functions
4836 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
4837 Other front ends might define more attributes
4838 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
4840 You may also specify attributes with @samp{__} preceding and following
4841 each keyword. This allows you to use them in header files without
4842 being concerned about a possible macro of the same name. For example,
4843 you may use @code{__aligned__} instead of @code{aligned}.
4845 @xref{Attribute Syntax}, for details of the exact syntax for using
4849 @cindex @code{aligned} attribute
4850 @item aligned (@var{alignment})
4851 This attribute specifies a minimum alignment for the variable or
4852 structure field, measured in bytes. For example, the declaration:
4855 int x __attribute__ ((aligned (16))) = 0;
4859 causes the compiler to allocate the global variable @code{x} on a
4860 16-byte boundary. On a 68040, this could be used in conjunction with
4861 an @code{asm} expression to access the @code{move16} instruction which
4862 requires 16-byte aligned operands.
4864 You can also specify the alignment of structure fields. For example, to
4865 create a double-word aligned @code{int} pair, you could write:
4868 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
4872 This is an alternative to creating a union with a @code{double} member,
4873 which forces the union to be double-word aligned.
4875 As in the preceding examples, you can explicitly specify the alignment
4876 (in bytes) that you wish the compiler to use for a given variable or
4877 structure field. Alternatively, you can leave out the alignment factor
4878 and just ask the compiler to align a variable or field to the
4879 default alignment for the target architecture you are compiling for.
4880 The default alignment is sufficient for all scalar types, but may not be
4881 enough for all vector types on a target that supports vector operations.
4882 The default alignment is fixed for a particular target ABI.
4884 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
4885 which is the largest alignment ever used for any data type on the
4886 target machine you are compiling for. For example, you could write:
4889 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
4892 The compiler automatically sets the alignment for the declared
4893 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
4894 often make copy operations more efficient, because the compiler can
4895 use whatever instructions copy the biggest chunks of memory when
4896 performing copies to or from the variables or fields that you have
4897 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
4898 may change depending on command-line options.
4900 When used on a struct, or struct member, the @code{aligned} attribute can
4901 only increase the alignment; in order to decrease it, the @code{packed}
4902 attribute must be specified as well. When used as part of a typedef, the
4903 @code{aligned} attribute can both increase and decrease alignment, and
4904 specifying the @code{packed} attribute generates a warning.
4906 Note that the effectiveness of @code{aligned} attributes may be limited
4907 by inherent limitations in your linker. On many systems, the linker is
4908 only able to arrange for variables to be aligned up to a certain maximum
4909 alignment. (For some linkers, the maximum supported alignment may
4910 be very very small.) If your linker is only able to align variables
4911 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
4912 in an @code{__attribute__} still only provides you with 8-byte
4913 alignment. See your linker documentation for further information.
4915 The @code{aligned} attribute can also be used for functions
4916 (@pxref{Function Attributes}.)
4918 @item cleanup (@var{cleanup_function})
4919 @cindex @code{cleanup} attribute
4920 The @code{cleanup} attribute runs a function when the variable goes
4921 out of scope. This attribute can only be applied to auto function
4922 scope variables; it may not be applied to parameters or variables
4923 with static storage duration. The function must take one parameter,
4924 a pointer to a type compatible with the variable. The return value
4925 of the function (if any) is ignored.
4927 If @option{-fexceptions} is enabled, then @var{cleanup_function}
4928 is run during the stack unwinding that happens during the
4929 processing of the exception. Note that the @code{cleanup} attribute
4930 does not allow the exception to be caught, only to perform an action.
4931 It is undefined what happens if @var{cleanup_function} does not
4936 @cindex @code{common} attribute
4937 @cindex @code{nocommon} attribute
4940 The @code{common} attribute requests GCC to place a variable in
4941 ``common'' storage. The @code{nocommon} attribute requests the
4942 opposite---to allocate space for it directly.
4944 These attributes override the default chosen by the
4945 @option{-fno-common} and @option{-fcommon} flags respectively.
4948 @itemx deprecated (@var{msg})
4949 @cindex @code{deprecated} attribute
4950 The @code{deprecated} attribute results in a warning if the variable
4951 is used anywhere in the source file. This is useful when identifying
4952 variables that are expected to be removed in a future version of a
4953 program. The warning also includes the location of the declaration
4954 of the deprecated variable, to enable users to easily find further
4955 information about why the variable is deprecated, or what they should
4956 do instead. Note that the warning only occurs for uses:
4959 extern int old_var __attribute__ ((deprecated));
4961 int new_fn () @{ return old_var; @}
4965 results in a warning on line 3 but not line 2. The optional @var{msg}
4966 argument, which must be a string, is printed in the warning if
4969 The @code{deprecated} attribute can also be used for functions and
4970 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
4972 @item mode (@var{mode})
4973 @cindex @code{mode} attribute
4974 This attribute specifies the data type for the declaration---whichever
4975 type corresponds to the mode @var{mode}. This in effect lets you
4976 request an integer or floating-point type according to its width.
4978 You may also specify a mode of @code{byte} or @code{__byte__} to
4979 indicate the mode corresponding to a one-byte integer, @code{word} or
4980 @code{__word__} for the mode of a one-word integer, and @code{pointer}
4981 or @code{__pointer__} for the mode used to represent pointers.
4984 @cindex @code{packed} attribute
4985 The @code{packed} attribute specifies that a variable or structure field
4986 should have the smallest possible alignment---one byte for a variable,
4987 and one bit for a field, unless you specify a larger value with the
4988 @code{aligned} attribute.
4990 Here is a structure in which the field @code{x} is packed, so that it
4991 immediately follows @code{a}:
4997 int x[2] __attribute__ ((packed));
5001 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5002 @code{packed} attribute on bit-fields of type @code{char}. This has
5003 been fixed in GCC 4.4 but the change can lead to differences in the
5004 structure layout. See the documentation of
5005 @option{-Wpacked-bitfield-compat} for more information.
5007 @item section ("@var{section-name}")
5008 @cindex @code{section} variable attribute
5009 Normally, the compiler places the objects it generates in sections like
5010 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5011 or you need certain particular variables to appear in special sections,
5012 for example to map to special hardware. The @code{section}
5013 attribute specifies that a variable (or function) lives in a particular
5014 section. For example, this small program uses several specific section names:
5017 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5018 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5019 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5020 int init_data __attribute__ ((section ("INITDATA")));
5024 /* @r{Initialize stack pointer} */
5025 init_sp (stack + sizeof (stack));
5027 /* @r{Initialize initialized data} */
5028 memcpy (&init_data, &data, &edata - &data);
5030 /* @r{Turn on the serial ports} */
5037 Use the @code{section} attribute with
5038 @emph{global} variables and not @emph{local} variables,
5039 as shown in the example.
5041 You may use the @code{section} attribute with initialized or
5042 uninitialized global variables but the linker requires
5043 each object be defined once, with the exception that uninitialized
5044 variables tentatively go in the @code{common} (or @code{bss}) section
5045 and can be multiply ``defined''. Using the @code{section} attribute
5046 changes what section the variable goes into and may cause the
5047 linker to issue an error if an uninitialized variable has multiple
5048 definitions. You can force a variable to be initialized with the
5049 @option{-fno-common} flag or the @code{nocommon} attribute.
5051 Some file formats do not support arbitrary sections so the @code{section}
5052 attribute is not available on all platforms.
5053 If you need to map the entire contents of a module to a particular
5054 section, consider using the facilities of the linker instead.
5057 @cindex @code{shared} variable attribute
5058 On Microsoft Windows, in addition to putting variable definitions in a named
5059 section, the section can also be shared among all running copies of an
5060 executable or DLL@. For example, this small program defines shared data
5061 by putting it in a named section @code{shared} and marking the section
5065 int foo __attribute__((section ("shared"), shared)) = 0;
5070 /* @r{Read and write foo. All running
5071 copies see the same value.} */
5077 You may only use the @code{shared} attribute along with @code{section}
5078 attribute with a fully-initialized global definition because of the way
5079 linkers work. See @code{section} attribute for more information.
5081 The @code{shared} attribute is only available on Microsoft Windows@.
5083 @item tls_model ("@var{tls_model}")
5084 @cindex @code{tls_model} attribute
5085 The @code{tls_model} attribute sets thread-local storage model
5086 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5087 overriding @option{-ftls-model=} command-line switch on a per-variable
5089 The @var{tls_model} argument should be one of @code{global-dynamic},
5090 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5092 Not all targets support this attribute.
5095 This attribute, attached to a variable, means that the variable is meant
5096 to be possibly unused. GCC does not produce a warning for this
5100 This attribute, attached to a variable with the static storage, means that
5101 the variable must be emitted even if it appears that the variable is not
5104 When applied to a static data member of a C++ class template, the
5105 attribute also means that the member is instantiated if the
5106 class itself is instantiated.
5108 @item vector_size (@var{bytes})
5109 This attribute specifies the vector size for the variable, measured in
5110 bytes. For example, the declaration:
5113 int foo __attribute__ ((vector_size (16)));
5117 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5118 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5119 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5121 This attribute is only applicable to integral and float scalars,
5122 although arrays, pointers, and function return values are allowed in
5123 conjunction with this construct.
5125 Aggregates with this attribute are invalid, even if they are of the same
5126 size as a corresponding scalar. For example, the declaration:
5129 struct S @{ int a; @};
5130 struct S __attribute__ ((vector_size (16))) foo;
5134 is invalid even if the size of the structure is the same as the size of
5138 The @code{selectany} attribute causes an initialized global variable to
5139 have link-once semantics. When multiple definitions of the variable are
5140 encountered by the linker, the first is selected and the remainder are
5141 discarded. Following usage by the Microsoft compiler, the linker is told
5142 @emph{not} to warn about size or content differences of the multiple
5145 Although the primary usage of this attribute is for POD types, the
5146 attribute can also be applied to global C++ objects that are initialized
5147 by a constructor. In this case, the static initialization and destruction
5148 code for the object is emitted in each translation defining the object,
5149 but the calls to the constructor and destructor are protected by a
5150 link-once guard variable.
5152 The @code{selectany} attribute is only available on Microsoft Windows
5153 targets. You can use @code{__declspec (selectany)} as a synonym for
5154 @code{__attribute__ ((selectany))} for compatibility with other
5158 The @code{weak} attribute is described in @ref{Function Attributes}.
5161 The @code{dllimport} attribute is described in @ref{Function Attributes}.
5164 The @code{dllexport} attribute is described in @ref{Function Attributes}.
5168 @anchor{AVR Variable Attributes}
5169 @subsection AVR Variable Attributes
5173 @cindex @code{progmem} AVR variable attribute
5174 The @code{progmem} attribute is used on the AVR to place read-only
5175 data in the non-volatile program memory (flash). The @code{progmem}
5176 attribute accomplishes this by putting respective variables into a
5177 section whose name starts with @code{.progmem}.
5179 This attribute works similar to the @code{section} attribute
5180 but adds additional checking. Notice that just like the
5181 @code{section} attribute, @code{progmem} affects the location
5182 of the data but not how this data is accessed.
5184 In order to read data located with the @code{progmem} attribute
5185 (inline) assembler must be used.
5187 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5188 #include <avr/pgmspace.h>
5190 /* Locate var in flash memory */
5191 const int var[2] PROGMEM = @{ 1, 2 @};
5193 int read_var (int i)
5195 /* Access var[] by accessor macro from avr/pgmspace.h */
5196 return (int) pgm_read_word (& var[i]);
5200 AVR is a Harvard architecture processor and data and read-only data
5201 normally resides in the data memory (RAM).
5203 See also the @ref{AVR Named Address Spaces} section for
5204 an alternate way to locate and access data in flash memory.
5207 @subsection Blackfin Variable Attributes
5209 Three attributes are currently defined for the Blackfin.
5215 @cindex @code{l1_data} variable attribute
5216 @cindex @code{l1_data_A} variable attribute
5217 @cindex @code{l1_data_B} variable attribute
5218 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5219 Variables with @code{l1_data} attribute are put into the specific section
5220 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5221 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5222 attribute are put into the specific section named @code{.l1.data.B}.
5225 @cindex @code{l2} variable attribute
5226 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5227 Variables with @code{l2} attribute are put into the specific section
5228 named @code{.l2.data}.
5231 @subsection M32R/D Variable Attributes
5233 One attribute is currently defined for the M32R/D@.
5236 @item model (@var{model-name})
5237 @cindex variable addressability on the M32R/D
5238 Use this attribute on the M32R/D to set the addressability of an object.
5239 The identifier @var{model-name} is one of @code{small}, @code{medium},
5240 or @code{large}, representing each of the code models.
5242 Small model objects live in the lower 16MB of memory (so that their
5243 addresses can be loaded with the @code{ld24} instruction).
5245 Medium and large model objects may live anywhere in the 32-bit address space
5246 (the compiler generates @code{seth/add3} instructions to load their
5250 @anchor{MeP Variable Attributes}
5251 @subsection MeP Variable Attributes
5253 The MeP target has a number of addressing modes and busses. The
5254 @code{near} space spans the standard memory space's first 16 megabytes
5255 (24 bits). The @code{far} space spans the entire 32-bit memory space.
5256 The @code{based} space is a 128-byte region in the memory space that
5257 is addressed relative to the @code{$tp} register. The @code{tiny}
5258 space is a 65536-byte region relative to the @code{$gp} register. In
5259 addition to these memory regions, the MeP target has a separate 16-bit
5260 control bus which is specified with @code{cb} attributes.
5265 Any variable with the @code{based} attribute is assigned to the
5266 @code{.based} section, and is accessed with relative to the
5267 @code{$tp} register.
5270 Likewise, the @code{tiny} attribute assigned variables to the
5271 @code{.tiny} section, relative to the @code{$gp} register.
5274 Variables with the @code{near} attribute are assumed to have addresses
5275 that fit in a 24-bit addressing mode. This is the default for large
5276 variables (@code{-mtiny=4} is the default) but this attribute can
5277 override @code{-mtiny=} for small variables, or override @code{-ml}.
5280 Variables with the @code{far} attribute are addressed using a full
5281 32-bit address. Since this covers the entire memory space, this
5282 allows modules to make no assumptions about where variables might be
5286 @itemx io (@var{addr})
5287 Variables with the @code{io} attribute are used to address
5288 memory-mapped peripherals. If an address is specified, the variable
5289 is assigned that address, else it is not assigned an address (it is
5290 assumed some other module assigns an address). Example:
5293 int timer_count __attribute__((io(0x123)));
5297 @itemx cb (@var{addr})
5298 Variables with the @code{cb} attribute are used to access the control
5299 bus, using special instructions. @code{addr} indicates the control bus
5303 int cpu_clock __attribute__((cb(0x123)));
5308 @anchor{i386 Variable Attributes}
5309 @subsection i386 Variable Attributes
5311 Two attributes are currently defined for i386 configurations:
5312 @code{ms_struct} and @code{gcc_struct}
5317 @cindex @code{ms_struct} attribute
5318 @cindex @code{gcc_struct} attribute
5320 If @code{packed} is used on a structure, or if bit-fields are used,
5321 it may be that the Microsoft ABI lays out the structure differently
5322 than the way GCC normally does. Particularly when moving packed
5323 data between functions compiled with GCC and the native Microsoft compiler
5324 (either via function call or as data in a file), it may be necessary to access
5327 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
5328 compilers to match the native Microsoft compiler.
5330 The Microsoft structure layout algorithm is fairly simple with the exception
5331 of the bit-field packing.
5332 The padding and alignment of members of structures and whether a bit-field
5333 can straddle a storage-unit boundary are determine by these rules:
5336 @item Structure members are stored sequentially in the order in which they are
5337 declared: the first member has the lowest memory address and the last member
5340 @item Every data object has an alignment requirement. The alignment requirement
5341 for all data except structures, unions, and arrays is either the size of the
5342 object or the current packing size (specified with either the
5343 @code{aligned} attribute or the @code{pack} pragma),
5344 whichever is less. For structures, unions, and arrays,
5345 the alignment requirement is the largest alignment requirement of its members.
5346 Every object is allocated an offset so that:
5349 offset % alignment_requirement == 0
5352 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
5353 unit if the integral types are the same size and if the next bit-field fits
5354 into the current allocation unit without crossing the boundary imposed by the
5355 common alignment requirements of the bit-fields.
5358 MSVC interprets zero-length bit-fields in the following ways:
5361 @item If a zero-length bit-field is inserted between two bit-fields that
5362 are normally coalesced, the bit-fields are not coalesced.
5369 unsigned long bf_1 : 12;
5371 unsigned long bf_2 : 12;
5376 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
5377 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
5379 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
5380 alignment of the zero-length bit-field is greater than the member that follows it,
5381 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
5402 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
5403 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
5404 bit-field does not affect the alignment of @code{bar} or, as a result, the size
5407 Taking this into account, it is important to note the following:
5410 @item If a zero-length bit-field follows a normal bit-field, the type of the
5411 zero-length bit-field may affect the alignment of the structure as whole. For
5412 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
5413 normal bit-field, and is of type short.
5415 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
5416 still affect the alignment of the structure:
5427 Here, @code{t4} takes up 4 bytes.
5430 @item Zero-length bit-fields following non-bit-field members are ignored:
5442 Here, @code{t5} takes up 2 bytes.
5446 @subsection PowerPC Variable Attributes
5448 Three attributes currently are defined for PowerPC configurations:
5449 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5451 For full documentation of the struct attributes please see the
5452 documentation in @ref{i386 Variable Attributes}.
5454 For documentation of @code{altivec} attribute please see the
5455 documentation in @ref{PowerPC Type Attributes}.
5457 @subsection SPU Variable Attributes
5459 The SPU supports the @code{spu_vector} attribute for variables. For
5460 documentation of this attribute please see the documentation in
5461 @ref{SPU Type Attributes}.
5463 @subsection Xstormy16 Variable Attributes
5465 One attribute is currently defined for xstormy16 configurations:
5470 @cindex @code{below100} attribute
5472 If a variable has the @code{below100} attribute (@code{BELOW100} is
5473 allowed also), GCC places the variable in the first 0x100 bytes of
5474 memory and use special opcodes to access it. Such variables are
5475 placed in either the @code{.bss_below100} section or the
5476 @code{.data_below100} section.
5480 @node Type Attributes
5481 @section Specifying Attributes of Types
5482 @cindex attribute of types
5483 @cindex type attributes
5485 The keyword @code{__attribute__} allows you to specify special
5486 attributes of @code{struct} and @code{union} types when you define
5487 such types. This keyword is followed by an attribute specification
5488 inside double parentheses. Seven attributes are currently defined for
5489 types: @code{aligned}, @code{packed}, @code{transparent_union},
5490 @code{unused}, @code{deprecated}, @code{visibility}, and
5491 @code{may_alias}. Other attributes are defined for functions
5492 (@pxref{Function Attributes}) and for variables (@pxref{Variable
5495 You may also specify any one of these attributes with @samp{__}
5496 preceding and following its keyword. This allows you to use these
5497 attributes in header files without being concerned about a possible
5498 macro of the same name. For example, you may use @code{__aligned__}
5499 instead of @code{aligned}.
5501 You may specify type attributes in an enum, struct or union type
5502 declaration or definition, or for other types in a @code{typedef}
5505 For an enum, struct or union type, you may specify attributes either
5506 between the enum, struct or union tag and the name of the type, or
5507 just past the closing curly brace of the @emph{definition}. The
5508 former syntax is preferred.
5510 @xref{Attribute Syntax}, for details of the exact syntax for using
5514 @cindex @code{aligned} attribute
5515 @item aligned (@var{alignment})
5516 This attribute specifies a minimum alignment (in bytes) for variables
5517 of the specified type. For example, the declarations:
5520 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
5521 typedef int more_aligned_int __attribute__ ((aligned (8)));
5525 force the compiler to ensure (as far as it can) that each variable whose
5526 type is @code{struct S} or @code{more_aligned_int} is allocated and
5527 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
5528 variables of type @code{struct S} aligned to 8-byte boundaries allows
5529 the compiler to use the @code{ldd} and @code{std} (doubleword load and
5530 store) instructions when copying one variable of type @code{struct S} to
5531 another, thus improving run-time efficiency.
5533 Note that the alignment of any given @code{struct} or @code{union} type
5534 is required by the ISO C standard to be at least a perfect multiple of
5535 the lowest common multiple of the alignments of all of the members of
5536 the @code{struct} or @code{union} in question. This means that you @emph{can}
5537 effectively adjust the alignment of a @code{struct} or @code{union}
5538 type by attaching an @code{aligned} attribute to any one of the members
5539 of such a type, but the notation illustrated in the example above is a
5540 more obvious, intuitive, and readable way to request the compiler to
5541 adjust the alignment of an entire @code{struct} or @code{union} type.
5543 As in the preceding example, you can explicitly specify the alignment
5544 (in bytes) that you wish the compiler to use for a given @code{struct}
5545 or @code{union} type. Alternatively, you can leave out the alignment factor
5546 and just ask the compiler to align a type to the maximum
5547 useful alignment for the target machine you are compiling for. For
5548 example, you could write:
5551 struct S @{ short f[3]; @} __attribute__ ((aligned));
5554 Whenever you leave out the alignment factor in an @code{aligned}
5555 attribute specification, the compiler automatically sets the alignment
5556 for the type to the largest alignment that is ever used for any data
5557 type on the target machine you are compiling for. Doing this can often
5558 make copy operations more efficient, because the compiler can use
5559 whatever instructions copy the biggest chunks of memory when performing
5560 copies to or from the variables that have types that you have aligned
5563 In the example above, if the size of each @code{short} is 2 bytes, then
5564 the size of the entire @code{struct S} type is 6 bytes. The smallest
5565 power of two that is greater than or equal to that is 8, so the
5566 compiler sets the alignment for the entire @code{struct S} type to 8
5569 Note that although you can ask the compiler to select a time-efficient
5570 alignment for a given type and then declare only individual stand-alone
5571 objects of that type, the compiler's ability to select a time-efficient
5572 alignment is primarily useful only when you plan to create arrays of
5573 variables having the relevant (efficiently aligned) type. If you
5574 declare or use arrays of variables of an efficiently-aligned type, then
5575 it is likely that your program also does pointer arithmetic (or
5576 subscripting, which amounts to the same thing) on pointers to the
5577 relevant type, and the code that the compiler generates for these
5578 pointer arithmetic operations is often more efficient for
5579 efficiently-aligned types than for other types.
5581 The @code{aligned} attribute can only increase the alignment; but you
5582 can decrease it by specifying @code{packed} as well. See below.
5584 Note that the effectiveness of @code{aligned} attributes may be limited
5585 by inherent limitations in your linker. On many systems, the linker is
5586 only able to arrange for variables to be aligned up to a certain maximum
5587 alignment. (For some linkers, the maximum supported alignment may
5588 be very very small.) If your linker is only able to align variables
5589 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5590 in an @code{__attribute__} still only provides you with 8-byte
5591 alignment. See your linker documentation for further information.
5594 This attribute, attached to @code{struct} or @code{union} type
5595 definition, specifies that each member (other than zero-width bit-fields)
5596 of the structure or union is placed to minimize the memory required. When
5597 attached to an @code{enum} definition, it indicates that the smallest
5598 integral type should be used.
5600 @opindex fshort-enums
5601 Specifying this attribute for @code{struct} and @code{union} types is
5602 equivalent to specifying the @code{packed} attribute on each of the
5603 structure or union members. Specifying the @option{-fshort-enums}
5604 flag on the line is equivalent to specifying the @code{packed}
5605 attribute on all @code{enum} definitions.
5607 In the following example @code{struct my_packed_struct}'s members are
5608 packed closely together, but the internal layout of its @code{s} member
5609 is not packed---to do that, @code{struct my_unpacked_struct} needs to
5613 struct my_unpacked_struct
5619 struct __attribute__ ((__packed__)) my_packed_struct
5623 struct my_unpacked_struct s;
5627 You may only specify this attribute on the definition of an @code{enum},
5628 @code{struct} or @code{union}, not on a @code{typedef} that does not
5629 also define the enumerated type, structure or union.
5631 @item transparent_union
5632 This attribute, attached to a @code{union} type definition, indicates
5633 that any function parameter having that union type causes calls to that
5634 function to be treated in a special way.
5636 First, the argument corresponding to a transparent union type can be of
5637 any type in the union; no cast is required. Also, if the union contains
5638 a pointer type, the corresponding argument can be a null pointer
5639 constant or a void pointer expression; and if the union contains a void
5640 pointer type, the corresponding argument can be any pointer expression.
5641 If the union member type is a pointer, qualifiers like @code{const} on
5642 the referenced type must be respected, just as with normal pointer
5645 Second, the argument is passed to the function using the calling
5646 conventions of the first member of the transparent union, not the calling
5647 conventions of the union itself. All members of the union must have the
5648 same machine representation; this is necessary for this argument passing
5651 Transparent unions are designed for library functions that have multiple
5652 interfaces for compatibility reasons. For example, suppose the
5653 @code{wait} function must accept either a value of type @code{int *} to
5654 comply with POSIX, or a value of type @code{union wait *} to comply with
5655 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
5656 @code{wait} would accept both kinds of arguments, but it would also
5657 accept any other pointer type and this would make argument type checking
5658 less useful. Instead, @code{<sys/wait.h>} might define the interface
5662 typedef union __attribute__ ((__transparent_union__))
5666 @} wait_status_ptr_t;
5668 pid_t wait (wait_status_ptr_t);
5672 This interface allows either @code{int *} or @code{union wait *}
5673 arguments to be passed, using the @code{int *} calling convention.
5674 The program can call @code{wait} with arguments of either type:
5677 int w1 () @{ int w; return wait (&w); @}
5678 int w2 () @{ union wait w; return wait (&w); @}
5682 With this interface, @code{wait}'s implementation might look like this:
5685 pid_t wait (wait_status_ptr_t p)
5687 return waitpid (-1, p.__ip, 0);
5692 When attached to a type (including a @code{union} or a @code{struct}),
5693 this attribute means that variables of that type are meant to appear
5694 possibly unused. GCC does not produce a warning for any variables of
5695 that type, even if the variable appears to do nothing. This is often
5696 the case with lock or thread classes, which are usually defined and then
5697 not referenced, but contain constructors and destructors that have
5698 nontrivial bookkeeping functions.
5701 @itemx deprecated (@var{msg})
5702 The @code{deprecated} attribute results in a warning if the type
5703 is used anywhere in the source file. This is useful when identifying
5704 types that are expected to be removed in a future version of a program.
5705 If possible, the warning also includes the location of the declaration
5706 of the deprecated type, to enable users to easily find further
5707 information about why the type is deprecated, or what they should do
5708 instead. Note that the warnings only occur for uses and then only
5709 if the type is being applied to an identifier that itself is not being
5710 declared as deprecated.
5713 typedef int T1 __attribute__ ((deprecated));
5717 typedef T1 T3 __attribute__ ((deprecated));
5718 T3 z __attribute__ ((deprecated));
5722 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
5723 warning is issued for line 4 because T2 is not explicitly
5724 deprecated. Line 5 has no warning because T3 is explicitly
5725 deprecated. Similarly for line 6. The optional @var{msg}
5726 argument, which must be a string, is printed in the warning if
5729 The @code{deprecated} attribute can also be used for functions and
5730 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
5733 Accesses through pointers to types with this attribute are not subject
5734 to type-based alias analysis, but are instead assumed to be able to alias
5735 any other type of objects.
5736 In the context of section 6.5 paragraph 7 of the C99 standard,
5737 an lvalue expression
5738 dereferencing such a pointer is treated like having a character type.
5739 See @option{-fstrict-aliasing} for more information on aliasing issues.
5740 This extension exists to support some vector APIs, in which pointers to
5741 one vector type are permitted to alias pointers to a different vector type.
5743 Note that an object of a type with this attribute does not have any
5749 typedef short __attribute__((__may_alias__)) short_a;
5755 short_a *b = (short_a *) &a;
5759 if (a == 0x12345678)
5767 If you replaced @code{short_a} with @code{short} in the variable
5768 declaration, the above program would abort when compiled with
5769 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
5770 above in recent GCC versions.
5773 In C++, attribute visibility (@pxref{Function Attributes}) can also be
5774 applied to class, struct, union and enum types. Unlike other type
5775 attributes, the attribute must appear between the initial keyword and
5776 the name of the type; it cannot appear after the body of the type.
5778 Note that the type visibility is applied to vague linkage entities
5779 associated with the class (vtable, typeinfo node, etc.). In
5780 particular, if a class is thrown as an exception in one shared object
5781 and caught in another, the class must have default visibility.
5782 Otherwise the two shared objects are unable to use the same
5783 typeinfo node and exception handling will break.
5787 To specify multiple attributes, separate them by commas within the
5788 double parentheses: for example, @samp{__attribute__ ((aligned (16),
5791 @subsection ARM Type Attributes
5793 On those ARM targets that support @code{dllimport} (such as Symbian
5794 OS), you can use the @code{notshared} attribute to indicate that the
5795 virtual table and other similar data for a class should not be
5796 exported from a DLL@. For example:
5799 class __declspec(notshared) C @{
5801 __declspec(dllimport) C();
5805 __declspec(dllexport)
5810 In this code, @code{C::C} is exported from the current DLL, but the
5811 virtual table for @code{C} is not exported. (You can use
5812 @code{__attribute__} instead of @code{__declspec} if you prefer, but
5813 most Symbian OS code uses @code{__declspec}.)
5815 @anchor{MeP Type Attributes}
5816 @subsection MeP Type Attributes
5818 Many of the MeP variable attributes may be applied to types as well.
5819 Specifically, the @code{based}, @code{tiny}, @code{near}, and
5820 @code{far} attributes may be applied to either. The @code{io} and
5821 @code{cb} attributes may not be applied to types.
5823 @anchor{i386 Type Attributes}
5824 @subsection i386 Type Attributes
5826 Two attributes are currently defined for i386 configurations:
5827 @code{ms_struct} and @code{gcc_struct}.
5833 @cindex @code{ms_struct}
5834 @cindex @code{gcc_struct}
5836 If @code{packed} is used on a structure, or if bit-fields are used
5837 it may be that the Microsoft ABI packs them differently
5838 than GCC normally packs them. Particularly when moving packed
5839 data between functions compiled with GCC and the native Microsoft compiler
5840 (either via function call or as data in a file), it may be necessary to access
5843 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
5844 compilers to match the native Microsoft compiler.
5847 @anchor{PowerPC Type Attributes}
5848 @subsection PowerPC Type Attributes
5850 Three attributes currently are defined for PowerPC configurations:
5851 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5853 For full documentation of the @code{ms_struct} and @code{gcc_struct}
5854 attributes please see the documentation in @ref{i386 Type Attributes}.
5856 The @code{altivec} attribute allows one to declare AltiVec vector data
5857 types supported by the AltiVec Programming Interface Manual. The
5858 attribute requires an argument to specify one of three vector types:
5859 @code{vector__}, @code{pixel__} (always followed by unsigned short),
5860 and @code{bool__} (always followed by unsigned).
5863 __attribute__((altivec(vector__)))
5864 __attribute__((altivec(pixel__))) unsigned short
5865 __attribute__((altivec(bool__))) unsigned
5868 These attributes mainly are intended to support the @code{__vector},
5869 @code{__pixel}, and @code{__bool} AltiVec keywords.
5871 @anchor{SPU Type Attributes}
5872 @subsection SPU Type Attributes
5874 The SPU supports the @code{spu_vector} attribute for types. This attribute
5875 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
5876 Language Extensions Specification. It is intended to support the
5877 @code{__vector} keyword.
5880 @section Inquiring on Alignment of Types or Variables
5882 @cindex type alignment
5883 @cindex variable alignment
5885 The keyword @code{__alignof__} allows you to inquire about how an object
5886 is aligned, or the minimum alignment usually required by a type. Its
5887 syntax is just like @code{sizeof}.
5889 For example, if the target machine requires a @code{double} value to be
5890 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
5891 This is true on many RISC machines. On more traditional machine
5892 designs, @code{__alignof__ (double)} is 4 or even 2.
5894 Some machines never actually require alignment; they allow reference to any
5895 data type even at an odd address. For these machines, @code{__alignof__}
5896 reports the smallest alignment that GCC gives the data type, usually as
5897 mandated by the target ABI.
5899 If the operand of @code{__alignof__} is an lvalue rather than a type,
5900 its value is the required alignment for its type, taking into account
5901 any minimum alignment specified with GCC's @code{__attribute__}
5902 extension (@pxref{Variable Attributes}). For example, after this
5906 struct foo @{ int x; char y; @} foo1;
5910 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
5911 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
5913 It is an error to ask for the alignment of an incomplete type.
5917 @section An Inline Function is As Fast As a Macro
5918 @cindex inline functions
5919 @cindex integrating function code
5921 @cindex macros, inline alternative
5923 By declaring a function inline, you can direct GCC to make
5924 calls to that function faster. One way GCC can achieve this is to
5925 integrate that function's code into the code for its callers. This
5926 makes execution faster by eliminating the function-call overhead; in
5927 addition, if any of the actual argument values are constant, their
5928 known values may permit simplifications at compile time so that not
5929 all of the inline function's code needs to be included. The effect on
5930 code size is less predictable; object code may be larger or smaller
5931 with function inlining, depending on the particular case. You can
5932 also direct GCC to try to integrate all ``simple enough'' functions
5933 into their callers with the option @option{-finline-functions}.
5935 GCC implements three different semantics of declaring a function
5936 inline. One is available with @option{-std=gnu89} or
5937 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
5938 on all inline declarations, another when
5939 @option{-std=c99}, @option{-std=c11},
5940 @option{-std=gnu99} or @option{-std=gnu11}
5941 (without @option{-fgnu89-inline}), and the third
5942 is used when compiling C++.
5944 To declare a function inline, use the @code{inline} keyword in its
5945 declaration, like this:
5955 If you are writing a header file to be included in ISO C90 programs, write
5956 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
5958 The three types of inlining behave similarly in two important cases:
5959 when the @code{inline} keyword is used on a @code{static} function,
5960 like the example above, and when a function is first declared without
5961 using the @code{inline} keyword and then is defined with
5962 @code{inline}, like this:
5965 extern int inc (int *a);
5973 In both of these common cases, the program behaves the same as if you
5974 had not used the @code{inline} keyword, except for its speed.
5976 @cindex inline functions, omission of
5977 @opindex fkeep-inline-functions
5978 When a function is both inline and @code{static}, if all calls to the
5979 function are integrated into the caller, and the function's address is
5980 never used, then the function's own assembler code is never referenced.
5981 In this case, GCC does not actually output assembler code for the
5982 function, unless you specify the option @option{-fkeep-inline-functions}.
5983 Some calls cannot be integrated for various reasons (in particular,
5984 calls that precede the function's definition cannot be integrated, and
5985 neither can recursive calls within the definition). If there is a
5986 nonintegrated call, then the function is compiled to assembler code as
5987 usual. The function must also be compiled as usual if the program
5988 refers to its address, because that can't be inlined.
5991 Note that certain usages in a function definition can make it unsuitable
5992 for inline substitution. Among these usages are: variadic functions, use of
5993 @code{alloca}, use of variable-length data types (@pxref{Variable Length}),
5994 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
5995 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
5996 warns when a function marked @code{inline} could not be substituted,
5997 and gives the reason for the failure.
5999 @cindex automatic @code{inline} for C++ member fns
6000 @cindex @code{inline} automatic for C++ member fns
6001 @cindex member fns, automatically @code{inline}
6002 @cindex C++ member fns, automatically @code{inline}
6003 @opindex fno-default-inline
6004 As required by ISO C++, GCC considers member functions defined within
6005 the body of a class to be marked inline even if they are
6006 not explicitly declared with the @code{inline} keyword. You can
6007 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
6008 Options,,Options Controlling C++ Dialect}.
6010 GCC does not inline any functions when not optimizing unless you specify
6011 the @samp{always_inline} attribute for the function, like this:
6014 /* @r{Prototype.} */
6015 inline void foo (const char) __attribute__((always_inline));
6018 The remainder of this section is specific to GNU C90 inlining.
6020 @cindex non-static inline function
6021 When an inline function is not @code{static}, then the compiler must assume
6022 that there may be calls from other source files; since a global symbol can
6023 be defined only once in any program, the function must not be defined in
6024 the other source files, so the calls therein cannot be integrated.
6025 Therefore, a non-@code{static} inline function is always compiled on its
6026 own in the usual fashion.
6028 If you specify both @code{inline} and @code{extern} in the function
6029 definition, then the definition is used only for inlining. In no case
6030 is the function compiled on its own, not even if you refer to its
6031 address explicitly. Such an address becomes an external reference, as
6032 if you had only declared the function, and had not defined it.
6034 This combination of @code{inline} and @code{extern} has almost the
6035 effect of a macro. The way to use it is to put a function definition in
6036 a header file with these keywords, and put another copy of the
6037 definition (lacking @code{inline} and @code{extern}) in a library file.
6038 The definition in the header file causes most calls to the function
6039 to be inlined. If any uses of the function remain, they refer to
6040 the single copy in the library.
6043 @section When is a Volatile Object Accessed?
6044 @cindex accessing volatiles
6045 @cindex volatile read
6046 @cindex volatile write
6047 @cindex volatile access
6049 C has the concept of volatile objects. These are normally accessed by
6050 pointers and used for accessing hardware or inter-thread
6051 communication. The standard encourages compilers to refrain from
6052 optimizations concerning accesses to volatile objects, but leaves it
6053 implementation defined as to what constitutes a volatile access. The
6054 minimum requirement is that at a sequence point all previous accesses
6055 to volatile objects have stabilized and no subsequent accesses have
6056 occurred. Thus an implementation is free to reorder and combine
6057 volatile accesses that occur between sequence points, but cannot do
6058 so for accesses across a sequence point. The use of volatile does
6059 not allow you to violate the restriction on updating objects multiple
6060 times between two sequence points.
6062 Accesses to non-volatile objects are not ordered with respect to
6063 volatile accesses. You cannot use a volatile object as a memory
6064 barrier to order a sequence of writes to non-volatile memory. For
6068 int *ptr = @var{something};
6070 *ptr = @var{something};
6075 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
6076 that the write to @var{*ptr} occurs by the time the update
6077 of @var{vobj} happens. If you need this guarantee, you must use
6078 a stronger memory barrier such as:
6081 int *ptr = @var{something};
6083 *ptr = @var{something};
6084 asm volatile ("" : : : "memory");
6088 A scalar volatile object is read when it is accessed in a void context:
6091 volatile int *src = @var{somevalue};
6095 Such expressions are rvalues, and GCC implements this as a
6096 read of the volatile object being pointed to.
6098 Assignments are also expressions and have an rvalue. However when
6099 assigning to a scalar volatile, the volatile object is not reread,
6100 regardless of whether the assignment expression's rvalue is used or
6101 not. If the assignment's rvalue is used, the value is that assigned
6102 to the volatile object. For instance, there is no read of @var{vobj}
6103 in all the following cases:
6108 vobj = @var{something};
6109 obj = vobj = @var{something};
6110 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
6111 obj = (@var{something}, vobj = @var{anotherthing});
6114 If you need to read the volatile object after an assignment has
6115 occurred, you must use a separate expression with an intervening
6118 As bit-fields are not individually addressable, volatile bit-fields may
6119 be implicitly read when written to, or when adjacent bit-fields are
6120 accessed. Bit-field operations may be optimized such that adjacent
6121 bit-fields are only partially accessed, if they straddle a storage unit
6122 boundary. For these reasons it is unwise to use volatile bit-fields to
6126 @section Assembler Instructions with C Expression Operands
6127 @cindex extended @code{asm}
6128 @cindex @code{asm} expressions
6129 @cindex assembler instructions
6132 In an assembler instruction using @code{asm}, you can specify the
6133 operands of the instruction using C expressions. This means you need not
6134 guess which registers or memory locations contain the data you want
6137 You must specify an assembler instruction template much like what
6138 appears in a machine description, plus an operand constraint string for
6141 For example, here is how to use the 68881's @code{fsinx} instruction:
6144 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
6148 Here @code{angle} is the C expression for the input operand while
6149 @code{result} is that of the output operand. Each has @samp{"f"} as its
6150 operand constraint, saying that a floating-point register is required.
6151 The @samp{=} in @samp{=f} indicates that the operand is an output; all
6152 output operands' constraints must use @samp{=}. The constraints use the
6153 same language used in the machine description (@pxref{Constraints}).
6155 Each operand is described by an operand-constraint string followed by
6156 the C expression in parentheses. A colon separates the assembler
6157 template from the first output operand and another separates the last
6158 output operand from the first input, if any. Commas separate the
6159 operands within each group. The total number of operands is currently
6160 limited to 30; this limitation may be lifted in some future version of
6163 If there are no output operands but there are input operands, you must
6164 place two consecutive colons surrounding the place where the output
6167 As of GCC version 3.1, it is also possible to specify input and output
6168 operands using symbolic names which can be referenced within the
6169 assembler code. These names are specified inside square brackets
6170 preceding the constraint string, and can be referenced inside the
6171 assembler code using @code{%[@var{name}]} instead of a percentage sign
6172 followed by the operand number. Using named operands the above example
6176 asm ("fsinx %[angle],%[output]"
6177 : [output] "=f" (result)
6178 : [angle] "f" (angle));
6182 Note that the symbolic operand names have no relation whatsoever to
6183 other C identifiers. You may use any name you like, even those of
6184 existing C symbols, but you must ensure that no two operands within the same
6185 assembler construct use the same symbolic name.
6187 Output operand expressions must be lvalues; the compiler can check this.
6188 The input operands need not be lvalues. The compiler cannot check
6189 whether the operands have data types that are reasonable for the
6190 instruction being executed. It does not parse the assembler instruction
6191 template and does not know what it means or even whether it is valid
6192 assembler input. The extended @code{asm} feature is most often used for
6193 machine instructions the compiler itself does not know exist. If
6194 the output expression cannot be directly addressed (for example, it is a
6195 bit-field), your constraint must allow a register. In that case, GCC
6196 uses the register as the output of the @code{asm}, and then stores
6197 that register into the output.
6199 The ordinary output operands must be write-only; GCC assumes that
6200 the values in these operands before the instruction are dead and need
6201 not be generated. Extended asm supports input-output or read-write
6202 operands. Use the constraint character @samp{+} to indicate such an
6203 operand and list it with the output operands.
6205 You may, as an alternative, logically split its function into two
6206 separate operands, one input operand and one write-only output
6207 operand. The connection between them is expressed by constraints
6208 that say they need to be in the same location when the instruction
6209 executes. You can use the same C expression for both operands, or
6210 different expressions. For example, here we write the (fictitious)
6211 @samp{combine} instruction with @code{bar} as its read-only source
6212 operand and @code{foo} as its read-write destination:
6215 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
6219 The constraint @samp{"0"} for operand 1 says that it must occupy the
6220 same location as operand 0. A number in constraint is allowed only in
6221 an input operand and it must refer to an output operand.
6223 Only a number in the constraint can guarantee that one operand is in
6224 the same place as another. The mere fact that @code{foo} is the value
6225 of both operands is not enough to guarantee that they are in the
6226 same place in the generated assembler code. The following does not
6230 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
6233 Various optimizations or reloading could cause operands 0 and 1 to be in
6234 different registers; GCC knows no reason not to do so. For example, the
6235 compiler might find a copy of the value of @code{foo} in one register and
6236 use it for operand 1, but generate the output operand 0 in a different
6237 register (copying it afterward to @code{foo}'s own address). Of course,
6238 since the register for operand 1 is not even mentioned in the assembler
6239 code, the result will not work, but GCC can't tell that.
6241 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
6242 the operand number for a matching constraint. For example:
6245 asm ("cmoveq %1,%2,%[result]"
6246 : [result] "=r"(result)
6247 : "r" (test), "r"(new), "[result]"(old));
6250 Sometimes you need to make an @code{asm} operand be a specific register,
6251 but there's no matching constraint letter for that register @emph{by
6252 itself}. To force the operand into that register, use a local variable
6253 for the operand and specify the register in the variable declaration.
6254 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
6255 register constraint letter that matches the register:
6258 register int *p1 asm ("r0") = @dots{};
6259 register int *p2 asm ("r1") = @dots{};
6260 register int *result asm ("r0");
6261 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
6264 @anchor{Example of asm with clobbered asm reg}
6265 In the above example, beware that a register that is call-clobbered by
6266 the target ABI will be overwritten by any function call in the
6267 assignment, including library calls for arithmetic operators.
6268 Also a register may be clobbered when generating some operations,
6269 like variable shift, memory copy or memory move on x86.
6270 Assuming it is a call-clobbered register, this may happen to @code{r0}
6271 above by the assignment to @code{p2}. If you have to use such a
6272 register, use temporary variables for expressions between the register
6277 register int *p1 asm ("r0") = @dots{};
6278 register int *p2 asm ("r1") = t1;
6279 register int *result asm ("r0");
6280 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
6283 Some instructions clobber specific hard registers. To describe this,
6284 write a third colon after the input operands, followed by the names of
6285 the clobbered hard registers (given as strings). Here is a realistic
6286 example for the VAX:
6289 asm volatile ("movc3 %0,%1,%2"
6290 : /* @r{no outputs} */
6291 : "g" (from), "g" (to), "g" (count)
6292 : "r0", "r1", "r2", "r3", "r4", "r5");
6295 You may not write a clobber description in a way that overlaps with an
6296 input or output operand. For example, you may not have an operand
6297 describing a register class with one member if you mention that register
6298 in the clobber list. Variables declared to live in specific registers
6299 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
6300 have no part mentioned in the clobber description.
6301 There is no way for you to specify that an input
6302 operand is modified without also specifying it as an output
6303 operand. Note that if all the output operands you specify are for this
6304 purpose (and hence unused), you then also need to specify
6305 @code{volatile} for the @code{asm} construct, as described below, to
6306 prevent GCC from deleting the @code{asm} statement as unused.
6308 If you refer to a particular hardware register from the assembler code,
6309 you probably have to list the register after the third colon to
6310 tell the compiler the register's value is modified. In some assemblers,
6311 the register names begin with @samp{%}; to produce one @samp{%} in the
6312 assembler code, you must write @samp{%%} in the input.
6314 If your assembler instruction can alter the condition code register, add
6315 @samp{cc} to the list of clobbered registers. GCC on some machines
6316 represents the condition codes as a specific hardware register;
6317 @samp{cc} serves to name this register. On other machines, the
6318 condition code is handled differently, and specifying @samp{cc} has no
6319 effect. But it is valid no matter what the machine.
6321 If your assembler instructions access memory in an unpredictable
6322 fashion, add @samp{memory} to the list of clobbered registers. This
6323 causes GCC to not keep memory values cached in registers across the
6324 assembler instruction and not optimize stores or loads to that memory.
6325 You also should add the @code{volatile} keyword if the memory
6326 affected is not listed in the inputs or outputs of the @code{asm}, as
6327 the @samp{memory} clobber does not count as a side-effect of the
6328 @code{asm}. If you know how large the accessed memory is, you can add
6329 it as input or output but if this is not known, you should add
6330 @samp{memory}. As an example, if you access ten bytes of a string, you
6331 can use a memory input like:
6334 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
6337 Note that in the following example the memory input is necessary,
6338 otherwise GCC might optimize the store to @code{x} away:
6345 asm ("magic stuff accessing an 'int' pointed to by '%1'"
6346 : "=&d" (r) : "a" (y), "m" (*y));
6351 You can put multiple assembler instructions together in a single
6352 @code{asm} template, separated by the characters normally used in assembly
6353 code for the system. A combination that works in most places is a newline
6354 to break the line, plus a tab character to move to the instruction field
6355 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
6356 assembler allows semicolons as a line-breaking character. Note that some
6357 assembler dialects use semicolons to start a comment.
6358 The input operands are guaranteed not to use any of the clobbered
6359 registers, and neither do the output operands' addresses, so you can
6360 read and write the clobbered registers as many times as you like. Here
6361 is an example of multiple instructions in a template; it assumes the
6362 subroutine @code{_foo} accepts arguments in registers 9 and 10:
6365 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
6367 : "g" (from), "g" (to)
6371 Unless an output operand has the @samp{&} constraint modifier, GCC
6372 may allocate it in the same register as an unrelated input operand, on
6373 the assumption the inputs are consumed before the outputs are produced.
6374 This assumption may be false if the assembler code actually consists of
6375 more than one instruction. In such a case, use @samp{&} for each output
6376 operand that may not overlap an input. @xref{Modifiers}.
6378 If you want to test the condition code produced by an assembler
6379 instruction, you must include a branch and a label in the @code{asm}
6380 construct, as follows:
6383 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
6389 This assumes your assembler supports local labels, as the GNU assembler
6390 and most Unix assemblers do.
6392 Speaking of labels, jumps from one @code{asm} to another are not
6393 supported. The compiler's optimizers do not know about these jumps, and
6394 therefore they cannot take account of them when deciding how to
6395 optimize. @xref{Extended asm with goto}.
6397 @cindex macros containing @code{asm}
6398 Usually the most convenient way to use these @code{asm} instructions is to
6399 encapsulate them in macros that look like functions. For example,
6403 (@{ double __value, __arg = (x); \
6404 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
6409 Here the variable @code{__arg} is used to make sure that the instruction
6410 operates on a proper @code{double} value, and to accept only those
6411 arguments @code{x} that can convert automatically to a @code{double}.
6413 Another way to make sure the instruction operates on the correct data
6414 type is to use a cast in the @code{asm}. This is different from using a
6415 variable @code{__arg} in that it converts more different types. For
6416 example, if the desired type is @code{int}, casting the argument to
6417 @code{int} accepts a pointer with no complaint, while assigning the
6418 argument to an @code{int} variable named @code{__arg} warns about
6419 using a pointer unless the caller explicitly casts it.
6421 If an @code{asm} has output operands, GCC assumes for optimization
6422 purposes the instruction has no side effects except to change the output
6423 operands. This does not mean instructions with a side effect cannot be
6424 used, but you must be careful, because the compiler may eliminate them
6425 if the output operands aren't used, or move them out of loops, or
6426 replace two with one if they constitute a common subexpression. Also,
6427 if your instruction does have a side effect on a variable that otherwise
6428 appears not to change, the old value of the variable may be reused later
6429 if it happens to be found in a register.
6431 You can prevent an @code{asm} instruction from being deleted
6432 by writing the keyword @code{volatile} after
6433 the @code{asm}. For example:
6436 #define get_and_set_priority(new) \
6438 asm volatile ("get_and_set_priority %0, %1" \
6439 : "=g" (__old) : "g" (new)); \
6444 The @code{volatile} keyword indicates that the instruction has
6445 important side-effects. GCC does not delete a volatile @code{asm} if
6446 it is reachable. (The instruction can still be deleted if GCC can
6447 prove that control flow never reaches the location of the
6448 instruction.) Note that even a volatile @code{asm} instruction
6449 can be moved relative to other code, including across jump
6450 instructions. For example, on many targets there is a system
6451 register that can be set to control the rounding mode of
6452 floating-point operations. You might try
6453 setting it with a volatile @code{asm}, like this PowerPC example:
6456 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
6461 This does not work reliably, as the compiler may move the addition back
6462 before the volatile @code{asm}. To make it work you need to add an
6463 artificial dependency to the @code{asm} referencing a variable in the code
6464 you don't want moved, for example:
6467 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
6471 Similarly, you can't expect a
6472 sequence of volatile @code{asm} instructions to remain perfectly
6473 consecutive. If you want consecutive output, use a single @code{asm}.
6474 Also, GCC performs some optimizations across a volatile @code{asm}
6475 instruction; GCC does not ``forget everything'' when it encounters
6476 a volatile @code{asm} instruction the way some other compilers do.
6478 An @code{asm} instruction without any output operands is treated
6479 identically to a volatile @code{asm} instruction.
6481 It is a natural idea to look for a way to give access to the condition
6482 code left by the assembler instruction. However, when we attempted to
6483 implement this, we found no way to make it work reliably. The problem
6484 is that output operands might need reloading, which result in
6485 additional following ``store'' instructions. On most machines, these
6486 instructions alter the condition code before there is time to
6487 test it. This problem doesn't arise for ordinary ``test'' and
6488 ``compare'' instructions because they don't have any output operands.
6490 For reasons similar to those described above, it is not possible to give
6491 an assembler instruction access to the condition code left by previous
6494 @anchor{Extended asm with goto}
6495 As of GCC version 4.5, @code{asm goto} may be used to have the assembly
6496 jump to one or more C labels. In this form, a fifth section after the
6497 clobber list contains a list of all C labels to which the assembly may jump.
6498 Each label operand is implicitly self-named. The @code{asm} is also assumed
6499 to fall through to the next statement.
6501 This form of @code{asm} is restricted to not have outputs. This is due
6502 to a internal restriction in the compiler that control transfer instructions
6503 cannot have outputs. This restriction on @code{asm goto} may be lifted
6504 in some future version of the compiler. In the meantime, @code{asm goto}
6505 may include a memory clobber, and so leave outputs in memory.
6511 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
6512 : : "r"(x), "r"(&y) : "r5", "memory" : error);
6520 In this (inefficient) example, the @code{frob} instruction sets the
6521 carry bit to indicate an error. The @code{jc} instruction detects
6522 this and branches to the @code{error} label. Finally, the output
6523 of the @code{frob} instruction (@code{%r5}) is stored into the memory
6524 for variable @code{y}, which is later read by the @code{return} statement.
6530 asm goto ("mfsr %%r1, 123; jmp %%r1;"
6531 ".pushsection doit_table;"
6532 ".long %l0, %l1, %l2, %l3;"
6534 : : : "r1" : label1, label2, label3, label4);
6535 __builtin_unreachable ();
6551 In this (also inefficient) example, the @code{mfsr} instruction reads
6552 an address from some out-of-band machine register, and the following
6553 @code{jmp} instruction branches to that address. The address read by
6554 the @code{mfsr} instruction is assumed to have been previously set via
6555 some application-specific mechanism to be one of the four values stored
6556 in the @code{doit_table} section. Finally, the @code{asm} is followed
6557 by a call to @code{__builtin_unreachable} to indicate that the @code{asm}
6558 does not in fact fall through.
6561 #define TRACE1(NUM) \
6563 asm goto ("0: nop;" \
6564 ".pushsection trace_table;" \
6567 : : : : trace#NUM); \
6568 if (0) @{ trace#NUM: trace(); @} \
6570 #define TRACE TRACE1(__COUNTER__)
6574 In this example (which in fact inspired the @code{asm goto} feature)
6575 we want on rare occasions to call the @code{trace} function; on other
6576 occasions we'd like to keep the overhead to the absolute minimum.
6577 The normal code path consists of a single @code{nop} instruction.
6578 However, we record the address of this @code{nop} together with the
6579 address of a label that calls the @code{trace} function. This allows
6580 the @code{nop} instruction to be patched at run time to be an
6581 unconditional branch to the stored label. It is assumed that an
6582 optimizing compiler moves the labeled block out of line, to
6583 optimize the fall through path from the @code{asm}.
6585 If you are writing a header file that should be includable in ISO C
6586 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
6589 @subsection Size of an @code{asm}
6591 Some targets require that GCC track the size of each instruction used in
6592 order to generate correct code. Because the final length of an
6593 @code{asm} is only known by the assembler, GCC must make an estimate as
6594 to how big it will be. The estimate is formed by counting the number of
6595 statements in the pattern of the @code{asm} and multiplying that by the
6596 length of the longest instruction on that processor. Statements in the
6597 @code{asm} are identified by newline characters and whatever statement
6598 separator characters are supported by the assembler; on most processors
6599 this is the @samp{;} character.
6601 Normally, GCC's estimate is perfectly adequate to ensure that correct
6602 code is generated, but it is possible to confuse the compiler if you use
6603 pseudo instructions or assembler macros that expand into multiple real
6604 instructions or if you use assembler directives that expand to more
6605 space in the object file than is needed for a single instruction.
6606 If this happens then the assembler produces a diagnostic saying that
6607 a label is unreachable.
6609 @subsection i386 floating-point asm operands
6611 On i386 targets, there are several rules on the usage of stack-like registers
6612 in the operands of an @code{asm}. These rules apply only to the operands
6613 that are stack-like registers:
6617 Given a set of input registers that die in an @code{asm}, it is
6618 necessary to know which are implicitly popped by the @code{asm}, and
6619 which must be explicitly popped by GCC@.
6621 An input register that is implicitly popped by the @code{asm} must be
6622 explicitly clobbered, unless it is constrained to match an
6626 For any input register that is implicitly popped by an @code{asm}, it is
6627 necessary to know how to adjust the stack to compensate for the pop.
6628 If any non-popped input is closer to the top of the reg-stack than
6629 the implicitly popped register, it would not be possible to know what the
6630 stack looked like---it's not clear how the rest of the stack ``slides
6633 All implicitly popped input registers must be closer to the top of
6634 the reg-stack than any input that is not implicitly popped.
6636 It is possible that if an input dies in an @code{asm}, the compiler might
6637 use the input register for an output reload. Consider this example:
6640 asm ("foo" : "=t" (a) : "f" (b));
6644 This code says that input @code{b} is not popped by the @code{asm}, and that
6645 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
6646 deeper after the @code{asm} than it was before. But, it is possible that
6647 reload may think that it can use the same register for both the input and
6650 To prevent this from happening,
6651 if any input operand uses the @code{f} constraint, all output register
6652 constraints must use the @code{&} early-clobber modifier.
6654 The example above would be correctly written as:
6657 asm ("foo" : "=&t" (a) : "f" (b));
6661 Some operands need to be in particular places on the stack. All
6662 output operands fall in this category---GCC has no other way to
6663 know which registers the outputs appear in unless you indicate
6664 this in the constraints.
6666 Output operands must specifically indicate which register an output
6667 appears in after an @code{asm}. @code{=f} is not allowed: the operand
6668 constraints must select a class with a single register.
6671 Output operands may not be ``inserted'' between existing stack registers.
6672 Since no 387 opcode uses a read/write operand, all output operands
6673 are dead before the @code{asm}, and are pushed by the @code{asm}.
6674 It makes no sense to push anywhere but the top of the reg-stack.
6676 Output operands must start at the top of the reg-stack: output
6677 operands may not ``skip'' a register.
6680 Some @code{asm} statements may need extra stack space for internal
6681 calculations. This can be guaranteed by clobbering stack registers
6682 unrelated to the inputs and outputs.
6686 Here are a couple of reasonable @code{asm}s to want to write. This
6688 takes one input, which is internally popped, and produces two outputs.
6691 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
6695 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
6696 and replaces them with one output. The @code{st(1)} clobber is necessary
6697 for the compiler to know that @code{fyl2xp1} pops both inputs.
6700 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
6706 @section Controlling Names Used in Assembler Code
6707 @cindex assembler names for identifiers
6708 @cindex names used in assembler code
6709 @cindex identifiers, names in assembler code
6711 You can specify the name to be used in the assembler code for a C
6712 function or variable by writing the @code{asm} (or @code{__asm__})
6713 keyword after the declarator as follows:
6716 int foo asm ("myfoo") = 2;
6720 This specifies that the name to be used for the variable @code{foo} in
6721 the assembler code should be @samp{myfoo} rather than the usual
6724 On systems where an underscore is normally prepended to the name of a C
6725 function or variable, this feature allows you to define names for the
6726 linker that do not start with an underscore.
6728 It does not make sense to use this feature with a non-static local
6729 variable since such variables do not have assembler names. If you are
6730 trying to put the variable in a particular register, see @ref{Explicit
6731 Reg Vars}. GCC presently accepts such code with a warning, but will
6732 probably be changed to issue an error, rather than a warning, in the
6735 You cannot use @code{asm} in this way in a function @emph{definition}; but
6736 you can get the same effect by writing a declaration for the function
6737 before its definition and putting @code{asm} there, like this:
6740 extern func () asm ("FUNC");
6747 It is up to you to make sure that the assembler names you choose do not
6748 conflict with any other assembler symbols. Also, you must not use a
6749 register name; that would produce completely invalid assembler code. GCC
6750 does not as yet have the ability to store static variables in registers.
6751 Perhaps that will be added.
6753 @node Explicit Reg Vars
6754 @section Variables in Specified Registers
6755 @cindex explicit register variables
6756 @cindex variables in specified registers
6757 @cindex specified registers
6758 @cindex registers, global allocation
6760 GNU C allows you to put a few global variables into specified hardware
6761 registers. You can also specify the register in which an ordinary
6762 register variable should be allocated.
6766 Global register variables reserve registers throughout the program.
6767 This may be useful in programs such as programming language
6768 interpreters that have a couple of global variables that are accessed
6772 Local register variables in specific registers do not reserve the
6773 registers, except at the point where they are used as input or output
6774 operands in an @code{asm} statement and the @code{asm} statement itself is
6775 not deleted. The compiler's data flow analysis is capable of determining
6776 where the specified registers contain live values, and where they are
6777 available for other uses. Stores into local register variables may be deleted
6778 when they appear to be dead according to dataflow analysis. References
6779 to local register variables may be deleted or moved or simplified.
6781 These local variables are sometimes convenient for use with the extended
6782 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
6783 output of the assembler instruction directly into a particular register.
6784 (This works provided the register you specify fits the constraints
6785 specified for that operand in the @code{asm}.)
6793 @node Global Reg Vars
6794 @subsection Defining Global Register Variables
6795 @cindex global register variables
6796 @cindex registers, global variables in
6798 You can define a global register variable in GNU C like this:
6801 register int *foo asm ("a5");
6805 Here @code{a5} is the name of the register that should be used. Choose a
6806 register that is normally saved and restored by function calls on your
6807 machine, so that library routines will not clobber it.
6809 Naturally the register name is cpu-dependent, so you need to
6810 conditionalize your program according to cpu type. The register
6811 @code{a5} is a good choice on a 68000 for a variable of pointer
6812 type. On machines with register windows, be sure to choose a ``global''
6813 register that is not affected magically by the function call mechanism.
6815 In addition, different operating systems on the same CPU may differ in how they
6816 name the registers; then you need additional conditionals. For
6817 example, some 68000 operating systems call this register @code{%a5}.
6819 Eventually there may be a way of asking the compiler to choose a register
6820 automatically, but first we need to figure out how it should choose and
6821 how to enable you to guide the choice. No solution is evident.
6823 Defining a global register variable in a certain register reserves that
6824 register entirely for this use, at least within the current compilation.
6825 The register is not allocated for any other purpose in the functions
6826 in the current compilation, and is not saved and restored by
6827 these functions. Stores into this register are never deleted even if they
6828 appear to be dead, but references may be deleted or moved or
6831 It is not safe to access the global register variables from signal
6832 handlers, or from more than one thread of control, because the system
6833 library routines may temporarily use the register for other things (unless
6834 you recompile them specially for the task at hand).
6836 @cindex @code{qsort}, and global register variables
6837 It is not safe for one function that uses a global register variable to
6838 call another such function @code{foo} by way of a third function
6839 @code{lose} that is compiled without knowledge of this variable (i.e.@: in a
6840 different source file in which the variable isn't declared). This is
6841 because @code{lose} might save the register and put some other value there.
6842 For example, you can't expect a global register variable to be available in
6843 the comparison-function that you pass to @code{qsort}, since @code{qsort}
6844 might have put something else in that register. (If you are prepared to
6845 recompile @code{qsort} with the same global register variable, you can
6846 solve this problem.)
6848 If you want to recompile @code{qsort} or other source files that do not
6849 actually use your global register variable, so that they do not use that
6850 register for any other purpose, then it suffices to specify the compiler
6851 option @option{-ffixed-@var{reg}}. You need not actually add a global
6852 register declaration to their source code.
6854 A function that can alter the value of a global register variable cannot
6855 safely be called from a function compiled without this variable, because it
6856 could clobber the value the caller expects to find there on return.
6857 Therefore, the function that is the entry point into the part of the
6858 program that uses the global register variable must explicitly save and
6859 restore the value that belongs to its caller.
6861 @cindex register variable after @code{longjmp}
6862 @cindex global register after @code{longjmp}
6863 @cindex value after @code{longjmp}
6866 On most machines, @code{longjmp} restores to each global register
6867 variable the value it had at the time of the @code{setjmp}. On some
6868 machines, however, @code{longjmp} does not change the value of global
6869 register variables. To be portable, the function that called @code{setjmp}
6870 should make other arrangements to save the values of the global register
6871 variables, and to restore them in a @code{longjmp}. This way, the same
6872 thing happens regardless of what @code{longjmp} does.
6874 All global register variable declarations must precede all function
6875 definitions. If such a declaration could appear after function
6876 definitions, the declaration would be too late to prevent the register from
6877 being used for other purposes in the preceding functions.
6879 Global register variables may not have initial values, because an
6880 executable file has no means to supply initial contents for a register.
6882 On the SPARC, there are reports that g3 @dots{} g7 are suitable
6883 registers, but certain library functions, such as @code{getwd}, as well
6884 as the subroutines for division and remainder, modify g3 and g4. g1 and
6885 g2 are local temporaries.
6887 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
6888 Of course, it does not do to use more than a few of those.
6890 @node Local Reg Vars
6891 @subsection Specifying Registers for Local Variables
6892 @cindex local variables, specifying registers
6893 @cindex specifying registers for local variables
6894 @cindex registers for local variables
6896 You can define a local register variable with a specified register
6900 register int *foo asm ("a5");
6904 Here @code{a5} is the name of the register that should be used. Note
6905 that this is the same syntax used for defining global register
6906 variables, but for a local variable it appears within a function.
6908 Naturally the register name is cpu-dependent, but this is not a
6909 problem, since specific registers are most often useful with explicit
6910 assembler instructions (@pxref{Extended Asm}). Both of these things
6911 generally require that you conditionalize your program according to
6914 In addition, operating systems on one type of cpu may differ in how they
6915 name the registers; then you need additional conditionals. For
6916 example, some 68000 operating systems call this register @code{%a5}.
6918 Defining such a register variable does not reserve the register; it
6919 remains available for other uses in places where flow control determines
6920 the variable's value is not live.
6922 This option does not guarantee that GCC generates code that has
6923 this variable in the register you specify at all times. You may not
6924 code an explicit reference to this register in the @emph{assembler
6925 instruction template} part of an @code{asm} statement and assume it
6926 always refers to this variable. However, using the variable as an
6927 @code{asm} @emph{operand} guarantees that the specified register is used
6930 Stores into local register variables may be deleted when they appear to be dead
6931 according to dataflow analysis. References to local register variables may
6932 be deleted or moved or simplified.
6934 As for global register variables, it's recommended that you choose a
6935 register that is normally saved and restored by function calls on
6936 your machine, so that library routines will not clobber it. A common
6937 pitfall is to initialize multiple call-clobbered registers with
6938 arbitrary expressions, where a function call or library call for an
6939 arithmetic operator overwrites a register value from a previous
6940 assignment, for example @code{r0} below:
6942 register int *p1 asm ("r0") = @dots{};
6943 register int *p2 asm ("r1") = @dots{};
6947 In those cases, a solution is to use a temporary variable for
6948 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
6950 @node Alternate Keywords
6951 @section Alternate Keywords
6952 @cindex alternate keywords
6953 @cindex keywords, alternate
6955 @option{-ansi} and the various @option{-std} options disable certain
6956 keywords. This causes trouble when you want to use GNU C extensions, or
6957 a general-purpose header file that should be usable by all programs,
6958 including ISO C programs. The keywords @code{asm}, @code{typeof} and
6959 @code{inline} are not available in programs compiled with
6960 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
6961 program compiled with @option{-std=c99} or @option{-std=c11}). The
6963 @code{restrict} is only available when @option{-std=gnu99} (which will
6964 eventually be the default) or @option{-std=c99} (or the equivalent
6965 @option{-std=iso9899:1999}), or an option for a later standard
6968 The way to solve these problems is to put @samp{__} at the beginning and
6969 end of each problematical keyword. For example, use @code{__asm__}
6970 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
6972 Other C compilers won't accept these alternative keywords; if you want to
6973 compile with another compiler, you can define the alternate keywords as
6974 macros to replace them with the customary keywords. It looks like this:
6982 @findex __extension__
6984 @option{-pedantic} and other options cause warnings for many GNU C extensions.
6986 prevent such warnings within one expression by writing
6987 @code{__extension__} before the expression. @code{__extension__} has no
6988 effect aside from this.
6990 @node Incomplete Enums
6991 @section Incomplete @code{enum} Types
6993 You can define an @code{enum} tag without specifying its possible values.
6994 This results in an incomplete type, much like what you get if you write
6995 @code{struct foo} without describing the elements. A later declaration
6996 that does specify the possible values completes the type.
6998 You can't allocate variables or storage using the type while it is
6999 incomplete. However, you can work with pointers to that type.
7001 This extension may not be very useful, but it makes the handling of
7002 @code{enum} more consistent with the way @code{struct} and @code{union}
7005 This extension is not supported by GNU C++.
7007 @node Function Names
7008 @section Function Names as Strings
7009 @cindex @code{__func__} identifier
7010 @cindex @code{__FUNCTION__} identifier
7011 @cindex @code{__PRETTY_FUNCTION__} identifier
7013 GCC provides three magic variables that hold the name of the current
7014 function, as a string. The first of these is @code{__func__}, which
7015 is part of the C99 standard:
7017 The identifier @code{__func__} is implicitly declared by the translator
7018 as if, immediately following the opening brace of each function
7019 definition, the declaration
7022 static const char __func__[] = "function-name";
7026 appeared, where function-name is the name of the lexically-enclosing
7027 function. This name is the unadorned name of the function.
7029 @code{__FUNCTION__} is another name for @code{__func__}. Older
7030 versions of GCC recognize only this name. However, it is not
7031 standardized. For maximum portability, we recommend you use
7032 @code{__func__}, but provide a fallback definition with the
7036 #if __STDC_VERSION__ < 199901L
7038 # define __func__ __FUNCTION__
7040 # define __func__ "<unknown>"
7045 In C, @code{__PRETTY_FUNCTION__} is yet another name for
7046 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
7047 the type signature of the function as well as its bare name. For
7048 example, this program:
7052 extern int printf (char *, ...);
7059 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
7060 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
7078 __PRETTY_FUNCTION__ = void a::sub(int)
7081 These identifiers are not preprocessor macros. In GCC 3.3 and
7082 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
7083 were treated as string literals; they could be used to initialize
7084 @code{char} arrays, and they could be concatenated with other string
7085 literals. GCC 3.4 and later treat them as variables, like
7086 @code{__func__}. In C++, @code{__FUNCTION__} and
7087 @code{__PRETTY_FUNCTION__} have always been variables.
7089 @node Return Address
7090 @section Getting the Return or Frame Address of a Function
7092 These functions may be used to get information about the callers of a
7095 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
7096 This function returns the return address of the current function, or of
7097 one of its callers. The @var{level} argument is number of frames to
7098 scan up the call stack. A value of @code{0} yields the return address
7099 of the current function, a value of @code{1} yields the return address
7100 of the caller of the current function, and so forth. When inlining
7101 the expected behavior is that the function returns the address of
7102 the function that is returned to. To work around this behavior use
7103 the @code{noinline} function attribute.
7105 The @var{level} argument must be a constant integer.
7107 On some machines it may be impossible to determine the return address of
7108 any function other than the current one; in such cases, or when the top
7109 of the stack has been reached, this function returns @code{0} or a
7110 random value. In addition, @code{__builtin_frame_address} may be used
7111 to determine if the top of the stack has been reached.
7113 Additional post-processing of the returned value may be needed, see
7114 @code{__builtin_extract_return_addr}.
7116 This function should only be used with a nonzero argument for debugging
7120 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
7121 The address as returned by @code{__builtin_return_address} may have to be fed
7122 through this function to get the actual encoded address. For example, on the
7123 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
7124 platforms an offset has to be added for the true next instruction to be
7127 If no fixup is needed, this function simply passes through @var{addr}.
7130 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
7131 This function does the reverse of @code{__builtin_extract_return_addr}.
7134 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
7135 This function is similar to @code{__builtin_return_address}, but it
7136 returns the address of the function frame rather than the return address
7137 of the function. Calling @code{__builtin_frame_address} with a value of
7138 @code{0} yields the frame address of the current function, a value of
7139 @code{1} yields the frame address of the caller of the current function,
7142 The frame is the area on the stack that holds local variables and saved
7143 registers. The frame address is normally the address of the first word
7144 pushed on to the stack by the function. However, the exact definition
7145 depends upon the processor and the calling convention. If the processor
7146 has a dedicated frame pointer register, and the function has a frame,
7147 then @code{__builtin_frame_address} returns the value of the frame
7150 On some machines it may be impossible to determine the frame address of
7151 any function other than the current one; in such cases, or when the top
7152 of the stack has been reached, this function returns @code{0} if
7153 the first frame pointer is properly initialized by the startup code.
7155 This function should only be used with a nonzero argument for debugging
7159 @node Vector Extensions
7160 @section Using Vector Instructions through Built-in Functions
7162 On some targets, the instruction set contains SIMD vector instructions which
7163 operate on multiple values contained in one large register at the same time.
7164 For example, on the i386 the MMX, 3DNow!@: and SSE extensions can be used
7167 The first step in using these extensions is to provide the necessary data
7168 types. This should be done using an appropriate @code{typedef}:
7171 typedef int v4si __attribute__ ((vector_size (16)));
7175 The @code{int} type specifies the base type, while the attribute specifies
7176 the vector size for the variable, measured in bytes. For example, the
7177 declaration above causes the compiler to set the mode for the @code{v4si}
7178 type to be 16 bytes wide and divided into @code{int} sized units. For
7179 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
7180 corresponding mode of @code{foo} is @acronym{V4SI}.
7182 The @code{vector_size} attribute is only applicable to integral and
7183 float scalars, although arrays, pointers, and function return values
7184 are allowed in conjunction with this construct. Only sizes that are
7185 a power of two are currently allowed.
7187 All the basic integer types can be used as base types, both as signed
7188 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
7189 @code{long long}. In addition, @code{float} and @code{double} can be
7190 used to build floating-point vector types.
7192 Specifying a combination that is not valid for the current architecture
7193 causes GCC to synthesize the instructions using a narrower mode.
7194 For example, if you specify a variable of type @code{V4SI} and your
7195 architecture does not allow for this specific SIMD type, GCC
7196 produces code that uses 4 @code{SIs}.
7198 The types defined in this manner can be used with a subset of normal C
7199 operations. Currently, GCC allows using the following operators
7200 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
7202 The operations behave like C++ @code{valarrays}. Addition is defined as
7203 the addition of the corresponding elements of the operands. For
7204 example, in the code below, each of the 4 elements in @var{a} is
7205 added to the corresponding 4 elements in @var{b} and the resulting
7206 vector is stored in @var{c}.
7209 typedef int v4si __attribute__ ((vector_size (16)));
7216 Subtraction, multiplication, division, and the logical operations
7217 operate in a similar manner. Likewise, the result of using the unary
7218 minus or complement operators on a vector type is a vector whose
7219 elements are the negative or complemented values of the corresponding
7220 elements in the operand.
7222 It is possible to use shifting operators @code{<<}, @code{>>} on
7223 integer-type vectors. The operation is defined as following: @code{@{a0,
7224 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
7225 @dots{}, an >> bn@}}@. Vector operands must have the same number of
7228 For convenience, it is allowed to use a binary vector operation
7229 where one operand is a scalar. In that case the compiler transforms
7230 the scalar operand into a vector where each element is the scalar from
7231 the operation. The transformation happens only if the scalar could be
7232 safely converted to the vector-element type.
7233 Consider the following code.
7236 typedef int v4si __attribute__ ((vector_size (16)));
7241 a = b + 1; /* a = b + @{1,1,1,1@}; */
7242 a = 2 * b; /* a = @{2,2,2,2@} * b; */
7244 a = l + a; /* Error, cannot convert long to int. */
7247 Vectors can be subscripted as if the vector were an array with
7248 the same number of elements and base type. Out of bound accesses
7249 invoke undefined behavior at run time. Warnings for out of bound
7250 accesses for vector subscription can be enabled with
7251 @option{-Warray-bounds}.
7253 Vector comparison is supported with standard comparison
7254 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
7255 vector expressions of integer-type or real-type. Comparison between
7256 integer-type vectors and real-type vectors are not supported. The
7257 result of the comparison is a vector of the same width and number of
7258 elements as the comparison operands with a signed integral element
7261 Vectors are compared element-wise producing 0 when comparison is false
7262 and -1 (constant of the appropriate type where all bits are set)
7263 otherwise. Consider the following example.
7266 typedef int v4si __attribute__ ((vector_size (16)));
7268 v4si a = @{1,2,3,4@};
7269 v4si b = @{3,2,1,4@};
7272 c = a > b; /* The result would be @{0, 0,-1, 0@} */
7273 c = a == b; /* The result would be @{0,-1, 0,-1@} */
7276 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
7277 @code{b} and @code{c} are vectors of the same type and @code{a} is an
7278 integer vector with the same number of elements of the same size as @code{b}
7279 and @code{c}, computes all three arguments and creates a vector
7280 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
7281 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
7282 As in the case of binary operations, this syntax is also accepted when
7283 one of @code{b} or @code{c} is a scalar that is then transformed into a
7284 vector. If both @code{b} and @code{c} are scalars and the type of
7285 @code{true?b:c} has the same size as the element type of @code{a}, then
7286 @code{b} and @code{c} are converted to a vector type whose elements have
7287 this type and with the same number of elements as @code{a}.
7289 Vector shuffling is available using functions
7290 @code{__builtin_shuffle (vec, mask)} and
7291 @code{__builtin_shuffle (vec0, vec1, mask)}.
7292 Both functions construct a permutation of elements from one or two
7293 vectors and return a vector of the same type as the input vector(s).
7294 The @var{mask} is an integral vector with the same width (@var{W})
7295 and element count (@var{N}) as the output vector.
7297 The elements of the input vectors are numbered in memory ordering of
7298 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
7299 elements of @var{mask} are considered modulo @var{N} in the single-operand
7300 case and modulo @math{2*@var{N}} in the two-operand case.
7302 Consider the following example,
7305 typedef int v4si __attribute__ ((vector_size (16)));
7307 v4si a = @{1,2,3,4@};
7308 v4si b = @{5,6,7,8@};
7309 v4si mask1 = @{0,1,1,3@};
7310 v4si mask2 = @{0,4,2,5@};
7313 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
7314 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
7317 Note that @code{__builtin_shuffle} is intentionally semantically
7318 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
7320 You can declare variables and use them in function calls and returns, as
7321 well as in assignments and some casts. You can specify a vector type as
7322 a return type for a function. Vector types can also be used as function
7323 arguments. It is possible to cast from one vector type to another,
7324 provided they are of the same size (in fact, you can also cast vectors
7325 to and from other datatypes of the same size).
7327 You cannot operate between vectors of different lengths or different
7328 signedness without a cast.
7332 @findex __builtin_offsetof
7334 GCC implements for both C and C++ a syntactic extension to implement
7335 the @code{offsetof} macro.
7339 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
7341 offsetof_member_designator:
7343 | offsetof_member_designator "." @code{identifier}
7344 | offsetof_member_designator "[" @code{expr} "]"
7347 This extension is sufficient such that
7350 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
7354 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
7355 may be dependent. In either case, @var{member} may consist of a single
7356 identifier, or a sequence of member accesses and array references.
7358 @node __sync Builtins
7359 @section Legacy __sync Built-in Functions for Atomic Memory Access
7361 The following built-in functions
7362 are intended to be compatible with those described
7363 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
7364 section 7.4. As such, they depart from the normal GCC practice of using
7365 the @samp{__builtin_} prefix, and further that they are overloaded such that
7366 they work on multiple types.
7368 The definition given in the Intel documentation allows only for the use of
7369 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
7370 counterparts. GCC allows any integral scalar or pointer type that is
7371 1, 2, 4 or 8 bytes in length.
7373 Not all operations are supported by all target processors. If a particular
7374 operation cannot be implemented on the target processor, a warning is
7375 generated and a call an external function is generated. The external
7376 function carries the same name as the built-in version,
7377 with an additional suffix
7378 @samp{_@var{n}} where @var{n} is the size of the data type.
7380 @c ??? Should we have a mechanism to suppress this warning? This is almost
7381 @c useful for implementing the operation under the control of an external
7384 In most cases, these built-in functions are considered a @dfn{full barrier}.
7386 no memory operand is moved across the operation, either forward or
7387 backward. Further, instructions are issued as necessary to prevent the
7388 processor from speculating loads across the operation and from queuing stores
7389 after the operation.
7391 All of the routines are described in the Intel documentation to take
7392 ``an optional list of variables protected by the memory barrier''. It's
7393 not clear what is meant by that; it could mean that @emph{only} the
7394 following variables are protected, or it could mean that these variables
7395 should in addition be protected. At present GCC ignores this list and
7396 protects all variables that are globally accessible. If in the future
7397 we make some use of this list, an empty list will continue to mean all
7398 globally accessible variables.
7401 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
7402 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
7403 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
7404 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
7405 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
7406 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
7407 @findex __sync_fetch_and_add
7408 @findex __sync_fetch_and_sub
7409 @findex __sync_fetch_and_or
7410 @findex __sync_fetch_and_and
7411 @findex __sync_fetch_and_xor
7412 @findex __sync_fetch_and_nand
7413 These built-in functions perform the operation suggested by the name, and
7414 returns the value that had previously been in memory. That is,
7417 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
7418 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
7421 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
7422 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
7424 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
7425 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
7426 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
7427 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
7428 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
7429 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
7430 @findex __sync_add_and_fetch
7431 @findex __sync_sub_and_fetch
7432 @findex __sync_or_and_fetch
7433 @findex __sync_and_and_fetch
7434 @findex __sync_xor_and_fetch
7435 @findex __sync_nand_and_fetch
7436 These built-in functions perform the operation suggested by the name, and
7437 return the new value. That is,
7440 @{ *ptr @var{op}= value; return *ptr; @}
7441 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
7444 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
7445 as @code{*ptr = ~(*ptr & value)} instead of
7446 @code{*ptr = ~*ptr & value}.
7448 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
7449 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
7450 @findex __sync_bool_compare_and_swap
7451 @findex __sync_val_compare_and_swap
7452 These built-in functions perform an atomic compare and swap.
7453 That is, if the current
7454 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
7457 The ``bool'' version returns true if the comparison is successful and
7458 @var{newval} is written. The ``val'' version returns the contents
7459 of @code{*@var{ptr}} before the operation.
7461 @item __sync_synchronize (...)
7462 @findex __sync_synchronize
7463 This built-in function issues a full memory barrier.
7465 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
7466 @findex __sync_lock_test_and_set
7467 This built-in function, as described by Intel, is not a traditional test-and-set
7468 operation, but rather an atomic exchange operation. It writes @var{value}
7469 into @code{*@var{ptr}}, and returns the previous contents of
7472 Many targets have only minimal support for such locks, and do not support
7473 a full exchange operation. In this case, a target may support reduced
7474 functionality here by which the @emph{only} valid value to store is the
7475 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
7476 is implementation defined.
7478 This built-in function is not a full barrier,
7479 but rather an @dfn{acquire barrier}.
7480 This means that references after the operation cannot move to (or be
7481 speculated to) before the operation, but previous memory stores may not
7482 be globally visible yet, and previous memory loads may not yet be
7485 @item void __sync_lock_release (@var{type} *ptr, ...)
7486 @findex __sync_lock_release
7487 This built-in function releases the lock acquired by
7488 @code{__sync_lock_test_and_set}.
7489 Normally this means writing the constant 0 to @code{*@var{ptr}}.
7491 This built-in function is not a full barrier,
7492 but rather a @dfn{release barrier}.
7493 This means that all previous memory stores are globally visible, and all
7494 previous memory loads have been satisfied, but following memory reads
7495 are not prevented from being speculated to before the barrier.
7498 @node __atomic Builtins
7499 @section Built-in functions for memory model aware atomic operations
7501 The following built-in functions approximately match the requirements for
7502 C++11 memory model. Many are similar to the @samp{__sync} prefixed built-in
7503 functions, but all also have a memory model parameter. These are all
7504 identified by being prefixed with @samp{__atomic}, and most are overloaded
7505 such that they work with multiple types.
7507 GCC allows any integral scalar or pointer type that is 1, 2, 4, or 8
7508 bytes in length. 16-byte integral types are also allowed if
7509 @samp{__int128} (@pxref{__int128}) is supported by the architecture.
7511 Target architectures are encouraged to provide their own patterns for
7512 each of these built-in functions. If no target is provided, the original
7513 non-memory model set of @samp{__sync} atomic built-in functions are
7514 utilized, along with any required synchronization fences surrounding it in
7515 order to achieve the proper behavior. Execution in this case is subject
7516 to the same restrictions as those built-in functions.
7518 If there is no pattern or mechanism to provide a lock free instruction
7519 sequence, a call is made to an external routine with the same parameters
7520 to be resolved at run time.
7522 The four non-arithmetic functions (load, store, exchange, and
7523 compare_exchange) all have a generic version as well. This generic
7524 version works on any data type. If the data type size maps to one
7525 of the integral sizes that may have lock free support, the generic
7526 version utilizes the lock free built-in function. Otherwise an
7527 external call is left to be resolved at run time. This external call is
7528 the same format with the addition of a @samp{size_t} parameter inserted
7529 as the first parameter indicating the size of the object being pointed to.
7530 All objects must be the same size.
7532 There are 6 different memory models that can be specified. These map
7533 to the same names in the C++11 standard. Refer there or to the
7534 @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki on
7535 atomic synchronization} for more detailed definitions. These memory
7536 models integrate both barriers to code motion as well as synchronization
7537 requirements with other threads. These are listed in approximately
7538 ascending order of strength. It is also possible to use target specific
7539 flags for memory model flags, like Hardware Lock Elision.
7542 @item __ATOMIC_RELAXED
7543 No barriers or synchronization.
7544 @item __ATOMIC_CONSUME
7545 Data dependency only for both barrier and synchronization with another
7547 @item __ATOMIC_ACQUIRE
7548 Barrier to hoisting of code and synchronizes with release (or stronger)
7549 semantic stores from another thread.
7550 @item __ATOMIC_RELEASE
7551 Barrier to sinking of code and synchronizes with acquire (or stronger)
7552 semantic loads from another thread.
7553 @item __ATOMIC_ACQ_REL
7554 Full barrier in both directions and synchronizes with acquire loads and
7555 release stores in another thread.
7556 @item __ATOMIC_SEQ_CST
7557 Full barrier in both directions and synchronizes with acquire loads and
7558 release stores in all threads.
7561 When implementing patterns for these built-in functions, the memory model
7562 parameter can be ignored as long as the pattern implements the most
7563 restrictive @code{__ATOMIC_SEQ_CST} model. Any of the other memory models
7564 execute correctly with this memory model but they may not execute as
7565 efficiently as they could with a more appropriate implementation of the
7566 relaxed requirements.
7568 Note that the C++11 standard allows for the memory model parameter to be
7569 determined at run time rather than at compile time. These built-in
7570 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
7571 than invoke a runtime library call or inline a switch statement. This is
7572 standard compliant, safe, and the simplest approach for now.
7574 The memory model parameter is a signed int, but only the lower 8 bits are
7575 reserved for the memory model. The remainder of the signed int is reserved
7576 for future use and should be 0. Use of the predefined atomic values
7577 ensures proper usage.
7579 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memmodel)
7580 This built-in function implements an atomic load operation. It returns the
7581 contents of @code{*@var{ptr}}.
7583 The valid memory model variants are
7584 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7585 and @code{__ATOMIC_CONSUME}.
7589 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memmodel)
7590 This is the generic version of an atomic load. It returns the
7591 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
7595 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memmodel)
7596 This built-in function implements an atomic store operation. It writes
7597 @code{@var{val}} into @code{*@var{ptr}}.
7599 The valid memory model variants are
7600 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
7604 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memmodel)
7605 This is the generic version of an atomic store. It stores the value
7606 of @code{*@var{val}} into @code{*@var{ptr}}.
7610 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memmodel)
7611 This built-in function implements an atomic exchange operation. It writes
7612 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
7615 The valid memory model variants are
7616 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7617 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
7621 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memmodel)
7622 This is the generic version of an atomic exchange. It stores the
7623 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
7624 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
7628 @deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memmodel, int failure_memmodel)
7629 This built-in function implements an atomic compare and exchange operation.
7630 This compares the contents of @code{*@var{ptr}} with the contents of
7631 @code{*@var{expected}} and if equal, writes @var{desired} into
7632 @code{*@var{ptr}}. If they are not equal, the current contents of
7633 @code{*@var{ptr}} is written into @code{*@var{expected}}. @var{weak} is true
7634 for weak compare_exchange, and false for the strong variation. Many targets
7635 only offer the strong variation and ignore the parameter. When in doubt, use
7636 the strong variation.
7638 True is returned if @var{desired} is written into
7639 @code{*@var{ptr}} and the execution is considered to conform to the
7640 memory model specified by @var{success_memmodel}. There are no
7641 restrictions on what memory model can be used here.
7643 False is returned otherwise, and the execution is considered to conform
7644 to @var{failure_memmodel}. This memory model cannot be
7645 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
7646 stronger model than that specified by @var{success_memmodel}.
7650 @deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memmodel, int failure_memmodel)
7651 This built-in function implements the generic version of
7652 @code{__atomic_compare_exchange}. The function is virtually identical to
7653 @code{__atomic_compare_exchange_n}, except the desired value is also a
7658 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7659 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7660 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7661 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7662 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7663 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7664 These built-in functions perform the operation suggested by the name, and
7665 return the result of the operation. That is,
7668 @{ *ptr @var{op}= val; return *ptr; @}
7671 All memory models are valid.
7675 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memmodel)
7676 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memmodel)
7677 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memmodel)
7678 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memmodel)
7679 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memmodel)
7680 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memmodel)
7681 These built-in functions perform the operation suggested by the name, and
7682 return the value that had previously been in @code{*@var{ptr}}. That is,
7685 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
7688 All memory models are valid.
7692 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memmodel)
7694 This built-in function performs an atomic test-and-set operation on
7695 the byte at @code{*@var{ptr}}. The byte is set to some implementation
7696 defined nonzero ``set'' value and the return value is @code{true} if and only
7697 if the previous contents were ``set''.
7698 It should be only used for operands of type @code{bool} or @code{char}. For
7699 other types only part of the value may be set.
7701 All memory models are valid.
7705 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memmodel)
7707 This built-in function performs an atomic clear operation on
7708 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
7709 It should be only used for operands of type @code{bool} or @code{char} and
7710 in conjunction with @code{__atomic_test_and_set}.
7711 For other types it may only clear partially. If the type is not @code{bool}
7712 prefer using @code{__atomic_store}.
7714 The valid memory model variants are
7715 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
7716 @code{__ATOMIC_RELEASE}.
7720 @deftypefn {Built-in Function} void __atomic_thread_fence (int memmodel)
7722 This built-in function acts as a synchronization fence between threads
7723 based on the specified memory model.
7725 All memory orders are valid.
7729 @deftypefn {Built-in Function} void __atomic_signal_fence (int memmodel)
7731 This built-in function acts as a synchronization fence between a thread
7732 and signal handlers based in the same thread.
7734 All memory orders are valid.
7738 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
7740 This built-in function returns true if objects of @var{size} bytes always
7741 generate lock free atomic instructions for the target architecture.
7742 @var{size} must resolve to a compile-time constant and the result also
7743 resolves to a compile-time constant.
7745 @var{ptr} is an optional pointer to the object that may be used to determine
7746 alignment. A value of 0 indicates typical alignment should be used. The
7747 compiler may also ignore this parameter.
7750 if (_atomic_always_lock_free (sizeof (long long), 0))
7755 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
7757 This built-in function returns true if objects of @var{size} bytes always
7758 generate lock free atomic instructions for the target architecture. If
7759 it is not known to be lock free a call is made to a runtime routine named
7760 @code{__atomic_is_lock_free}.
7762 @var{ptr} is an optional pointer to the object that may be used to determine
7763 alignment. A value of 0 indicates typical alignment should be used. The
7764 compiler may also ignore this parameter.
7767 @node x86 specific memory model extensions for transactional memory
7768 @section x86 specific memory model extensions for transactional memory
7770 The i386 architecture supports additional memory ordering flags
7771 to mark lock critical sections for hardware lock elision.
7772 These must be specified in addition to an existing memory model to
7776 @item __ATOMIC_HLE_ACQUIRE
7777 Start lock elision on a lock variable.
7778 Memory model must be @code{__ATOMIC_ACQUIRE} or stronger.
7779 @item __ATOMIC_HLE_RELEASE
7780 End lock elision on a lock variable.
7781 Memory model must be @code{__ATOMIC_RELEASE} or stronger.
7784 When a lock acquire fails it is required for good performance to abort
7785 the transaction quickly. This can be done with a @code{_mm_pause}
7788 #include <immintrin.h> // For _mm_pause
7792 /* Acquire lock with lock elision */
7793 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
7794 _mm_pause(); /* Abort failed transaction */
7796 /* Free lock with lock elision */
7797 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
7800 @node Object Size Checking
7801 @section Object Size Checking Built-in Functions
7802 @findex __builtin_object_size
7803 @findex __builtin___memcpy_chk
7804 @findex __builtin___mempcpy_chk
7805 @findex __builtin___memmove_chk
7806 @findex __builtin___memset_chk
7807 @findex __builtin___strcpy_chk
7808 @findex __builtin___stpcpy_chk
7809 @findex __builtin___strncpy_chk
7810 @findex __builtin___strcat_chk
7811 @findex __builtin___strncat_chk
7812 @findex __builtin___sprintf_chk
7813 @findex __builtin___snprintf_chk
7814 @findex __builtin___vsprintf_chk
7815 @findex __builtin___vsnprintf_chk
7816 @findex __builtin___printf_chk
7817 @findex __builtin___vprintf_chk
7818 @findex __builtin___fprintf_chk
7819 @findex __builtin___vfprintf_chk
7821 GCC implements a limited buffer overflow protection mechanism
7822 that can prevent some buffer overflow attacks.
7824 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
7825 is a built-in construct that returns a constant number of bytes from
7826 @var{ptr} to the end of the object @var{ptr} pointer points to
7827 (if known at compile time). @code{__builtin_object_size} never evaluates
7828 its arguments for side-effects. If there are any side-effects in them, it
7829 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7830 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
7831 point to and all of them are known at compile time, the returned number
7832 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
7833 0 and minimum if nonzero. If it is not possible to determine which objects
7834 @var{ptr} points to at compile time, @code{__builtin_object_size} should
7835 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7836 for @var{type} 2 or 3.
7838 @var{type} is an integer constant from 0 to 3. If the least significant
7839 bit is clear, objects are whole variables, if it is set, a closest
7840 surrounding subobject is considered the object a pointer points to.
7841 The second bit determines if maximum or minimum of remaining bytes
7845 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
7846 char *p = &var.buf1[1], *q = &var.b;
7848 /* Here the object p points to is var. */
7849 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
7850 /* The subobject p points to is var.buf1. */
7851 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
7852 /* The object q points to is var. */
7853 assert (__builtin_object_size (q, 0)
7854 == (char *) (&var + 1) - (char *) &var.b);
7855 /* The subobject q points to is var.b. */
7856 assert (__builtin_object_size (q, 1) == sizeof (var.b));
7860 There are built-in functions added for many common string operation
7861 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
7862 built-in is provided. This built-in has an additional last argument,
7863 which is the number of bytes remaining in object the @var{dest}
7864 argument points to or @code{(size_t) -1} if the size is not known.
7866 The built-in functions are optimized into the normal string functions
7867 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
7868 it is known at compile time that the destination object will not
7869 be overflown. If the compiler can determine at compile time the
7870 object will be always overflown, it issues a warning.
7872 The intended use can be e.g.@:
7876 #define bos0(dest) __builtin_object_size (dest, 0)
7877 #define memcpy(dest, src, n) \
7878 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
7882 /* It is unknown what object p points to, so this is optimized
7883 into plain memcpy - no checking is possible. */
7884 memcpy (p, "abcde", n);
7885 /* Destination is known and length too. It is known at compile
7886 time there will be no overflow. */
7887 memcpy (&buf[5], "abcde", 5);
7888 /* Destination is known, but the length is not known at compile time.
7889 This will result in __memcpy_chk call that can check for overflow
7891 memcpy (&buf[5], "abcde", n);
7892 /* Destination is known and it is known at compile time there will
7893 be overflow. There will be a warning and __memcpy_chk call that
7894 will abort the program at run time. */
7895 memcpy (&buf[6], "abcde", 5);
7898 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
7899 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
7900 @code{strcat} and @code{strncat}.
7902 There are also checking built-in functions for formatted output functions.
7904 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
7905 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7906 const char *fmt, ...);
7907 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
7909 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7910 const char *fmt, va_list ap);
7913 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
7914 etc.@: functions and can contain implementation specific flags on what
7915 additional security measures the checking function might take, such as
7916 handling @code{%n} differently.
7918 The @var{os} argument is the object size @var{s} points to, like in the
7919 other built-in functions. There is a small difference in the behavior
7920 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
7921 optimized into the non-checking functions only if @var{flag} is 0, otherwise
7922 the checking function is called with @var{os} argument set to
7925 In addition to this, there are checking built-in functions
7926 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
7927 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
7928 These have just one additional argument, @var{flag}, right before
7929 format string @var{fmt}. If the compiler is able to optimize them to
7930 @code{fputc} etc.@: functions, it does, otherwise the checking function
7931 is called and the @var{flag} argument passed to it.
7933 @node Cilk Plus Builtins
7934 @section Cilk Plus C/C++ language extension Built-in Functions.
7936 GCC provides support for the following built-in reduction funtions if Cilk Plus
7937 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
7940 @item __sec_implicit_index
7942 @item __sec_reduce_add
7943 @item __sec_reduce_all_nonzero
7944 @item __sec_reduce_all_zero
7945 @item __sec_reduce_any_nonzero
7946 @item __sec_reduce_any_zero
7947 @item __sec_reduce_max
7948 @item __sec_reduce_min
7949 @item __sec_reduce_max_ind
7950 @item __sec_reduce_min_ind
7951 @item __sec_reduce_mul
7952 @item __sec_reduce_mutating
7955 Further details and examples about these built-in functions are described
7956 in the Cilk Plus language manual which can be found at
7957 @uref{http://www.cilkplus.org}.
7959 @node Other Builtins
7960 @section Other Built-in Functions Provided by GCC
7961 @cindex built-in functions
7962 @findex __builtin_fpclassify
7963 @findex __builtin_isfinite
7964 @findex __builtin_isnormal
7965 @findex __builtin_isgreater
7966 @findex __builtin_isgreaterequal
7967 @findex __builtin_isinf_sign
7968 @findex __builtin_isless
7969 @findex __builtin_islessequal
7970 @findex __builtin_islessgreater
7971 @findex __builtin_isunordered
7972 @findex __builtin_powi
7973 @findex __builtin_powif
7974 @findex __builtin_powil
8132 @findex fprintf_unlocked
8134 @findex fputs_unlocked
8251 @findex printf_unlocked
8283 @findex significandf
8284 @findex significandl
8355 GCC provides a large number of built-in functions other than the ones
8356 mentioned above. Some of these are for internal use in the processing
8357 of exceptions or variable-length argument lists and are not
8358 documented here because they may change from time to time; we do not
8359 recommend general use of these functions.
8361 The remaining functions are provided for optimization purposes.
8363 @opindex fno-builtin
8364 GCC includes built-in versions of many of the functions in the standard
8365 C library. The versions prefixed with @code{__builtin_} are always
8366 treated as having the same meaning as the C library function even if you
8367 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
8368 Many of these functions are only optimized in certain cases; if they are
8369 not optimized in a particular case, a call to the library function is
8374 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
8375 @option{-std=c99} or @option{-std=c11}), the functions
8376 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
8377 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
8378 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
8379 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
8380 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
8381 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
8382 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
8383 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
8384 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
8385 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
8386 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
8387 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
8388 @code{signbitd64}, @code{signbitd128}, @code{significandf},
8389 @code{significandl}, @code{significand}, @code{sincosf},
8390 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
8391 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
8392 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
8393 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
8395 may be handled as built-in functions.
8396 All these functions have corresponding versions
8397 prefixed with @code{__builtin_}, which may be used even in strict C90
8400 The ISO C99 functions
8401 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
8402 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
8403 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
8404 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
8405 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
8406 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
8407 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
8408 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
8409 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
8410 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
8411 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
8412 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
8413 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
8414 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
8415 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
8416 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
8417 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
8418 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
8419 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
8420 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
8421 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
8422 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
8423 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
8424 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
8425 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
8426 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
8427 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
8428 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
8429 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
8430 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
8431 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
8432 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
8433 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
8434 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
8435 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
8436 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
8437 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
8438 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
8439 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
8440 are handled as built-in functions
8441 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
8443 There are also built-in versions of the ISO C99 functions
8444 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
8445 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
8446 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
8447 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
8448 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
8449 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
8450 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
8451 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
8452 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
8453 that are recognized in any mode since ISO C90 reserves these names for
8454 the purpose to which ISO C99 puts them. All these functions have
8455 corresponding versions prefixed with @code{__builtin_}.
8457 The ISO C94 functions
8458 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
8459 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
8460 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
8462 are handled as built-in functions
8463 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
8465 The ISO C90 functions
8466 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
8467 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
8468 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
8469 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
8470 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
8471 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
8472 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
8473 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
8474 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
8475 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
8476 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
8477 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
8478 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
8479 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
8480 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
8481 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
8482 are all recognized as built-in functions unless
8483 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
8484 is specified for an individual function). All of these functions have
8485 corresponding versions prefixed with @code{__builtin_}.
8487 GCC provides built-in versions of the ISO C99 floating-point comparison
8488 macros that avoid raising exceptions for unordered operands. They have
8489 the same names as the standard macros ( @code{isgreater},
8490 @code{isgreaterequal}, @code{isless}, @code{islessequal},
8491 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
8492 prefixed. We intend for a library implementor to be able to simply
8493 @code{#define} each standard macro to its built-in equivalent.
8494 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
8495 @code{isinf_sign} and @code{isnormal} built-ins used with
8496 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
8497 built-in functions appear both with and without the @code{__builtin_} prefix.
8499 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
8501 You can use the built-in function @code{__builtin_types_compatible_p} to
8502 determine whether two types are the same.
8504 This built-in function returns 1 if the unqualified versions of the
8505 types @var{type1} and @var{type2} (which are types, not expressions) are
8506 compatible, 0 otherwise. The result of this built-in function can be
8507 used in integer constant expressions.
8509 This built-in function ignores top level qualifiers (e.g., @code{const},
8510 @code{volatile}). For example, @code{int} is equivalent to @code{const
8513 The type @code{int[]} and @code{int[5]} are compatible. On the other
8514 hand, @code{int} and @code{char *} are not compatible, even if the size
8515 of their types, on the particular architecture are the same. Also, the
8516 amount of pointer indirection is taken into account when determining
8517 similarity. Consequently, @code{short *} is not similar to
8518 @code{short **}. Furthermore, two types that are typedefed are
8519 considered compatible if their underlying types are compatible.
8521 An @code{enum} type is not considered to be compatible with another
8522 @code{enum} type even if both are compatible with the same integer
8523 type; this is what the C standard specifies.
8524 For example, @code{enum @{foo, bar@}} is not similar to
8525 @code{enum @{hot, dog@}}.
8527 You typically use this function in code whose execution varies
8528 depending on the arguments' types. For example:
8533 typeof (x) tmp = (x); \
8534 if (__builtin_types_compatible_p (typeof (x), long double)) \
8535 tmp = foo_long_double (tmp); \
8536 else if (__builtin_types_compatible_p (typeof (x), double)) \
8537 tmp = foo_double (tmp); \
8538 else if (__builtin_types_compatible_p (typeof (x), float)) \
8539 tmp = foo_float (tmp); \
8546 @emph{Note:} This construct is only available for C@.
8550 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
8552 You can use the built-in function @code{__builtin_choose_expr} to
8553 evaluate code depending on the value of a constant expression. This
8554 built-in function returns @var{exp1} if @var{const_exp}, which is an
8555 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
8557 This built-in function is analogous to the @samp{? :} operator in C,
8558 except that the expression returned has its type unaltered by promotion
8559 rules. Also, the built-in function does not evaluate the expression
8560 that is not chosen. For example, if @var{const_exp} evaluates to true,
8561 @var{exp2} is not evaluated even if it has side-effects.
8563 This built-in function can return an lvalue if the chosen argument is an
8566 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
8567 type. Similarly, if @var{exp2} is returned, its return type is the same
8574 __builtin_choose_expr ( \
8575 __builtin_types_compatible_p (typeof (x), double), \
8577 __builtin_choose_expr ( \
8578 __builtin_types_compatible_p (typeof (x), float), \
8580 /* @r{The void expression results in a compile-time error} \
8581 @r{when assigning the result to something.} */ \
8585 @emph{Note:} This construct is only available for C@. Furthermore, the
8586 unused expression (@var{exp1} or @var{exp2} depending on the value of
8587 @var{const_exp}) may still generate syntax errors. This may change in
8592 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
8594 The built-in function @code{__builtin_complex} is provided for use in
8595 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
8596 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
8597 real binary floating-point type, and the result has the corresponding
8598 complex type with real and imaginary parts @var{real} and @var{imag}.
8599 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
8600 infinities, NaNs and negative zeros are involved.
8604 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
8605 You can use the built-in function @code{__builtin_constant_p} to
8606 determine if a value is known to be constant at compile time and hence
8607 that GCC can perform constant-folding on expressions involving that
8608 value. The argument of the function is the value to test. The function
8609 returns the integer 1 if the argument is known to be a compile-time
8610 constant and 0 if it is not known to be a compile-time constant. A
8611 return of 0 does not indicate that the value is @emph{not} a constant,
8612 but merely that GCC cannot prove it is a constant with the specified
8613 value of the @option{-O} option.
8615 You typically use this function in an embedded application where
8616 memory is a critical resource. If you have some complex calculation,
8617 you may want it to be folded if it involves constants, but need to call
8618 a function if it does not. For example:
8621 #define Scale_Value(X) \
8622 (__builtin_constant_p (X) \
8623 ? ((X) * SCALE + OFFSET) : Scale (X))
8626 You may use this built-in function in either a macro or an inline
8627 function. However, if you use it in an inlined function and pass an
8628 argument of the function as the argument to the built-in, GCC
8629 never returns 1 when you call the inline function with a string constant
8630 or compound literal (@pxref{Compound Literals}) and does not return 1
8631 when you pass a constant numeric value to the inline function unless you
8632 specify the @option{-O} option.
8634 You may also use @code{__builtin_constant_p} in initializers for static
8635 data. For instance, you can write
8638 static const int table[] = @{
8639 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
8645 This is an acceptable initializer even if @var{EXPRESSION} is not a
8646 constant expression, including the case where
8647 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
8648 folded to a constant but @var{EXPRESSION} contains operands that are
8649 not otherwise permitted in a static initializer (for example,
8650 @code{0 && foo ()}). GCC must be more conservative about evaluating the
8651 built-in in this case, because it has no opportunity to perform
8654 Previous versions of GCC did not accept this built-in in data
8655 initializers. The earliest version where it is completely safe is
8659 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
8660 @opindex fprofile-arcs
8661 You may use @code{__builtin_expect} to provide the compiler with
8662 branch prediction information. In general, you should prefer to
8663 use actual profile feedback for this (@option{-fprofile-arcs}), as
8664 programmers are notoriously bad at predicting how their programs
8665 actually perform. However, there are applications in which this
8666 data is hard to collect.
8668 The return value is the value of @var{exp}, which should be an integral
8669 expression. The semantics of the built-in are that it is expected that
8670 @var{exp} == @var{c}. For example:
8673 if (__builtin_expect (x, 0))
8678 indicates that we do not expect to call @code{foo}, since
8679 we expect @code{x} to be zero. Since you are limited to integral
8680 expressions for @var{exp}, you should use constructions such as
8683 if (__builtin_expect (ptr != NULL, 1))
8688 when testing pointer or floating-point values.
8691 @deftypefn {Built-in Function} void __builtin_trap (void)
8692 This function causes the program to exit abnormally. GCC implements
8693 this function by using a target-dependent mechanism (such as
8694 intentionally executing an illegal instruction) or by calling
8695 @code{abort}. The mechanism used may vary from release to release so
8696 you should not rely on any particular implementation.
8699 @deftypefn {Built-in Function} void __builtin_unreachable (void)
8700 If control flow reaches the point of the @code{__builtin_unreachable},
8701 the program is undefined. It is useful in situations where the
8702 compiler cannot deduce the unreachability of the code.
8704 One such case is immediately following an @code{asm} statement that
8705 either never terminates, or one that transfers control elsewhere
8706 and never returns. In this example, without the
8707 @code{__builtin_unreachable}, GCC issues a warning that control
8708 reaches the end of a non-void function. It also generates code
8709 to return after the @code{asm}.
8712 int f (int c, int v)
8720 asm("jmp error_handler");
8721 __builtin_unreachable ();
8727 Because the @code{asm} statement unconditionally transfers control out
8728 of the function, control never reaches the end of the function
8729 body. The @code{__builtin_unreachable} is in fact unreachable and
8730 communicates this fact to the compiler.
8732 Another use for @code{__builtin_unreachable} is following a call a
8733 function that never returns but that is not declared
8734 @code{__attribute__((noreturn))}, as in this example:
8737 void function_that_never_returns (void);
8747 function_that_never_returns ();
8748 __builtin_unreachable ();
8755 @deftypefn {Built-in Function} void *__builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
8756 This function returns its first argument, and allows the compiler
8757 to assume that the returned pointer is at least @var{align} bytes
8758 aligned. This built-in can have either two or three arguments,
8759 if it has three, the third argument should have integer type, and
8760 if it is nonzero means misalignment offset. For example:
8763 void *x = __builtin_assume_aligned (arg, 16);
8767 means that the compiler can assume @code{x}, set to @code{arg}, is at least
8768 16-byte aligned, while:
8771 void *x = __builtin_assume_aligned (arg, 32, 8);
8775 means that the compiler can assume for @code{x}, set to @code{arg}, that
8776 @code{(char *) x - 8} is 32-byte aligned.
8779 @deftypefn {Built-in Function} int __builtin_LINE ()
8780 This function is the equivalent to the preprocessor @code{__LINE__}
8781 macro and returns the line number of the invocation of the built-in.
8782 In a C++ default argument for a function @var{F}, it gets the line number of
8783 the call to @var{F}.
8786 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
8787 This function is the equivalent to the preprocessor @code{__FUNCTION__}
8788 macro and returns the function name the invocation of the built-in is in.
8791 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
8792 This function is the equivalent to the preprocessor @code{__FILE__}
8793 macro and returns the file name the invocation of the built-in is in.
8794 In a C++ default argument for a function @var{F}, it gets the file name of
8795 the call to @var{F}.
8798 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
8799 This function is used to flush the processor's instruction cache for
8800 the region of memory between @var{begin} inclusive and @var{end}
8801 exclusive. Some targets require that the instruction cache be
8802 flushed, after modifying memory containing code, in order to obtain
8803 deterministic behavior.
8805 If the target does not require instruction cache flushes,
8806 @code{__builtin___clear_cache} has no effect. Otherwise either
8807 instructions are emitted in-line to clear the instruction cache or a
8808 call to the @code{__clear_cache} function in libgcc is made.
8811 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
8812 This function is used to minimize cache-miss latency by moving data into
8813 a cache before it is accessed.
8814 You can insert calls to @code{__builtin_prefetch} into code for which
8815 you know addresses of data in memory that is likely to be accessed soon.
8816 If the target supports them, data prefetch instructions are generated.
8817 If the prefetch is done early enough before the access then the data will
8818 be in the cache by the time it is accessed.
8820 The value of @var{addr} is the address of the memory to prefetch.
8821 There are two optional arguments, @var{rw} and @var{locality}.
8822 The value of @var{rw} is a compile-time constant one or zero; one
8823 means that the prefetch is preparing for a write to the memory address
8824 and zero, the default, means that the prefetch is preparing for a read.
8825 The value @var{locality} must be a compile-time constant integer between
8826 zero and three. A value of zero means that the data has no temporal
8827 locality, so it need not be left in the cache after the access. A value
8828 of three means that the data has a high degree of temporal locality and
8829 should be left in all levels of cache possible. Values of one and two
8830 mean, respectively, a low or moderate degree of temporal locality. The
8834 for (i = 0; i < n; i++)
8837 __builtin_prefetch (&a[i+j], 1, 1);
8838 __builtin_prefetch (&b[i+j], 0, 1);
8843 Data prefetch does not generate faults if @var{addr} is invalid, but
8844 the address expression itself must be valid. For example, a prefetch
8845 of @code{p->next} does not fault if @code{p->next} is not a valid
8846 address, but evaluation faults if @code{p} is not a valid address.
8848 If the target does not support data prefetch, the address expression
8849 is evaluated if it includes side effects but no other code is generated
8850 and GCC does not issue a warning.
8853 @deftypefn {Built-in Function} double __builtin_huge_val (void)
8854 Returns a positive infinity, if supported by the floating-point format,
8855 else @code{DBL_MAX}. This function is suitable for implementing the
8856 ISO C macro @code{HUGE_VAL}.
8859 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
8860 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
8863 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
8864 Similar to @code{__builtin_huge_val}, except the return
8865 type is @code{long double}.
8868 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
8869 This built-in implements the C99 fpclassify functionality. The first
8870 five int arguments should be the target library's notion of the
8871 possible FP classes and are used for return values. They must be
8872 constant values and they must appear in this order: @code{FP_NAN},
8873 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
8874 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
8875 to classify. GCC treats the last argument as type-generic, which
8876 means it does not do default promotion from float to double.
8879 @deftypefn {Built-in Function} double __builtin_inf (void)
8880 Similar to @code{__builtin_huge_val}, except a warning is generated
8881 if the target floating-point format does not support infinities.
8884 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
8885 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
8888 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
8889 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
8892 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
8893 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
8896 @deftypefn {Built-in Function} float __builtin_inff (void)
8897 Similar to @code{__builtin_inf}, except the return type is @code{float}.
8898 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
8901 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
8902 Similar to @code{__builtin_inf}, except the return
8903 type is @code{long double}.
8906 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
8907 Similar to @code{isinf}, except the return value is -1 for
8908 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
8909 Note while the parameter list is an
8910 ellipsis, this function only accepts exactly one floating-point
8911 argument. GCC treats this parameter as type-generic, which means it
8912 does not do default promotion from float to double.
8915 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
8916 This is an implementation of the ISO C99 function @code{nan}.
8918 Since ISO C99 defines this function in terms of @code{strtod}, which we
8919 do not implement, a description of the parsing is in order. The string
8920 is parsed as by @code{strtol}; that is, the base is recognized by
8921 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
8922 in the significand such that the least significant bit of the number
8923 is at the least significant bit of the significand. The number is
8924 truncated to fit the significand field provided. The significand is
8925 forced to be a quiet NaN@.
8927 This function, if given a string literal all of which would have been
8928 consumed by @code{strtol}, is evaluated early enough that it is considered a
8929 compile-time constant.
8932 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
8933 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
8936 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
8937 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
8940 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
8941 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
8944 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
8945 Similar to @code{__builtin_nan}, except the return type is @code{float}.
8948 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
8949 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
8952 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
8953 Similar to @code{__builtin_nan}, except the significand is forced
8954 to be a signaling NaN@. The @code{nans} function is proposed by
8955 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
8958 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
8959 Similar to @code{__builtin_nans}, except the return type is @code{float}.
8962 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
8963 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
8966 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
8967 Returns one plus the index of the least significant 1-bit of @var{x}, or
8968 if @var{x} is zero, returns zero.
8971 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
8972 Returns the number of leading 0-bits in @var{x}, starting at the most
8973 significant bit position. If @var{x} is 0, the result is undefined.
8976 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
8977 Returns the number of trailing 0-bits in @var{x}, starting at the least
8978 significant bit position. If @var{x} is 0, the result is undefined.
8981 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
8982 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
8983 number of bits following the most significant bit that are identical
8984 to it. There are no special cases for 0 or other values.
8987 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
8988 Returns the number of 1-bits in @var{x}.
8991 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
8992 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
8996 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
8997 Similar to @code{__builtin_ffs}, except the argument type is
8998 @code{unsigned long}.
9001 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
9002 Similar to @code{__builtin_clz}, except the argument type is
9003 @code{unsigned long}.
9006 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
9007 Similar to @code{__builtin_ctz}, except the argument type is
9008 @code{unsigned long}.
9011 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
9012 Similar to @code{__builtin_clrsb}, except the argument type is
9016 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
9017 Similar to @code{__builtin_popcount}, except the argument type is
9018 @code{unsigned long}.
9021 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
9022 Similar to @code{__builtin_parity}, except the argument type is
9023 @code{unsigned long}.
9026 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
9027 Similar to @code{__builtin_ffs}, except the argument type is
9028 @code{unsigned long long}.
9031 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
9032 Similar to @code{__builtin_clz}, except the argument type is
9033 @code{unsigned long long}.
9036 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
9037 Similar to @code{__builtin_ctz}, except the argument type is
9038 @code{unsigned long long}.
9041 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
9042 Similar to @code{__builtin_clrsb}, except the argument type is
9046 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
9047 Similar to @code{__builtin_popcount}, except the argument type is
9048 @code{unsigned long long}.
9051 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
9052 Similar to @code{__builtin_parity}, except the argument type is
9053 @code{unsigned long long}.
9056 @deftypefn {Built-in Function} double __builtin_powi (double, int)
9057 Returns the first argument raised to the power of the second. Unlike the
9058 @code{pow} function no guarantees about precision and rounding are made.
9061 @deftypefn {Built-in Function} float __builtin_powif (float, int)
9062 Similar to @code{__builtin_powi}, except the argument and return types
9066 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
9067 Similar to @code{__builtin_powi}, except the argument and return types
9068 are @code{long double}.
9071 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
9072 Returns @var{x} with the order of the bytes reversed; for example,
9073 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
9077 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
9078 Similar to @code{__builtin_bswap16}, except the argument and return types
9082 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
9083 Similar to @code{__builtin_bswap32}, except the argument and return types
9087 @node Target Builtins
9088 @section Built-in Functions Specific to Particular Target Machines
9090 On some target machines, GCC supports many built-in functions specific
9091 to those machines. Generally these generate calls to specific machine
9092 instructions, but allow the compiler to schedule those calls.
9095 * Alpha Built-in Functions::
9096 * Altera Nios II Built-in Functions::
9097 * ARC Built-in Functions::
9098 * ARC SIMD Built-in Functions::
9099 * ARM iWMMXt Built-in Functions::
9100 * ARM NEON Intrinsics::
9101 * ARM ACLE Intrinsics::
9102 * AVR Built-in Functions::
9103 * Blackfin Built-in Functions::
9104 * FR-V Built-in Functions::
9105 * X86 Built-in Functions::
9106 * X86 transactional memory intrinsics::
9107 * MIPS DSP Built-in Functions::
9108 * MIPS Paired-Single Support::
9109 * MIPS Loongson Built-in Functions::
9110 * Other MIPS Built-in Functions::
9111 * MSP430 Built-in Functions::
9112 * NDS32 Built-in Functions::
9113 * picoChip Built-in Functions::
9114 * PowerPC Built-in Functions::
9115 * PowerPC AltiVec/VSX Built-in Functions::
9116 * PowerPC Hardware Transactional Memory Built-in Functions::
9117 * RX Built-in Functions::
9118 * S/390 System z Built-in Functions::
9119 * SH Built-in Functions::
9120 * SPARC VIS Built-in Functions::
9121 * SPU Built-in Functions::
9122 * TI C6X Built-in Functions::
9123 * TILE-Gx Built-in Functions::
9124 * TILEPro Built-in Functions::
9127 @node Alpha Built-in Functions
9128 @subsection Alpha Built-in Functions
9130 These built-in functions are available for the Alpha family of
9131 processors, depending on the command-line switches used.
9133 The following built-in functions are always available. They
9134 all generate the machine instruction that is part of the name.
9137 long __builtin_alpha_implver (void)
9138 long __builtin_alpha_rpcc (void)
9139 long __builtin_alpha_amask (long)
9140 long __builtin_alpha_cmpbge (long, long)
9141 long __builtin_alpha_extbl (long, long)
9142 long __builtin_alpha_extwl (long, long)
9143 long __builtin_alpha_extll (long, long)
9144 long __builtin_alpha_extql (long, long)
9145 long __builtin_alpha_extwh (long, long)
9146 long __builtin_alpha_extlh (long, long)
9147 long __builtin_alpha_extqh (long, long)
9148 long __builtin_alpha_insbl (long, long)
9149 long __builtin_alpha_inswl (long, long)
9150 long __builtin_alpha_insll (long, long)
9151 long __builtin_alpha_insql (long, long)
9152 long __builtin_alpha_inswh (long, long)
9153 long __builtin_alpha_inslh (long, long)
9154 long __builtin_alpha_insqh (long, long)
9155 long __builtin_alpha_mskbl (long, long)
9156 long __builtin_alpha_mskwl (long, long)
9157 long __builtin_alpha_mskll (long, long)
9158 long __builtin_alpha_mskql (long, long)
9159 long __builtin_alpha_mskwh (long, long)
9160 long __builtin_alpha_msklh (long, long)
9161 long __builtin_alpha_mskqh (long, long)
9162 long __builtin_alpha_umulh (long, long)
9163 long __builtin_alpha_zap (long, long)
9164 long __builtin_alpha_zapnot (long, long)
9167 The following built-in functions are always with @option{-mmax}
9168 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
9169 later. They all generate the machine instruction that is part
9173 long __builtin_alpha_pklb (long)
9174 long __builtin_alpha_pkwb (long)
9175 long __builtin_alpha_unpkbl (long)
9176 long __builtin_alpha_unpkbw (long)
9177 long __builtin_alpha_minub8 (long, long)
9178 long __builtin_alpha_minsb8 (long, long)
9179 long __builtin_alpha_minuw4 (long, long)
9180 long __builtin_alpha_minsw4 (long, long)
9181 long __builtin_alpha_maxub8 (long, long)
9182 long __builtin_alpha_maxsb8 (long, long)
9183 long __builtin_alpha_maxuw4 (long, long)
9184 long __builtin_alpha_maxsw4 (long, long)
9185 long __builtin_alpha_perr (long, long)
9188 The following built-in functions are always with @option{-mcix}
9189 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
9190 later. They all generate the machine instruction that is part
9194 long __builtin_alpha_cttz (long)
9195 long __builtin_alpha_ctlz (long)
9196 long __builtin_alpha_ctpop (long)
9199 The following built-in functions are available on systems that use the OSF/1
9200 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
9201 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
9202 @code{rdval} and @code{wrval}.
9205 void *__builtin_thread_pointer (void)
9206 void __builtin_set_thread_pointer (void *)
9209 @node Altera Nios II Built-in Functions
9210 @subsection Altera Nios II Built-in Functions
9212 These built-in functions are available for the Altera Nios II
9213 family of processors.
9215 The following built-in functions are always available. They
9216 all generate the machine instruction that is part of the name.
9219 int __builtin_ldbio (volatile const void *)
9220 int __builtin_ldbuio (volatile const void *)
9221 int __builtin_ldhio (volatile const void *)
9222 int __builtin_ldhuio (volatile const void *)
9223 int __builtin_ldwio (volatile const void *)
9224 void __builtin_stbio (volatile void *, int)
9225 void __builtin_sthio (volatile void *, int)
9226 void __builtin_stwio (volatile void *, int)
9227 void __builtin_sync (void)
9228 int __builtin_rdctl (int)
9229 void __builtin_wrctl (int, int)
9232 The following built-in functions are always available. They
9233 all generate a Nios II Custom Instruction. The name of the
9234 function represents the types that the function takes and
9235 returns. The letter before the @code{n} is the return type
9236 or void if absent. The @code{n} represents the first parameter
9237 to all the custom instructions, the custom instruction number.
9238 The two letters after the @code{n} represent the up to two
9239 parameters to the function.
9241 The letters represent the following data types:
9244 @code{void} for return type and no parameter for parameter types.
9247 @code{int} for return type and parameter type
9250 @code{float} for return type and parameter type
9253 @code{void *} for return type and parameter type
9257 And the function names are:
9259 void __builtin_custom_n (void)
9260 void __builtin_custom_ni (int)
9261 void __builtin_custom_nf (float)
9262 void __builtin_custom_np (void *)
9263 void __builtin_custom_nii (int, int)
9264 void __builtin_custom_nif (int, float)
9265 void __builtin_custom_nip (int, void *)
9266 void __builtin_custom_nfi (float, int)
9267 void __builtin_custom_nff (float, float)
9268 void __builtin_custom_nfp (float, void *)
9269 void __builtin_custom_npi (void *, int)
9270 void __builtin_custom_npf (void *, float)
9271 void __builtin_custom_npp (void *, void *)
9272 int __builtin_custom_in (void)
9273 int __builtin_custom_ini (int)
9274 int __builtin_custom_inf (float)
9275 int __builtin_custom_inp (void *)
9276 int __builtin_custom_inii (int, int)
9277 int __builtin_custom_inif (int, float)
9278 int __builtin_custom_inip (int, void *)
9279 int __builtin_custom_infi (float, int)
9280 int __builtin_custom_inff (float, float)
9281 int __builtin_custom_infp (float, void *)
9282 int __builtin_custom_inpi (void *, int)
9283 int __builtin_custom_inpf (void *, float)
9284 int __builtin_custom_inpp (void *, void *)
9285 float __builtin_custom_fn (void)
9286 float __builtin_custom_fni (int)
9287 float __builtin_custom_fnf (float)
9288 float __builtin_custom_fnp (void *)
9289 float __builtin_custom_fnii (int, int)
9290 float __builtin_custom_fnif (int, float)
9291 float __builtin_custom_fnip (int, void *)
9292 float __builtin_custom_fnfi (float, int)
9293 float __builtin_custom_fnff (float, float)
9294 float __builtin_custom_fnfp (float, void *)
9295 float __builtin_custom_fnpi (void *, int)
9296 float __builtin_custom_fnpf (void *, float)
9297 float __builtin_custom_fnpp (void *, void *)
9298 void * __builtin_custom_pn (void)
9299 void * __builtin_custom_pni (int)
9300 void * __builtin_custom_pnf (float)
9301 void * __builtin_custom_pnp (void *)
9302 void * __builtin_custom_pnii (int, int)
9303 void * __builtin_custom_pnif (int, float)
9304 void * __builtin_custom_pnip (int, void *)
9305 void * __builtin_custom_pnfi (float, int)
9306 void * __builtin_custom_pnff (float, float)
9307 void * __builtin_custom_pnfp (float, void *)
9308 void * __builtin_custom_pnpi (void *, int)
9309 void * __builtin_custom_pnpf (void *, float)
9310 void * __builtin_custom_pnpp (void *, void *)
9313 @node ARC Built-in Functions
9314 @subsection ARC Built-in Functions
9316 The following built-in functions are provided for ARC targets. The
9317 built-ins generate the corresponding assembly instructions. In the
9318 examples given below, the generated code often requires an operand or
9319 result to be in a register. Where necessary further code will be
9320 generated to ensure this is true, but for brevity this is not
9321 described in each case.
9323 @emph{Note:} Using a built-in to generate an instruction not supported
9324 by a target may cause problems. At present the compiler is not
9325 guaranteed to detect such misuse, and as a result an internal compiler
9326 error may be generated.
9328 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
9329 Return 1 if @var{val} is known to have the byte alignment given
9330 by @var{alignval}, otherwise return 0.
9331 Note that this is different from
9333 __alignof__(*(char *)@var{val}) >= alignval
9335 because __alignof__ sees only the type of the dereference, whereas
9336 __builtin_arc_align uses alignment information from the pointer
9337 as well as from the pointed-to type.
9338 The information available will depend on optimization level.
9341 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
9348 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
9349 The operand is the number of a register to be read. Generates:
9351 mov @var{dest}, r@var{regno}
9353 where the value in @var{dest} will be the result returned from the
9357 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
9358 The first operand is the number of a register to be written, the
9359 second operand is a compile time constant to write into that
9360 register. Generates:
9362 mov r@var{regno}, @var{val}
9366 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
9367 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
9370 divaw @var{dest}, @var{a}, @var{b}
9372 where the value in @var{dest} will be the result returned from the
9376 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
9383 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
9384 The operand, @var{auxv}, is the address of an auxiliary register and
9385 must be a compile time constant. Generates:
9387 lr @var{dest}, [@var{auxr}]
9389 Where the value in @var{dest} will be the result returned from the
9393 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
9394 Only available with @option{-mmul64}. Generates:
9396 mul64 @var{a}, @var{b}
9400 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
9401 Only available with @option{-mmul64}. Generates:
9403 mulu64 @var{a}, @var{b}
9407 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
9414 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
9415 Only valid if the @samp{norm} instruction is available through the
9416 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
9419 norm @var{dest}, @var{src}
9421 Where the value in @var{dest} will be the result returned from the
9425 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
9426 Only valid if the @samp{normw} instruction is available through the
9427 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
9430 normw @var{dest}, @var{src}
9432 Where the value in @var{dest} will be the result returned from the
9436 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
9443 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
9450 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
9451 The first argument, @var{auxv}, is the address of an auxiliary
9452 register, the second argument, @var{val}, is a compile time constant
9453 to be written to the register. Generates:
9455 sr @var{auxr}, [@var{val}]
9459 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
9460 Only valid with @option{-mswap}. Generates:
9462 swap @var{dest}, @var{src}
9464 Where the value in @var{dest} will be the result returned from the
9468 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
9475 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
9476 Only available with @option{-mcpu=ARC700}. Generates:
9482 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
9483 Only available with @option{-mcpu=ARC700}. Generates:
9489 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
9490 Only available with @option{-mcpu=ARC700}. Generates:
9496 The instructions generated by the following builtins are not
9497 considered as candidates for scheduling. They are not moved around by
9498 the compiler during scheduling, and thus can be expected to appear
9499 where they are put in the C code:
9502 __builtin_arc_core_read()
9503 __builtin_arc_core_write()
9504 __builtin_arc_flag()
9506 __builtin_arc_sleep()
9511 @node ARC SIMD Built-in Functions
9512 @subsection ARC SIMD Built-in Functions
9514 SIMD builtins provided by the compiler can be used to generate the
9515 vector instructions. This section describes the available builtins
9516 and their usage in programs. With the @option{-msimd} option, the
9517 compiler provides 128-bit vector types, which can be specified using
9518 the @code{vector_size} attribute. The header file @file{arc-simd.h}
9519 can be included to use the following predefined types:
9521 typedef int __v4si __attribute__((vector_size(16)));
9522 typedef short __v8hi __attribute__((vector_size(16)));
9525 These types can be used to define 128-bit variables. The built-in
9526 functions listed in the following section can be used on these
9527 variables to generate the vector operations.
9529 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
9530 @file{arc-simd.h} also provides equivalent macros called
9531 @code{_@var{someinsn}} that can be used for programming ease and
9532 improved readability. The following macros for DMA control are also
9535 #define _setup_dma_in_channel_reg _vdiwr
9536 #define _setup_dma_out_channel_reg _vdowr
9539 The following is a complete list of all the SIMD built-ins provided
9540 for ARC, grouped by calling signature.
9542 The following take two @code{__v8hi} arguments and return a
9543 @code{__v8hi} result:
9545 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
9546 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
9547 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
9548 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
9549 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
9550 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
9551 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
9552 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
9553 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
9554 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
9555 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
9556 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
9557 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
9558 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
9559 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
9560 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
9561 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
9562 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
9563 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
9564 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
9565 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
9566 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
9567 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
9568 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
9569 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
9570 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
9571 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
9572 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
9573 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
9574 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
9575 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
9576 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
9577 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
9578 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
9579 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
9580 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
9581 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
9582 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
9583 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
9584 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
9585 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
9586 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
9587 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
9588 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
9589 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
9590 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
9591 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
9592 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
9595 The following take one @code{__v8hi} and one @code{int} argument and return a
9596 @code{__v8hi} result:
9599 __v8hi __builtin_arc_vbaddw (__v8hi, int)
9600 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
9601 __v8hi __builtin_arc_vbminw (__v8hi, int)
9602 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
9603 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
9604 __v8hi __builtin_arc_vbmulw (__v8hi, int)
9605 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
9606 __v8hi __builtin_arc_vbsubw (__v8hi, int)
9609 The following take one @code{__v8hi} argument and one @code{int} argument which
9610 must be a 3-bit compile time constant indicating a register number
9611 I0-I7. They return a @code{__v8hi} result.
9613 __v8hi __builtin_arc_vasrw (__v8hi, const int)
9614 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
9615 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
9618 The following take one @code{__v8hi} argument and one @code{int}
9619 argument which must be a 6-bit compile time constant. They return a
9620 @code{__v8hi} result.
9622 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
9623 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
9624 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
9625 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
9626 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
9627 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
9628 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
9631 The following take one @code{__v8hi} argument and one @code{int} argument which
9632 must be a 8-bit compile time constant. They return a @code{__v8hi}
9635 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
9636 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
9637 __v8hi __builtin_arc_vmvw (__v8hi, const int)
9638 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
9641 The following take two @code{int} arguments, the second of which which
9642 must be a 8-bit compile time constant. They return a @code{__v8hi}
9645 __v8hi __builtin_arc_vmovaw (int, const int)
9646 __v8hi __builtin_arc_vmovw (int, const int)
9647 __v8hi __builtin_arc_vmovzw (int, const int)
9650 The following take a single @code{__v8hi} argument and return a
9651 @code{__v8hi} result:
9653 __v8hi __builtin_arc_vabsaw (__v8hi)
9654 __v8hi __builtin_arc_vabsw (__v8hi)
9655 __v8hi __builtin_arc_vaddsuw (__v8hi)
9656 __v8hi __builtin_arc_vexch1 (__v8hi)
9657 __v8hi __builtin_arc_vexch2 (__v8hi)
9658 __v8hi __builtin_arc_vexch4 (__v8hi)
9659 __v8hi __builtin_arc_vsignw (__v8hi)
9660 __v8hi __builtin_arc_vupbaw (__v8hi)
9661 __v8hi __builtin_arc_vupbw (__v8hi)
9662 __v8hi __builtin_arc_vupsbaw (__v8hi)
9663 __v8hi __builtin_arc_vupsbw (__v8hi)
9666 The followign take two @code{int} arguments and return no result:
9668 void __builtin_arc_vdirun (int, int)
9669 void __builtin_arc_vdorun (int, int)
9672 The following take two @code{int} arguments and return no result. The
9673 first argument must a 3-bit compile time constant indicating one of
9674 the DR0-DR7 DMA setup channels:
9676 void __builtin_arc_vdiwr (const int, int)
9677 void __builtin_arc_vdowr (const int, int)
9680 The following take an @code{int} argument and return no result:
9682 void __builtin_arc_vendrec (int)
9683 void __builtin_arc_vrec (int)
9684 void __builtin_arc_vrecrun (int)
9685 void __builtin_arc_vrun (int)
9688 The following take a @code{__v8hi} argument and two @code{int}
9689 arguments and return a @code{__v8hi} result. The second argument must
9690 be a 3-bit compile time constants, indicating one the registers I0-I7,
9691 and the third argument must be an 8-bit compile time constant.
9693 @emph{Note:} Although the equivalent hardware instructions do not take
9694 an SIMD register as an operand, these builtins overwrite the relevant
9695 bits of the @code{__v8hi} register provided as the first argument with
9696 the value loaded from the @code{[Ib, u8]} location in the SDM.
9699 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
9700 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
9701 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
9702 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
9705 The following take two @code{int} arguments and return a @code{__v8hi}
9706 result. The first argument must be a 3-bit compile time constants,
9707 indicating one the registers I0-I7, and the second argument must be an
9708 8-bit compile time constant.
9711 __v8hi __builtin_arc_vld128 (const int, const int)
9712 __v8hi __builtin_arc_vld64w (const int, const int)
9715 The following take a @code{__v8hi} argument and two @code{int}
9716 arguments and return no result. The second argument must be a 3-bit
9717 compile time constants, indicating one the registers I0-I7, and the
9718 third argument must be an 8-bit compile time constant.
9721 void __builtin_arc_vst128 (__v8hi, const int, const int)
9722 void __builtin_arc_vst64 (__v8hi, const int, const int)
9725 The following take a @code{__v8hi} argument and three @code{int}
9726 arguments and return no result. The second argument must be a 3-bit
9727 compile-time constant, identifying the 16-bit sub-register to be
9728 stored, the third argument must be a 3-bit compile time constants,
9729 indicating one the registers I0-I7, and the fourth argument must be an
9730 8-bit compile time constant.
9733 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
9734 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
9737 @node ARM iWMMXt Built-in Functions
9738 @subsection ARM iWMMXt Built-in Functions
9740 These built-in functions are available for the ARM family of
9741 processors when the @option{-mcpu=iwmmxt} switch is used:
9744 typedef int v2si __attribute__ ((vector_size (8)));
9745 typedef short v4hi __attribute__ ((vector_size (8)));
9746 typedef char v8qi __attribute__ ((vector_size (8)));
9748 int __builtin_arm_getwcgr0 (void)
9749 void __builtin_arm_setwcgr0 (int)
9750 int __builtin_arm_getwcgr1 (void)
9751 void __builtin_arm_setwcgr1 (int)
9752 int __builtin_arm_getwcgr2 (void)
9753 void __builtin_arm_setwcgr2 (int)
9754 int __builtin_arm_getwcgr3 (void)
9755 void __builtin_arm_setwcgr3 (int)
9756 int __builtin_arm_textrmsb (v8qi, int)
9757 int __builtin_arm_textrmsh (v4hi, int)
9758 int __builtin_arm_textrmsw (v2si, int)
9759 int __builtin_arm_textrmub (v8qi, int)
9760 int __builtin_arm_textrmuh (v4hi, int)
9761 int __builtin_arm_textrmuw (v2si, int)
9762 v8qi __builtin_arm_tinsrb (v8qi, int, int)
9763 v4hi __builtin_arm_tinsrh (v4hi, int, int)
9764 v2si __builtin_arm_tinsrw (v2si, int, int)
9765 long long __builtin_arm_tmia (long long, int, int)
9766 long long __builtin_arm_tmiabb (long long, int, int)
9767 long long __builtin_arm_tmiabt (long long, int, int)
9768 long long __builtin_arm_tmiaph (long long, int, int)
9769 long long __builtin_arm_tmiatb (long long, int, int)
9770 long long __builtin_arm_tmiatt (long long, int, int)
9771 int __builtin_arm_tmovmskb (v8qi)
9772 int __builtin_arm_tmovmskh (v4hi)
9773 int __builtin_arm_tmovmskw (v2si)
9774 long long __builtin_arm_waccb (v8qi)
9775 long long __builtin_arm_wacch (v4hi)
9776 long long __builtin_arm_waccw (v2si)
9777 v8qi __builtin_arm_waddb (v8qi, v8qi)
9778 v8qi __builtin_arm_waddbss (v8qi, v8qi)
9779 v8qi __builtin_arm_waddbus (v8qi, v8qi)
9780 v4hi __builtin_arm_waddh (v4hi, v4hi)
9781 v4hi __builtin_arm_waddhss (v4hi, v4hi)
9782 v4hi __builtin_arm_waddhus (v4hi, v4hi)
9783 v2si __builtin_arm_waddw (v2si, v2si)
9784 v2si __builtin_arm_waddwss (v2si, v2si)
9785 v2si __builtin_arm_waddwus (v2si, v2si)
9786 v8qi __builtin_arm_walign (v8qi, v8qi, int)
9787 long long __builtin_arm_wand(long long, long long)
9788 long long __builtin_arm_wandn (long long, long long)
9789 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
9790 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
9791 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
9792 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
9793 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
9794 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
9795 v2si __builtin_arm_wcmpeqw (v2si, v2si)
9796 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
9797 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
9798 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
9799 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
9800 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
9801 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
9802 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
9803 long long __builtin_arm_wmacsz (v4hi, v4hi)
9804 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
9805 long long __builtin_arm_wmacuz (v4hi, v4hi)
9806 v4hi __builtin_arm_wmadds (v4hi, v4hi)
9807 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
9808 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
9809 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
9810 v2si __builtin_arm_wmaxsw (v2si, v2si)
9811 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
9812 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
9813 v2si __builtin_arm_wmaxuw (v2si, v2si)
9814 v8qi __builtin_arm_wminsb (v8qi, v8qi)
9815 v4hi __builtin_arm_wminsh (v4hi, v4hi)
9816 v2si __builtin_arm_wminsw (v2si, v2si)
9817 v8qi __builtin_arm_wminub (v8qi, v8qi)
9818 v4hi __builtin_arm_wminuh (v4hi, v4hi)
9819 v2si __builtin_arm_wminuw (v2si, v2si)
9820 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
9821 v4hi __builtin_arm_wmulul (v4hi, v4hi)
9822 v4hi __builtin_arm_wmulum (v4hi, v4hi)
9823 long long __builtin_arm_wor (long long, long long)
9824 v2si __builtin_arm_wpackdss (long long, long long)
9825 v2si __builtin_arm_wpackdus (long long, long long)
9826 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
9827 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
9828 v4hi __builtin_arm_wpackwss (v2si, v2si)
9829 v4hi __builtin_arm_wpackwus (v2si, v2si)
9830 long long __builtin_arm_wrord (long long, long long)
9831 long long __builtin_arm_wrordi (long long, int)
9832 v4hi __builtin_arm_wrorh (v4hi, long long)
9833 v4hi __builtin_arm_wrorhi (v4hi, int)
9834 v2si __builtin_arm_wrorw (v2si, long long)
9835 v2si __builtin_arm_wrorwi (v2si, int)
9836 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
9837 v2si __builtin_arm_wsadbz (v8qi, v8qi)
9838 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
9839 v2si __builtin_arm_wsadhz (v4hi, v4hi)
9840 v4hi __builtin_arm_wshufh (v4hi, int)
9841 long long __builtin_arm_wslld (long long, long long)
9842 long long __builtin_arm_wslldi (long long, int)
9843 v4hi __builtin_arm_wsllh (v4hi, long long)
9844 v4hi __builtin_arm_wsllhi (v4hi, int)
9845 v2si __builtin_arm_wsllw (v2si, long long)
9846 v2si __builtin_arm_wsllwi (v2si, int)
9847 long long __builtin_arm_wsrad (long long, long long)
9848 long long __builtin_arm_wsradi (long long, int)
9849 v4hi __builtin_arm_wsrah (v4hi, long long)
9850 v4hi __builtin_arm_wsrahi (v4hi, int)
9851 v2si __builtin_arm_wsraw (v2si, long long)
9852 v2si __builtin_arm_wsrawi (v2si, int)
9853 long long __builtin_arm_wsrld (long long, long long)
9854 long long __builtin_arm_wsrldi (long long, int)
9855 v4hi __builtin_arm_wsrlh (v4hi, long long)
9856 v4hi __builtin_arm_wsrlhi (v4hi, int)
9857 v2si __builtin_arm_wsrlw (v2si, long long)
9858 v2si __builtin_arm_wsrlwi (v2si, int)
9859 v8qi __builtin_arm_wsubb (v8qi, v8qi)
9860 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
9861 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
9862 v4hi __builtin_arm_wsubh (v4hi, v4hi)
9863 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
9864 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
9865 v2si __builtin_arm_wsubw (v2si, v2si)
9866 v2si __builtin_arm_wsubwss (v2si, v2si)
9867 v2si __builtin_arm_wsubwus (v2si, v2si)
9868 v4hi __builtin_arm_wunpckehsb (v8qi)
9869 v2si __builtin_arm_wunpckehsh (v4hi)
9870 long long __builtin_arm_wunpckehsw (v2si)
9871 v4hi __builtin_arm_wunpckehub (v8qi)
9872 v2si __builtin_arm_wunpckehuh (v4hi)
9873 long long __builtin_arm_wunpckehuw (v2si)
9874 v4hi __builtin_arm_wunpckelsb (v8qi)
9875 v2si __builtin_arm_wunpckelsh (v4hi)
9876 long long __builtin_arm_wunpckelsw (v2si)
9877 v4hi __builtin_arm_wunpckelub (v8qi)
9878 v2si __builtin_arm_wunpckeluh (v4hi)
9879 long long __builtin_arm_wunpckeluw (v2si)
9880 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
9881 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
9882 v2si __builtin_arm_wunpckihw (v2si, v2si)
9883 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
9884 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
9885 v2si __builtin_arm_wunpckilw (v2si, v2si)
9886 long long __builtin_arm_wxor (long long, long long)
9887 long long __builtin_arm_wzero ()
9890 @node ARM NEON Intrinsics
9891 @subsection ARM NEON Intrinsics
9893 These built-in intrinsics for the ARM Advanced SIMD extension are available
9894 when the @option{-mfpu=neon} switch is used:
9896 @include arm-neon-intrinsics.texi
9898 @node ARM ACLE Intrinsics
9899 @subsection ARM ACLE Intrinsics
9901 These built-in intrinsics for the ARMv8-A CRC32 extension are available when
9902 the @option{-march=armv8-a+crc} switch is used:
9904 @include arm-acle-intrinsics.texi
9906 @node AVR Built-in Functions
9907 @subsection AVR Built-in Functions
9909 For each built-in function for AVR, there is an equally named,
9910 uppercase built-in macro defined. That way users can easily query if
9911 or if not a specific built-in is implemented or not. For example, if
9912 @code{__builtin_avr_nop} is available the macro
9913 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
9915 The following built-in functions map to the respective machine
9916 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
9917 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
9918 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
9919 as library call if no hardware multiplier is available.
9922 void __builtin_avr_nop (void)
9923 void __builtin_avr_sei (void)
9924 void __builtin_avr_cli (void)
9925 void __builtin_avr_sleep (void)
9926 void __builtin_avr_wdr (void)
9927 unsigned char __builtin_avr_swap (unsigned char)
9928 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
9929 int __builtin_avr_fmuls (char, char)
9930 int __builtin_avr_fmulsu (char, unsigned char)
9933 In order to delay execution for a specific number of cycles, GCC
9936 void __builtin_avr_delay_cycles (unsigned long ticks)
9940 @code{ticks} is the number of ticks to delay execution. Note that this
9941 built-in does not take into account the effect of interrupts that
9942 might increase delay time. @code{ticks} must be a compile-time
9943 integer constant; delays with a variable number of cycles are not supported.
9946 char __builtin_avr_flash_segment (const __memx void*)
9950 This built-in takes a byte address to the 24-bit
9951 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
9952 the number of the flash segment (the 64 KiB chunk) where the address
9953 points to. Counting starts at @code{0}.
9954 If the address does not point to flash memory, return @code{-1}.
9957 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
9961 Insert bits from @var{bits} into @var{val} and return the resulting
9962 value. The nibbles of @var{map} determine how the insertion is
9963 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
9965 @item If @var{X} is @code{0xf},
9966 then the @var{n}-th bit of @var{val} is returned unaltered.
9968 @item If X is in the range 0@dots{}7,
9969 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
9971 @item If X is in the range 8@dots{}@code{0xe},
9972 then the @var{n}-th result bit is undefined.
9976 One typical use case for this built-in is adjusting input and
9977 output values to non-contiguous port layouts. Some examples:
9980 // same as val, bits is unused
9981 __builtin_avr_insert_bits (0xffffffff, bits, val)
9985 // same as bits, val is unused
9986 __builtin_avr_insert_bits (0x76543210, bits, val)
9990 // same as rotating bits by 4
9991 __builtin_avr_insert_bits (0x32107654, bits, 0)
9995 // high nibble of result is the high nibble of val
9996 // low nibble of result is the low nibble of bits
9997 __builtin_avr_insert_bits (0xffff3210, bits, val)
10001 // reverse the bit order of bits
10002 __builtin_avr_insert_bits (0x01234567, bits, 0)
10005 @node Blackfin Built-in Functions
10006 @subsection Blackfin Built-in Functions
10008 Currently, there are two Blackfin-specific built-in functions. These are
10009 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
10010 using inline assembly; by using these built-in functions the compiler can
10011 automatically add workarounds for hardware errata involving these
10012 instructions. These functions are named as follows:
10015 void __builtin_bfin_csync (void)
10016 void __builtin_bfin_ssync (void)
10019 @node FR-V Built-in Functions
10020 @subsection FR-V Built-in Functions
10022 GCC provides many FR-V-specific built-in functions. In general,
10023 these functions are intended to be compatible with those described
10024 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
10025 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
10026 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
10027 pointer rather than by value.
10029 Most of the functions are named after specific FR-V instructions.
10030 Such functions are said to be ``directly mapped'' and are summarized
10031 here in tabular form.
10035 * Directly-mapped Integer Functions::
10036 * Directly-mapped Media Functions::
10037 * Raw read/write Functions::
10038 * Other Built-in Functions::
10041 @node Argument Types
10042 @subsubsection Argument Types
10044 The arguments to the built-in functions can be divided into three groups:
10045 register numbers, compile-time constants and run-time values. In order
10046 to make this classification clear at a glance, the arguments and return
10047 values are given the following pseudo types:
10049 @multitable @columnfractions .20 .30 .15 .35
10050 @item Pseudo type @tab Real C type @tab Constant? @tab Description
10051 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
10052 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
10053 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
10054 @item @code{uw2} @tab @code{unsigned long long} @tab No
10055 @tab an unsigned doubleword
10056 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
10057 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
10058 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
10059 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
10062 These pseudo types are not defined by GCC, they are simply a notational
10063 convenience used in this manual.
10065 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
10066 and @code{sw2} are evaluated at run time. They correspond to
10067 register operands in the underlying FR-V instructions.
10069 @code{const} arguments represent immediate operands in the underlying
10070 FR-V instructions. They must be compile-time constants.
10072 @code{acc} arguments are evaluated at compile time and specify the number
10073 of an accumulator register. For example, an @code{acc} argument of 2
10074 selects the ACC2 register.
10076 @code{iacc} arguments are similar to @code{acc} arguments but specify the
10077 number of an IACC register. See @pxref{Other Built-in Functions}
10080 @node Directly-mapped Integer Functions
10081 @subsubsection Directly-mapped Integer Functions
10083 The functions listed below map directly to FR-V I-type instructions.
10085 @multitable @columnfractions .45 .32 .23
10086 @item Function prototype @tab Example usage @tab Assembly output
10087 @item @code{sw1 __ADDSS (sw1, sw1)}
10088 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
10089 @tab @code{ADDSS @var{a},@var{b},@var{c}}
10090 @item @code{sw1 __SCAN (sw1, sw1)}
10091 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
10092 @tab @code{SCAN @var{a},@var{b},@var{c}}
10093 @item @code{sw1 __SCUTSS (sw1)}
10094 @tab @code{@var{b} = __SCUTSS (@var{a})}
10095 @tab @code{SCUTSS @var{a},@var{b}}
10096 @item @code{sw1 __SLASS (sw1, sw1)}
10097 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
10098 @tab @code{SLASS @var{a},@var{b},@var{c}}
10099 @item @code{void __SMASS (sw1, sw1)}
10100 @tab @code{__SMASS (@var{a}, @var{b})}
10101 @tab @code{SMASS @var{a},@var{b}}
10102 @item @code{void __SMSSS (sw1, sw1)}
10103 @tab @code{__SMSSS (@var{a}, @var{b})}
10104 @tab @code{SMSSS @var{a},@var{b}}
10105 @item @code{void __SMU (sw1, sw1)}
10106 @tab @code{__SMU (@var{a}, @var{b})}
10107 @tab @code{SMU @var{a},@var{b}}
10108 @item @code{sw2 __SMUL (sw1, sw1)}
10109 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
10110 @tab @code{SMUL @var{a},@var{b},@var{c}}
10111 @item @code{sw1 __SUBSS (sw1, sw1)}
10112 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
10113 @tab @code{SUBSS @var{a},@var{b},@var{c}}
10114 @item @code{uw2 __UMUL (uw1, uw1)}
10115 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
10116 @tab @code{UMUL @var{a},@var{b},@var{c}}
10119 @node Directly-mapped Media Functions
10120 @subsubsection Directly-mapped Media Functions
10122 The functions listed below map directly to FR-V M-type instructions.
10124 @multitable @columnfractions .45 .32 .23
10125 @item Function prototype @tab Example usage @tab Assembly output
10126 @item @code{uw1 __MABSHS (sw1)}
10127 @tab @code{@var{b} = __MABSHS (@var{a})}
10128 @tab @code{MABSHS @var{a},@var{b}}
10129 @item @code{void __MADDACCS (acc, acc)}
10130 @tab @code{__MADDACCS (@var{b}, @var{a})}
10131 @tab @code{MADDACCS @var{a},@var{b}}
10132 @item @code{sw1 __MADDHSS (sw1, sw1)}
10133 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
10134 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
10135 @item @code{uw1 __MADDHUS (uw1, uw1)}
10136 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
10137 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
10138 @item @code{uw1 __MAND (uw1, uw1)}
10139 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
10140 @tab @code{MAND @var{a},@var{b},@var{c}}
10141 @item @code{void __MASACCS (acc, acc)}
10142 @tab @code{__MASACCS (@var{b}, @var{a})}
10143 @tab @code{MASACCS @var{a},@var{b}}
10144 @item @code{uw1 __MAVEH (uw1, uw1)}
10145 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
10146 @tab @code{MAVEH @var{a},@var{b},@var{c}}
10147 @item @code{uw2 __MBTOH (uw1)}
10148 @tab @code{@var{b} = __MBTOH (@var{a})}
10149 @tab @code{MBTOH @var{a},@var{b}}
10150 @item @code{void __MBTOHE (uw1 *, uw1)}
10151 @tab @code{__MBTOHE (&@var{b}, @var{a})}
10152 @tab @code{MBTOHE @var{a},@var{b}}
10153 @item @code{void __MCLRACC (acc)}
10154 @tab @code{__MCLRACC (@var{a})}
10155 @tab @code{MCLRACC @var{a}}
10156 @item @code{void __MCLRACCA (void)}
10157 @tab @code{__MCLRACCA ()}
10158 @tab @code{MCLRACCA}
10159 @item @code{uw1 __Mcop1 (uw1, uw1)}
10160 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
10161 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
10162 @item @code{uw1 __Mcop2 (uw1, uw1)}
10163 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
10164 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
10165 @item @code{uw1 __MCPLHI (uw2, const)}
10166 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
10167 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
10168 @item @code{uw1 __MCPLI (uw2, const)}
10169 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
10170 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
10171 @item @code{void __MCPXIS (acc, sw1, sw1)}
10172 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
10173 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
10174 @item @code{void __MCPXIU (acc, uw1, uw1)}
10175 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
10176 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
10177 @item @code{void __MCPXRS (acc, sw1, sw1)}
10178 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
10179 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
10180 @item @code{void __MCPXRU (acc, uw1, uw1)}
10181 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
10182 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
10183 @item @code{uw1 __MCUT (acc, uw1)}
10184 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
10185 @tab @code{MCUT @var{a},@var{b},@var{c}}
10186 @item @code{uw1 __MCUTSS (acc, sw1)}
10187 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
10188 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
10189 @item @code{void __MDADDACCS (acc, acc)}
10190 @tab @code{__MDADDACCS (@var{b}, @var{a})}
10191 @tab @code{MDADDACCS @var{a},@var{b}}
10192 @item @code{void __MDASACCS (acc, acc)}
10193 @tab @code{__MDASACCS (@var{b}, @var{a})}
10194 @tab @code{MDASACCS @var{a},@var{b}}
10195 @item @code{uw2 __MDCUTSSI (acc, const)}
10196 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
10197 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
10198 @item @code{uw2 __MDPACKH (uw2, uw2)}
10199 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
10200 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
10201 @item @code{uw2 __MDROTLI (uw2, const)}
10202 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
10203 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
10204 @item @code{void __MDSUBACCS (acc, acc)}
10205 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
10206 @tab @code{MDSUBACCS @var{a},@var{b}}
10207 @item @code{void __MDUNPACKH (uw1 *, uw2)}
10208 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
10209 @tab @code{MDUNPACKH @var{a},@var{b}}
10210 @item @code{uw2 __MEXPDHD (uw1, const)}
10211 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
10212 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
10213 @item @code{uw1 __MEXPDHW (uw1, const)}
10214 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
10215 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
10216 @item @code{uw1 __MHDSETH (uw1, const)}
10217 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
10218 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
10219 @item @code{sw1 __MHDSETS (const)}
10220 @tab @code{@var{b} = __MHDSETS (@var{a})}
10221 @tab @code{MHDSETS #@var{a},@var{b}}
10222 @item @code{uw1 __MHSETHIH (uw1, const)}
10223 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
10224 @tab @code{MHSETHIH #@var{a},@var{b}}
10225 @item @code{sw1 __MHSETHIS (sw1, const)}
10226 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
10227 @tab @code{MHSETHIS #@var{a},@var{b}}
10228 @item @code{uw1 __MHSETLOH (uw1, const)}
10229 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
10230 @tab @code{MHSETLOH #@var{a},@var{b}}
10231 @item @code{sw1 __MHSETLOS (sw1, const)}
10232 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
10233 @tab @code{MHSETLOS #@var{a},@var{b}}
10234 @item @code{uw1 __MHTOB (uw2)}
10235 @tab @code{@var{b} = __MHTOB (@var{a})}
10236 @tab @code{MHTOB @var{a},@var{b}}
10237 @item @code{void __MMACHS (acc, sw1, sw1)}
10238 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
10239 @tab @code{MMACHS @var{a},@var{b},@var{c}}
10240 @item @code{void __MMACHU (acc, uw1, uw1)}
10241 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
10242 @tab @code{MMACHU @var{a},@var{b},@var{c}}
10243 @item @code{void __MMRDHS (acc, sw1, sw1)}
10244 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
10245 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
10246 @item @code{void __MMRDHU (acc, uw1, uw1)}
10247 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
10248 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
10249 @item @code{void __MMULHS (acc, sw1, sw1)}
10250 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
10251 @tab @code{MMULHS @var{a},@var{b},@var{c}}
10252 @item @code{void __MMULHU (acc, uw1, uw1)}
10253 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
10254 @tab @code{MMULHU @var{a},@var{b},@var{c}}
10255 @item @code{void __MMULXHS (acc, sw1, sw1)}
10256 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
10257 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
10258 @item @code{void __MMULXHU (acc, uw1, uw1)}
10259 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
10260 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
10261 @item @code{uw1 __MNOT (uw1)}
10262 @tab @code{@var{b} = __MNOT (@var{a})}
10263 @tab @code{MNOT @var{a},@var{b}}
10264 @item @code{uw1 __MOR (uw1, uw1)}
10265 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
10266 @tab @code{MOR @var{a},@var{b},@var{c}}
10267 @item @code{uw1 __MPACKH (uh, uh)}
10268 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
10269 @tab @code{MPACKH @var{a},@var{b},@var{c}}
10270 @item @code{sw2 __MQADDHSS (sw2, sw2)}
10271 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
10272 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
10273 @item @code{uw2 __MQADDHUS (uw2, uw2)}
10274 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
10275 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
10276 @item @code{void __MQCPXIS (acc, sw2, sw2)}
10277 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
10278 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
10279 @item @code{void __MQCPXIU (acc, uw2, uw2)}
10280 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
10281 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
10282 @item @code{void __MQCPXRS (acc, sw2, sw2)}
10283 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
10284 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
10285 @item @code{void __MQCPXRU (acc, uw2, uw2)}
10286 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
10287 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
10288 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
10289 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
10290 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
10291 @item @code{sw2 __MQLMTHS (sw2, sw2)}
10292 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
10293 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
10294 @item @code{void __MQMACHS (acc, sw2, sw2)}
10295 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
10296 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
10297 @item @code{void __MQMACHU (acc, uw2, uw2)}
10298 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
10299 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
10300 @item @code{void __MQMACXHS (acc, sw2, sw2)}
10301 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
10302 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
10303 @item @code{void __MQMULHS (acc, sw2, sw2)}
10304 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
10305 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
10306 @item @code{void __MQMULHU (acc, uw2, uw2)}
10307 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
10308 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
10309 @item @code{void __MQMULXHS (acc, sw2, sw2)}
10310 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
10311 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
10312 @item @code{void __MQMULXHU (acc, uw2, uw2)}
10313 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
10314 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
10315 @item @code{sw2 __MQSATHS (sw2, sw2)}
10316 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
10317 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
10318 @item @code{uw2 __MQSLLHI (uw2, int)}
10319 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
10320 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
10321 @item @code{sw2 __MQSRAHI (sw2, int)}
10322 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
10323 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
10324 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
10325 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
10326 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
10327 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
10328 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
10329 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
10330 @item @code{void __MQXMACHS (acc, sw2, sw2)}
10331 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
10332 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
10333 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
10334 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
10335 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
10336 @item @code{uw1 __MRDACC (acc)}
10337 @tab @code{@var{b} = __MRDACC (@var{a})}
10338 @tab @code{MRDACC @var{a},@var{b}}
10339 @item @code{uw1 __MRDACCG (acc)}
10340 @tab @code{@var{b} = __MRDACCG (@var{a})}
10341 @tab @code{MRDACCG @var{a},@var{b}}
10342 @item @code{uw1 __MROTLI (uw1, const)}
10343 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
10344 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
10345 @item @code{uw1 __MROTRI (uw1, const)}
10346 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
10347 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
10348 @item @code{sw1 __MSATHS (sw1, sw1)}
10349 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
10350 @tab @code{MSATHS @var{a},@var{b},@var{c}}
10351 @item @code{uw1 __MSATHU (uw1, uw1)}
10352 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
10353 @tab @code{MSATHU @var{a},@var{b},@var{c}}
10354 @item @code{uw1 __MSLLHI (uw1, const)}
10355 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
10356 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
10357 @item @code{sw1 __MSRAHI (sw1, const)}
10358 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
10359 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
10360 @item @code{uw1 __MSRLHI (uw1, const)}
10361 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
10362 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
10363 @item @code{void __MSUBACCS (acc, acc)}
10364 @tab @code{__MSUBACCS (@var{b}, @var{a})}
10365 @tab @code{MSUBACCS @var{a},@var{b}}
10366 @item @code{sw1 __MSUBHSS (sw1, sw1)}
10367 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
10368 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
10369 @item @code{uw1 __MSUBHUS (uw1, uw1)}
10370 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
10371 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
10372 @item @code{void __MTRAP (void)}
10373 @tab @code{__MTRAP ()}
10375 @item @code{uw2 __MUNPACKH (uw1)}
10376 @tab @code{@var{b} = __MUNPACKH (@var{a})}
10377 @tab @code{MUNPACKH @var{a},@var{b}}
10378 @item @code{uw1 __MWCUT (uw2, uw1)}
10379 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
10380 @tab @code{MWCUT @var{a},@var{b},@var{c}}
10381 @item @code{void __MWTACC (acc, uw1)}
10382 @tab @code{__MWTACC (@var{b}, @var{a})}
10383 @tab @code{MWTACC @var{a},@var{b}}
10384 @item @code{void __MWTACCG (acc, uw1)}
10385 @tab @code{__MWTACCG (@var{b}, @var{a})}
10386 @tab @code{MWTACCG @var{a},@var{b}}
10387 @item @code{uw1 __MXOR (uw1, uw1)}
10388 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
10389 @tab @code{MXOR @var{a},@var{b},@var{c}}
10392 @node Raw read/write Functions
10393 @subsubsection Raw read/write Functions
10395 This sections describes built-in functions related to read and write
10396 instructions to access memory. These functions generate
10397 @code{membar} instructions to flush the I/O load and stores where
10398 appropriate, as described in Fujitsu's manual described above.
10402 @item unsigned char __builtin_read8 (void *@var{data})
10403 @item unsigned short __builtin_read16 (void *@var{data})
10404 @item unsigned long __builtin_read32 (void *@var{data})
10405 @item unsigned long long __builtin_read64 (void *@var{data})
10407 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
10408 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
10409 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
10410 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
10413 @node Other Built-in Functions
10414 @subsubsection Other Built-in Functions
10416 This section describes built-in functions that are not named after
10417 a specific FR-V instruction.
10420 @item sw2 __IACCreadll (iacc @var{reg})
10421 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
10422 for future expansion and must be 0.
10424 @item sw1 __IACCreadl (iacc @var{reg})
10425 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
10426 Other values of @var{reg} are rejected as invalid.
10428 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
10429 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
10430 is reserved for future expansion and must be 0.
10432 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
10433 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
10434 is 1. Other values of @var{reg} are rejected as invalid.
10436 @item void __data_prefetch0 (const void *@var{x})
10437 Use the @code{dcpl} instruction to load the contents of address @var{x}
10438 into the data cache.
10440 @item void __data_prefetch (const void *@var{x})
10441 Use the @code{nldub} instruction to load the contents of address @var{x}
10442 into the data cache. The instruction is issued in slot I1@.
10445 @node X86 Built-in Functions
10446 @subsection X86 Built-in Functions
10448 These built-in functions are available for the i386 and x86-64 family
10449 of computers, depending on the command-line switches used.
10451 If you specify command-line switches such as @option{-msse},
10452 the compiler could use the extended instruction sets even if the built-ins
10453 are not used explicitly in the program. For this reason, applications
10454 that perform run-time CPU detection must compile separate files for each
10455 supported architecture, using the appropriate flags. In particular,
10456 the file containing the CPU detection code should be compiled without
10459 The following machine modes are available for use with MMX built-in functions
10460 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
10461 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
10462 vector of eight 8-bit integers. Some of the built-in functions operate on
10463 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
10465 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
10466 of two 32-bit floating-point values.
10468 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
10469 floating-point values. Some instructions use a vector of four 32-bit
10470 integers, these use @code{V4SI}. Finally, some instructions operate on an
10471 entire vector register, interpreting it as a 128-bit integer, these use mode
10474 In 64-bit mode, the x86-64 family of processors uses additional built-in
10475 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
10476 floating point and @code{TC} 128-bit complex floating-point values.
10478 The following floating-point built-in functions are available in 64-bit
10479 mode. All of them implement the function that is part of the name.
10482 __float128 __builtin_fabsq (__float128)
10483 __float128 __builtin_copysignq (__float128, __float128)
10486 The following built-in function is always available.
10489 @item void __builtin_ia32_pause (void)
10490 Generates the @code{pause} machine instruction with a compiler memory
10494 The following floating-point built-in functions are made available in the
10498 @item __float128 __builtin_infq (void)
10499 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
10500 @findex __builtin_infq
10502 @item __float128 __builtin_huge_valq (void)
10503 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
10504 @findex __builtin_huge_valq
10507 The following built-in functions are always available and can be used to
10508 check the target platform type.
10510 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
10511 This function runs the CPU detection code to check the type of CPU and the
10512 features supported. This built-in function needs to be invoked along with the built-in functions
10513 to check CPU type and features, @code{__builtin_cpu_is} and
10514 @code{__builtin_cpu_supports}, only when used in a function that is
10515 executed before any constructors are called. The CPU detection code is
10516 automatically executed in a very high priority constructor.
10518 For example, this function has to be used in @code{ifunc} resolvers that
10519 check for CPU type using the built-in functions @code{__builtin_cpu_is}
10520 and @code{__builtin_cpu_supports}, or in constructors on targets that
10521 don't support constructor priority.
10524 static void (*resolve_memcpy (void)) (void)
10526 // ifunc resolvers fire before constructors, explicitly call the init
10528 __builtin_cpu_init ();
10529 if (__builtin_cpu_supports ("ssse3"))
10530 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
10532 return default_memcpy;
10535 void *memcpy (void *, const void *, size_t)
10536 __attribute__ ((ifunc ("resolve_memcpy")));
10541 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
10542 This function returns a positive integer if the run-time CPU
10543 is of type @var{cpuname}
10544 and returns @code{0} otherwise. The following CPU names can be detected:
10560 Intel Core i7 Nehalem CPU.
10563 Intel Core i7 Westmere CPU.
10566 Intel Core i7 Sandy Bridge CPU.
10572 AMD Family 10h CPU.
10575 AMD Family 10h Barcelona CPU.
10578 AMD Family 10h Shanghai CPU.
10581 AMD Family 10h Istanbul CPU.
10584 AMD Family 14h CPU.
10587 AMD Family 15h CPU.
10590 AMD Family 15h Bulldozer version 1.
10593 AMD Family 15h Bulldozer version 2.
10596 AMD Family 15h Bulldozer version 3.
10599 AMD Family 15h Bulldozer version 4.
10602 AMD Family 16h CPU.
10605 Here is an example:
10607 if (__builtin_cpu_is ("corei7"))
10609 do_corei7 (); // Core i7 specific implementation.
10613 do_generic (); // Generic implementation.
10618 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
10619 This function returns a positive integer if the run-time CPU
10620 supports @var{feature}
10621 and returns @code{0} otherwise. The following features can be detected:
10629 POPCNT instruction.
10637 SSSE3 instructions.
10639 SSE4.1 instructions.
10641 SSE4.2 instructions.
10648 Here is an example:
10650 if (__builtin_cpu_supports ("popcnt"))
10652 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
10656 count = generic_countbits (n); //generic implementation.
10662 The following built-in functions are made available by @option{-mmmx}.
10663 All of them generate the machine instruction that is part of the name.
10666 v8qi __builtin_ia32_paddb (v8qi, v8qi)
10667 v4hi __builtin_ia32_paddw (v4hi, v4hi)
10668 v2si __builtin_ia32_paddd (v2si, v2si)
10669 v8qi __builtin_ia32_psubb (v8qi, v8qi)
10670 v4hi __builtin_ia32_psubw (v4hi, v4hi)
10671 v2si __builtin_ia32_psubd (v2si, v2si)
10672 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
10673 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
10674 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
10675 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
10676 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
10677 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
10678 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
10679 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
10680 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
10681 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
10682 di __builtin_ia32_pand (di, di)
10683 di __builtin_ia32_pandn (di,di)
10684 di __builtin_ia32_por (di, di)
10685 di __builtin_ia32_pxor (di, di)
10686 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
10687 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
10688 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
10689 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
10690 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
10691 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
10692 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
10693 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
10694 v2si __builtin_ia32_punpckhdq (v2si, v2si)
10695 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
10696 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
10697 v2si __builtin_ia32_punpckldq (v2si, v2si)
10698 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
10699 v4hi __builtin_ia32_packssdw (v2si, v2si)
10700 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
10702 v4hi __builtin_ia32_psllw (v4hi, v4hi)
10703 v2si __builtin_ia32_pslld (v2si, v2si)
10704 v1di __builtin_ia32_psllq (v1di, v1di)
10705 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
10706 v2si __builtin_ia32_psrld (v2si, v2si)
10707 v1di __builtin_ia32_psrlq (v1di, v1di)
10708 v4hi __builtin_ia32_psraw (v4hi, v4hi)
10709 v2si __builtin_ia32_psrad (v2si, v2si)
10710 v4hi __builtin_ia32_psllwi (v4hi, int)
10711 v2si __builtin_ia32_pslldi (v2si, int)
10712 v1di __builtin_ia32_psllqi (v1di, int)
10713 v4hi __builtin_ia32_psrlwi (v4hi, int)
10714 v2si __builtin_ia32_psrldi (v2si, int)
10715 v1di __builtin_ia32_psrlqi (v1di, int)
10716 v4hi __builtin_ia32_psrawi (v4hi, int)
10717 v2si __builtin_ia32_psradi (v2si, int)
10721 The following built-in functions are made available either with
10722 @option{-msse}, or with a combination of @option{-m3dnow} and
10723 @option{-march=athlon}. All of them generate the machine
10724 instruction that is part of the name.
10727 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
10728 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
10729 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
10730 v1di __builtin_ia32_psadbw (v8qi, v8qi)
10731 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
10732 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
10733 v8qi __builtin_ia32_pminub (v8qi, v8qi)
10734 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
10735 int __builtin_ia32_pmovmskb (v8qi)
10736 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
10737 void __builtin_ia32_movntq (di *, di)
10738 void __builtin_ia32_sfence (void)
10741 The following built-in functions are available when @option{-msse} is used.
10742 All of them generate the machine instruction that is part of the name.
10745 int __builtin_ia32_comieq (v4sf, v4sf)
10746 int __builtin_ia32_comineq (v4sf, v4sf)
10747 int __builtin_ia32_comilt (v4sf, v4sf)
10748 int __builtin_ia32_comile (v4sf, v4sf)
10749 int __builtin_ia32_comigt (v4sf, v4sf)
10750 int __builtin_ia32_comige (v4sf, v4sf)
10751 int __builtin_ia32_ucomieq (v4sf, v4sf)
10752 int __builtin_ia32_ucomineq (v4sf, v4sf)
10753 int __builtin_ia32_ucomilt (v4sf, v4sf)
10754 int __builtin_ia32_ucomile (v4sf, v4sf)
10755 int __builtin_ia32_ucomigt (v4sf, v4sf)
10756 int __builtin_ia32_ucomige (v4sf, v4sf)
10757 v4sf __builtin_ia32_addps (v4sf, v4sf)
10758 v4sf __builtin_ia32_subps (v4sf, v4sf)
10759 v4sf __builtin_ia32_mulps (v4sf, v4sf)
10760 v4sf __builtin_ia32_divps (v4sf, v4sf)
10761 v4sf __builtin_ia32_addss (v4sf, v4sf)
10762 v4sf __builtin_ia32_subss (v4sf, v4sf)
10763 v4sf __builtin_ia32_mulss (v4sf, v4sf)
10764 v4sf __builtin_ia32_divss (v4sf, v4sf)
10765 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
10766 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
10767 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
10768 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
10769 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
10770 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
10771 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
10772 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
10773 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
10774 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
10775 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
10776 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
10777 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
10778 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
10779 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
10780 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
10781 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
10782 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
10783 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
10784 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
10785 v4sf __builtin_ia32_maxps (v4sf, v4sf)
10786 v4sf __builtin_ia32_maxss (v4sf, v4sf)
10787 v4sf __builtin_ia32_minps (v4sf, v4sf)
10788 v4sf __builtin_ia32_minss (v4sf, v4sf)
10789 v4sf __builtin_ia32_andps (v4sf, v4sf)
10790 v4sf __builtin_ia32_andnps (v4sf, v4sf)
10791 v4sf __builtin_ia32_orps (v4sf, v4sf)
10792 v4sf __builtin_ia32_xorps (v4sf, v4sf)
10793 v4sf __builtin_ia32_movss (v4sf, v4sf)
10794 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
10795 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
10796 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
10797 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
10798 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
10799 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
10800 v2si __builtin_ia32_cvtps2pi (v4sf)
10801 int __builtin_ia32_cvtss2si (v4sf)
10802 v2si __builtin_ia32_cvttps2pi (v4sf)
10803 int __builtin_ia32_cvttss2si (v4sf)
10804 v4sf __builtin_ia32_rcpps (v4sf)
10805 v4sf __builtin_ia32_rsqrtps (v4sf)
10806 v4sf __builtin_ia32_sqrtps (v4sf)
10807 v4sf __builtin_ia32_rcpss (v4sf)
10808 v4sf __builtin_ia32_rsqrtss (v4sf)
10809 v4sf __builtin_ia32_sqrtss (v4sf)
10810 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
10811 void __builtin_ia32_movntps (float *, v4sf)
10812 int __builtin_ia32_movmskps (v4sf)
10815 The following built-in functions are available when @option{-msse} is used.
10818 @item v4sf __builtin_ia32_loadups (float *)
10819 Generates the @code{movups} machine instruction as a load from memory.
10820 @item void __builtin_ia32_storeups (float *, v4sf)
10821 Generates the @code{movups} machine instruction as a store to memory.
10822 @item v4sf __builtin_ia32_loadss (float *)
10823 Generates the @code{movss} machine instruction as a load from memory.
10824 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
10825 Generates the @code{movhps} machine instruction as a load from memory.
10826 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
10827 Generates the @code{movlps} machine instruction as a load from memory
10828 @item void __builtin_ia32_storehps (v2sf *, v4sf)
10829 Generates the @code{movhps} machine instruction as a store to memory.
10830 @item void __builtin_ia32_storelps (v2sf *, v4sf)
10831 Generates the @code{movlps} machine instruction as a store to memory.
10834 The following built-in functions are available when @option{-msse2} is used.
10835 All of them generate the machine instruction that is part of the name.
10838 int __builtin_ia32_comisdeq (v2df, v2df)
10839 int __builtin_ia32_comisdlt (v2df, v2df)
10840 int __builtin_ia32_comisdle (v2df, v2df)
10841 int __builtin_ia32_comisdgt (v2df, v2df)
10842 int __builtin_ia32_comisdge (v2df, v2df)
10843 int __builtin_ia32_comisdneq (v2df, v2df)
10844 int __builtin_ia32_ucomisdeq (v2df, v2df)
10845 int __builtin_ia32_ucomisdlt (v2df, v2df)
10846 int __builtin_ia32_ucomisdle (v2df, v2df)
10847 int __builtin_ia32_ucomisdgt (v2df, v2df)
10848 int __builtin_ia32_ucomisdge (v2df, v2df)
10849 int __builtin_ia32_ucomisdneq (v2df, v2df)
10850 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
10851 v2df __builtin_ia32_cmpltpd (v2df, v2df)
10852 v2df __builtin_ia32_cmplepd (v2df, v2df)
10853 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
10854 v2df __builtin_ia32_cmpgepd (v2df, v2df)
10855 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
10856 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
10857 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
10858 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
10859 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
10860 v2df __builtin_ia32_cmpngepd (v2df, v2df)
10861 v2df __builtin_ia32_cmpordpd (v2df, v2df)
10862 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
10863 v2df __builtin_ia32_cmpltsd (v2df, v2df)
10864 v2df __builtin_ia32_cmplesd (v2df, v2df)
10865 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
10866 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
10867 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
10868 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
10869 v2df __builtin_ia32_cmpordsd (v2df, v2df)
10870 v2di __builtin_ia32_paddq (v2di, v2di)
10871 v2di __builtin_ia32_psubq (v2di, v2di)
10872 v2df __builtin_ia32_addpd (v2df, v2df)
10873 v2df __builtin_ia32_subpd (v2df, v2df)
10874 v2df __builtin_ia32_mulpd (v2df, v2df)
10875 v2df __builtin_ia32_divpd (v2df, v2df)
10876 v2df __builtin_ia32_addsd (v2df, v2df)
10877 v2df __builtin_ia32_subsd (v2df, v2df)
10878 v2df __builtin_ia32_mulsd (v2df, v2df)
10879 v2df __builtin_ia32_divsd (v2df, v2df)
10880 v2df __builtin_ia32_minpd (v2df, v2df)
10881 v2df __builtin_ia32_maxpd (v2df, v2df)
10882 v2df __builtin_ia32_minsd (v2df, v2df)
10883 v2df __builtin_ia32_maxsd (v2df, v2df)
10884 v2df __builtin_ia32_andpd (v2df, v2df)
10885 v2df __builtin_ia32_andnpd (v2df, v2df)
10886 v2df __builtin_ia32_orpd (v2df, v2df)
10887 v2df __builtin_ia32_xorpd (v2df, v2df)
10888 v2df __builtin_ia32_movsd (v2df, v2df)
10889 v2df __builtin_ia32_unpckhpd (v2df, v2df)
10890 v2df __builtin_ia32_unpcklpd (v2df, v2df)
10891 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
10892 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
10893 v4si __builtin_ia32_paddd128 (v4si, v4si)
10894 v2di __builtin_ia32_paddq128 (v2di, v2di)
10895 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
10896 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
10897 v4si __builtin_ia32_psubd128 (v4si, v4si)
10898 v2di __builtin_ia32_psubq128 (v2di, v2di)
10899 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
10900 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
10901 v2di __builtin_ia32_pand128 (v2di, v2di)
10902 v2di __builtin_ia32_pandn128 (v2di, v2di)
10903 v2di __builtin_ia32_por128 (v2di, v2di)
10904 v2di __builtin_ia32_pxor128 (v2di, v2di)
10905 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
10906 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
10907 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
10908 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
10909 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
10910 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
10911 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
10912 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
10913 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
10914 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
10915 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
10916 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
10917 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
10918 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
10919 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
10920 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
10921 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
10922 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
10923 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
10924 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
10925 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
10926 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
10927 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
10928 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
10929 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
10930 v2df __builtin_ia32_loadupd (double *)
10931 void __builtin_ia32_storeupd (double *, v2df)
10932 v2df __builtin_ia32_loadhpd (v2df, double const *)
10933 v2df __builtin_ia32_loadlpd (v2df, double const *)
10934 int __builtin_ia32_movmskpd (v2df)
10935 int __builtin_ia32_pmovmskb128 (v16qi)
10936 void __builtin_ia32_movnti (int *, int)
10937 void __builtin_ia32_movnti64 (long long int *, long long int)
10938 void __builtin_ia32_movntpd (double *, v2df)
10939 void __builtin_ia32_movntdq (v2df *, v2df)
10940 v4si __builtin_ia32_pshufd (v4si, int)
10941 v8hi __builtin_ia32_pshuflw (v8hi, int)
10942 v8hi __builtin_ia32_pshufhw (v8hi, int)
10943 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
10944 v2df __builtin_ia32_sqrtpd (v2df)
10945 v2df __builtin_ia32_sqrtsd (v2df)
10946 v2df __builtin_ia32_shufpd (v2df, v2df, int)
10947 v2df __builtin_ia32_cvtdq2pd (v4si)
10948 v4sf __builtin_ia32_cvtdq2ps (v4si)
10949 v4si __builtin_ia32_cvtpd2dq (v2df)
10950 v2si __builtin_ia32_cvtpd2pi (v2df)
10951 v4sf __builtin_ia32_cvtpd2ps (v2df)
10952 v4si __builtin_ia32_cvttpd2dq (v2df)
10953 v2si __builtin_ia32_cvttpd2pi (v2df)
10954 v2df __builtin_ia32_cvtpi2pd (v2si)
10955 int __builtin_ia32_cvtsd2si (v2df)
10956 int __builtin_ia32_cvttsd2si (v2df)
10957 long long __builtin_ia32_cvtsd2si64 (v2df)
10958 long long __builtin_ia32_cvttsd2si64 (v2df)
10959 v4si __builtin_ia32_cvtps2dq (v4sf)
10960 v2df __builtin_ia32_cvtps2pd (v4sf)
10961 v4si __builtin_ia32_cvttps2dq (v4sf)
10962 v2df __builtin_ia32_cvtsi2sd (v2df, int)
10963 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
10964 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
10965 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
10966 void __builtin_ia32_clflush (const void *)
10967 void __builtin_ia32_lfence (void)
10968 void __builtin_ia32_mfence (void)
10969 v16qi __builtin_ia32_loaddqu (const char *)
10970 void __builtin_ia32_storedqu (char *, v16qi)
10971 v1di __builtin_ia32_pmuludq (v2si, v2si)
10972 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
10973 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
10974 v4si __builtin_ia32_pslld128 (v4si, v4si)
10975 v2di __builtin_ia32_psllq128 (v2di, v2di)
10976 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
10977 v4si __builtin_ia32_psrld128 (v4si, v4si)
10978 v2di __builtin_ia32_psrlq128 (v2di, v2di)
10979 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
10980 v4si __builtin_ia32_psrad128 (v4si, v4si)
10981 v2di __builtin_ia32_pslldqi128 (v2di, int)
10982 v8hi __builtin_ia32_psllwi128 (v8hi, int)
10983 v4si __builtin_ia32_pslldi128 (v4si, int)
10984 v2di __builtin_ia32_psllqi128 (v2di, int)
10985 v2di __builtin_ia32_psrldqi128 (v2di, int)
10986 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
10987 v4si __builtin_ia32_psrldi128 (v4si, int)
10988 v2di __builtin_ia32_psrlqi128 (v2di, int)
10989 v8hi __builtin_ia32_psrawi128 (v8hi, int)
10990 v4si __builtin_ia32_psradi128 (v4si, int)
10991 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
10992 v2di __builtin_ia32_movq128 (v2di)
10995 The following built-in functions are available when @option{-msse3} is used.
10996 All of them generate the machine instruction that is part of the name.
10999 v2df __builtin_ia32_addsubpd (v2df, v2df)
11000 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
11001 v2df __builtin_ia32_haddpd (v2df, v2df)
11002 v4sf __builtin_ia32_haddps (v4sf, v4sf)
11003 v2df __builtin_ia32_hsubpd (v2df, v2df)
11004 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
11005 v16qi __builtin_ia32_lddqu (char const *)
11006 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
11007 v4sf __builtin_ia32_movshdup (v4sf)
11008 v4sf __builtin_ia32_movsldup (v4sf)
11009 void __builtin_ia32_mwait (unsigned int, unsigned int)
11012 The following built-in functions are available when @option{-mssse3} is used.
11013 All of them generate the machine instruction that is part of the name.
11016 v2si __builtin_ia32_phaddd (v2si, v2si)
11017 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
11018 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
11019 v2si __builtin_ia32_phsubd (v2si, v2si)
11020 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
11021 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
11022 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
11023 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
11024 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
11025 v8qi __builtin_ia32_psignb (v8qi, v8qi)
11026 v2si __builtin_ia32_psignd (v2si, v2si)
11027 v4hi __builtin_ia32_psignw (v4hi, v4hi)
11028 v1di __builtin_ia32_palignr (v1di, v1di, int)
11029 v8qi __builtin_ia32_pabsb (v8qi)
11030 v2si __builtin_ia32_pabsd (v2si)
11031 v4hi __builtin_ia32_pabsw (v4hi)
11034 The following built-in functions are available when @option{-mssse3} is used.
11035 All of them generate the machine instruction that is part of the name.
11038 v4si __builtin_ia32_phaddd128 (v4si, v4si)
11039 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
11040 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
11041 v4si __builtin_ia32_phsubd128 (v4si, v4si)
11042 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
11043 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
11044 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
11045 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
11046 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
11047 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
11048 v4si __builtin_ia32_psignd128 (v4si, v4si)
11049 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
11050 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
11051 v16qi __builtin_ia32_pabsb128 (v16qi)
11052 v4si __builtin_ia32_pabsd128 (v4si)
11053 v8hi __builtin_ia32_pabsw128 (v8hi)
11056 The following built-in functions are available when @option{-msse4.1} is
11057 used. All of them generate the machine instruction that is part of the
11061 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
11062 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
11063 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
11064 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
11065 v2df __builtin_ia32_dppd (v2df, v2df, const int)
11066 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
11067 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
11068 v2di __builtin_ia32_movntdqa (v2di *);
11069 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
11070 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
11071 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
11072 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
11073 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
11074 v8hi __builtin_ia32_phminposuw128 (v8hi)
11075 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
11076 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
11077 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
11078 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
11079 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
11080 v4si __builtin_ia32_pminsd128 (v4si, v4si)
11081 v4si __builtin_ia32_pminud128 (v4si, v4si)
11082 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
11083 v4si __builtin_ia32_pmovsxbd128 (v16qi)
11084 v2di __builtin_ia32_pmovsxbq128 (v16qi)
11085 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
11086 v2di __builtin_ia32_pmovsxdq128 (v4si)
11087 v4si __builtin_ia32_pmovsxwd128 (v8hi)
11088 v2di __builtin_ia32_pmovsxwq128 (v8hi)
11089 v4si __builtin_ia32_pmovzxbd128 (v16qi)
11090 v2di __builtin_ia32_pmovzxbq128 (v16qi)
11091 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
11092 v2di __builtin_ia32_pmovzxdq128 (v4si)
11093 v4si __builtin_ia32_pmovzxwd128 (v8hi)
11094 v2di __builtin_ia32_pmovzxwq128 (v8hi)
11095 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
11096 v4si __builtin_ia32_pmulld128 (v4si, v4si)
11097 int __builtin_ia32_ptestc128 (v2di, v2di)
11098 int __builtin_ia32_ptestnzc128 (v2di, v2di)
11099 int __builtin_ia32_ptestz128 (v2di, v2di)
11100 v2df __builtin_ia32_roundpd (v2df, const int)
11101 v4sf __builtin_ia32_roundps (v4sf, const int)
11102 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
11103 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
11106 The following built-in functions are available when @option{-msse4.1} is
11110 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
11111 Generates the @code{insertps} machine instruction.
11112 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
11113 Generates the @code{pextrb} machine instruction.
11114 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
11115 Generates the @code{pinsrb} machine instruction.
11116 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
11117 Generates the @code{pinsrd} machine instruction.
11118 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
11119 Generates the @code{pinsrq} machine instruction in 64bit mode.
11122 The following built-in functions are changed to generate new SSE4.1
11123 instructions when @option{-msse4.1} is used.
11126 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
11127 Generates the @code{extractps} machine instruction.
11128 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
11129 Generates the @code{pextrd} machine instruction.
11130 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
11131 Generates the @code{pextrq} machine instruction in 64bit mode.
11134 The following built-in functions are available when @option{-msse4.2} is
11135 used. All of them generate the machine instruction that is part of the
11139 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
11140 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
11141 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
11142 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
11143 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
11144 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
11145 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
11146 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
11147 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
11148 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
11149 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
11150 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
11151 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
11152 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
11153 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
11156 The following built-in functions are available when @option{-msse4.2} is
11160 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
11161 Generates the @code{crc32b} machine instruction.
11162 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
11163 Generates the @code{crc32w} machine instruction.
11164 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
11165 Generates the @code{crc32l} machine instruction.
11166 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
11167 Generates the @code{crc32q} machine instruction.
11170 The following built-in functions are changed to generate new SSE4.2
11171 instructions when @option{-msse4.2} is used.
11174 @item int __builtin_popcount (unsigned int)
11175 Generates the @code{popcntl} machine instruction.
11176 @item int __builtin_popcountl (unsigned long)
11177 Generates the @code{popcntl} or @code{popcntq} machine instruction,
11178 depending on the size of @code{unsigned long}.
11179 @item int __builtin_popcountll (unsigned long long)
11180 Generates the @code{popcntq} machine instruction.
11183 The following built-in functions are available when @option{-mavx} is
11184 used. All of them generate the machine instruction that is part of the
11188 v4df __builtin_ia32_addpd256 (v4df,v4df)
11189 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
11190 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
11191 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
11192 v4df __builtin_ia32_andnpd256 (v4df,v4df)
11193 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
11194 v4df __builtin_ia32_andpd256 (v4df,v4df)
11195 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
11196 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
11197 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
11198 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
11199 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
11200 v2df __builtin_ia32_cmppd (v2df,v2df,int)
11201 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
11202 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
11203 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
11204 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
11205 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
11206 v4df __builtin_ia32_cvtdq2pd256 (v4si)
11207 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
11208 v4si __builtin_ia32_cvtpd2dq256 (v4df)
11209 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
11210 v8si __builtin_ia32_cvtps2dq256 (v8sf)
11211 v4df __builtin_ia32_cvtps2pd256 (v4sf)
11212 v4si __builtin_ia32_cvttpd2dq256 (v4df)
11213 v8si __builtin_ia32_cvttps2dq256 (v8sf)
11214 v4df __builtin_ia32_divpd256 (v4df,v4df)
11215 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
11216 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
11217 v4df __builtin_ia32_haddpd256 (v4df,v4df)
11218 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
11219 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
11220 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
11221 v32qi __builtin_ia32_lddqu256 (pcchar)
11222 v32qi __builtin_ia32_loaddqu256 (pcchar)
11223 v4df __builtin_ia32_loadupd256 (pcdouble)
11224 v8sf __builtin_ia32_loadups256 (pcfloat)
11225 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
11226 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
11227 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
11228 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
11229 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
11230 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
11231 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
11232 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
11233 v4df __builtin_ia32_maxpd256 (v4df,v4df)
11234 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
11235 v4df __builtin_ia32_minpd256 (v4df,v4df)
11236 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
11237 v4df __builtin_ia32_movddup256 (v4df)
11238 int __builtin_ia32_movmskpd256 (v4df)
11239 int __builtin_ia32_movmskps256 (v8sf)
11240 v8sf __builtin_ia32_movshdup256 (v8sf)
11241 v8sf __builtin_ia32_movsldup256 (v8sf)
11242 v4df __builtin_ia32_mulpd256 (v4df,v4df)
11243 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
11244 v4df __builtin_ia32_orpd256 (v4df,v4df)
11245 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
11246 v2df __builtin_ia32_pd_pd256 (v4df)
11247 v4df __builtin_ia32_pd256_pd (v2df)
11248 v4sf __builtin_ia32_ps_ps256 (v8sf)
11249 v8sf __builtin_ia32_ps256_ps (v4sf)
11250 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
11251 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
11252 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
11253 v8sf __builtin_ia32_rcpps256 (v8sf)
11254 v4df __builtin_ia32_roundpd256 (v4df,int)
11255 v8sf __builtin_ia32_roundps256 (v8sf,int)
11256 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
11257 v8sf __builtin_ia32_rsqrtps256 (v8sf)
11258 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
11259 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
11260 v4si __builtin_ia32_si_si256 (v8si)
11261 v8si __builtin_ia32_si256_si (v4si)
11262 v4df __builtin_ia32_sqrtpd256 (v4df)
11263 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
11264 v8sf __builtin_ia32_sqrtps256 (v8sf)
11265 void __builtin_ia32_storedqu256 (pchar,v32qi)
11266 void __builtin_ia32_storeupd256 (pdouble,v4df)
11267 void __builtin_ia32_storeups256 (pfloat,v8sf)
11268 v4df __builtin_ia32_subpd256 (v4df,v4df)
11269 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
11270 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
11271 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
11272 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
11273 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
11274 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
11275 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
11276 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
11277 v4sf __builtin_ia32_vbroadcastss (pcfloat)
11278 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
11279 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
11280 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
11281 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
11282 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
11283 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
11284 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
11285 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
11286 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
11287 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
11288 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
11289 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
11290 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
11291 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
11292 v2df __builtin_ia32_vpermilpd (v2df,int)
11293 v4df __builtin_ia32_vpermilpd256 (v4df,int)
11294 v4sf __builtin_ia32_vpermilps (v4sf,int)
11295 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
11296 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
11297 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
11298 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
11299 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
11300 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
11301 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
11302 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
11303 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
11304 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
11305 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
11306 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
11307 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
11308 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
11309 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
11310 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
11311 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
11312 void __builtin_ia32_vzeroall (void)
11313 void __builtin_ia32_vzeroupper (void)
11314 v4df __builtin_ia32_xorpd256 (v4df,v4df)
11315 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
11318 The following built-in functions are available when @option{-mavx2} is
11319 used. All of them generate the machine instruction that is part of the
11323 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,v32qi,int)
11324 v32qi __builtin_ia32_pabsb256 (v32qi)
11325 v16hi __builtin_ia32_pabsw256 (v16hi)
11326 v8si __builtin_ia32_pabsd256 (v8si)
11327 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
11328 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
11329 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
11330 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
11331 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
11332 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
11333 v8si __builtin_ia32_paddd256 (v8si,v8si)
11334 v4di __builtin_ia32_paddq256 (v4di,v4di)
11335 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
11336 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
11337 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
11338 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
11339 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
11340 v4di __builtin_ia32_andsi256 (v4di,v4di)
11341 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
11342 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
11343 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
11344 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
11345 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
11346 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
11347 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
11348 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
11349 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
11350 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
11351 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
11352 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
11353 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
11354 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
11355 v8si __builtin_ia32_phaddd256 (v8si,v8si)
11356 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
11357 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
11358 v8si __builtin_ia32_phsubd256 (v8si,v8si)
11359 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
11360 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
11361 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
11362 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
11363 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
11364 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
11365 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
11366 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
11367 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
11368 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
11369 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
11370 v8si __builtin_ia32_pminsd256 (v8si,v8si)
11371 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
11372 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
11373 v8si __builtin_ia32_pminud256 (v8si,v8si)
11374 int __builtin_ia32_pmovmskb256 (v32qi)
11375 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
11376 v8si __builtin_ia32_pmovsxbd256 (v16qi)
11377 v4di __builtin_ia32_pmovsxbq256 (v16qi)
11378 v8si __builtin_ia32_pmovsxwd256 (v8hi)
11379 v4di __builtin_ia32_pmovsxwq256 (v8hi)
11380 v4di __builtin_ia32_pmovsxdq256 (v4si)
11381 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
11382 v8si __builtin_ia32_pmovzxbd256 (v16qi)
11383 v4di __builtin_ia32_pmovzxbq256 (v16qi)
11384 v8si __builtin_ia32_pmovzxwd256 (v8hi)
11385 v4di __builtin_ia32_pmovzxwq256 (v8hi)
11386 v4di __builtin_ia32_pmovzxdq256 (v4si)
11387 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
11388 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
11389 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
11390 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
11391 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
11392 v8si __builtin_ia32_pmulld256 (v8si,v8si)
11393 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
11394 v4di __builtin_ia32_por256 (v4di,v4di)
11395 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
11396 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
11397 v8si __builtin_ia32_pshufd256 (v8si,int)
11398 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
11399 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
11400 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
11401 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
11402 v8si __builtin_ia32_psignd256 (v8si,v8si)
11403 v4di __builtin_ia32_pslldqi256 (v4di,int)
11404 v16hi __builtin_ia32_psllwi256 (16hi,int)
11405 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
11406 v8si __builtin_ia32_pslldi256 (v8si,int)
11407 v8si __builtin_ia32_pslld256(v8si,v4si)
11408 v4di __builtin_ia32_psllqi256 (v4di,int)
11409 v4di __builtin_ia32_psllq256(v4di,v2di)
11410 v16hi __builtin_ia32_psrawi256 (v16hi,int)
11411 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
11412 v8si __builtin_ia32_psradi256 (v8si,int)
11413 v8si __builtin_ia32_psrad256 (v8si,v4si)
11414 v4di __builtin_ia32_psrldqi256 (v4di, int)
11415 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
11416 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
11417 v8si __builtin_ia32_psrldi256 (v8si,int)
11418 v8si __builtin_ia32_psrld256 (v8si,v4si)
11419 v4di __builtin_ia32_psrlqi256 (v4di,int)
11420 v4di __builtin_ia32_psrlq256(v4di,v2di)
11421 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
11422 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
11423 v8si __builtin_ia32_psubd256 (v8si,v8si)
11424 v4di __builtin_ia32_psubq256 (v4di,v4di)
11425 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
11426 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
11427 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
11428 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
11429 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
11430 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
11431 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
11432 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
11433 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
11434 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
11435 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
11436 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
11437 v4di __builtin_ia32_pxor256 (v4di,v4di)
11438 v4di __builtin_ia32_movntdqa256 (pv4di)
11439 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
11440 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
11441 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
11442 v4di __builtin_ia32_vbroadcastsi256 (v2di)
11443 v4si __builtin_ia32_pblendd128 (v4si,v4si)
11444 v8si __builtin_ia32_pblendd256 (v8si,v8si)
11445 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
11446 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
11447 v8si __builtin_ia32_pbroadcastd256 (v4si)
11448 v4di __builtin_ia32_pbroadcastq256 (v2di)
11449 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
11450 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
11451 v4si __builtin_ia32_pbroadcastd128 (v4si)
11452 v2di __builtin_ia32_pbroadcastq128 (v2di)
11453 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
11454 v4df __builtin_ia32_permdf256 (v4df,int)
11455 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
11456 v4di __builtin_ia32_permdi256 (v4di,int)
11457 v4di __builtin_ia32_permti256 (v4di,v4di,int)
11458 v4di __builtin_ia32_extract128i256 (v4di,int)
11459 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
11460 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
11461 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
11462 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
11463 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
11464 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
11465 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
11466 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
11467 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
11468 v8si __builtin_ia32_psllv8si (v8si,v8si)
11469 v4si __builtin_ia32_psllv4si (v4si,v4si)
11470 v4di __builtin_ia32_psllv4di (v4di,v4di)
11471 v2di __builtin_ia32_psllv2di (v2di,v2di)
11472 v8si __builtin_ia32_psrav8si (v8si,v8si)
11473 v4si __builtin_ia32_psrav4si (v4si,v4si)
11474 v8si __builtin_ia32_psrlv8si (v8si,v8si)
11475 v4si __builtin_ia32_psrlv4si (v4si,v4si)
11476 v4di __builtin_ia32_psrlv4di (v4di,v4di)
11477 v2di __builtin_ia32_psrlv2di (v2di,v2di)
11478 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
11479 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
11480 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
11481 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
11482 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
11483 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
11484 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
11485 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
11486 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
11487 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
11488 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
11489 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
11490 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
11491 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
11492 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
11493 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
11496 The following built-in functions are available when @option{-maes} is
11497 used. All of them generate the machine instruction that is part of the
11501 v2di __builtin_ia32_aesenc128 (v2di, v2di)
11502 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
11503 v2di __builtin_ia32_aesdec128 (v2di, v2di)
11504 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
11505 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
11506 v2di __builtin_ia32_aesimc128 (v2di)
11509 The following built-in function is available when @option{-mpclmul} is
11513 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
11514 Generates the @code{pclmulqdq} machine instruction.
11517 The following built-in function is available when @option{-mfsgsbase} is
11518 used. All of them generate the machine instruction that is part of the
11522 unsigned int __builtin_ia32_rdfsbase32 (void)
11523 unsigned long long __builtin_ia32_rdfsbase64 (void)
11524 unsigned int __builtin_ia32_rdgsbase32 (void)
11525 unsigned long long __builtin_ia32_rdgsbase64 (void)
11526 void _writefsbase_u32 (unsigned int)
11527 void _writefsbase_u64 (unsigned long long)
11528 void _writegsbase_u32 (unsigned int)
11529 void _writegsbase_u64 (unsigned long long)
11532 The following built-in function is available when @option{-mrdrnd} is
11533 used. All of them generate the machine instruction that is part of the
11537 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
11538 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
11539 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
11542 The following built-in functions are available when @option{-msse4a} is used.
11543 All of them generate the machine instruction that is part of the name.
11546 void __builtin_ia32_movntsd (double *, v2df)
11547 void __builtin_ia32_movntss (float *, v4sf)
11548 v2di __builtin_ia32_extrq (v2di, v16qi)
11549 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
11550 v2di __builtin_ia32_insertq (v2di, v2di)
11551 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
11554 The following built-in functions are available when @option{-mxop} is used.
11556 v2df __builtin_ia32_vfrczpd (v2df)
11557 v4sf __builtin_ia32_vfrczps (v4sf)
11558 v2df __builtin_ia32_vfrczsd (v2df, v2df)
11559 v4sf __builtin_ia32_vfrczss (v4sf, v4sf)
11560 v4df __builtin_ia32_vfrczpd256 (v4df)
11561 v8sf __builtin_ia32_vfrczps256 (v8sf)
11562 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
11563 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
11564 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
11565 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
11566 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
11567 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
11568 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
11569 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
11570 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
11571 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
11572 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
11573 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
11574 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
11575 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
11576 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
11577 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
11578 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
11579 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
11580 v4si __builtin_ia32_vpcomequd (v4si, v4si)
11581 v2di __builtin_ia32_vpcomequq (v2di, v2di)
11582 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
11583 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
11584 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
11585 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
11586 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
11587 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
11588 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
11589 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
11590 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
11591 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
11592 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
11593 v4si __builtin_ia32_vpcomged (v4si, v4si)
11594 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
11595 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
11596 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
11597 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
11598 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
11599 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
11600 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
11601 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
11602 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
11603 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
11604 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
11605 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
11606 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
11607 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
11608 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
11609 v4si __builtin_ia32_vpcomled (v4si, v4si)
11610 v2di __builtin_ia32_vpcomleq (v2di, v2di)
11611 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
11612 v4si __builtin_ia32_vpcomleud (v4si, v4si)
11613 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
11614 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
11615 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
11616 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
11617 v4si __builtin_ia32_vpcomltd (v4si, v4si)
11618 v2di __builtin_ia32_vpcomltq (v2di, v2di)
11619 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
11620 v4si __builtin_ia32_vpcomltud (v4si, v4si)
11621 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
11622 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
11623 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
11624 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
11625 v4si __builtin_ia32_vpcomned (v4si, v4si)
11626 v2di __builtin_ia32_vpcomneq (v2di, v2di)
11627 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
11628 v4si __builtin_ia32_vpcomneud (v4si, v4si)
11629 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
11630 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
11631 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
11632 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
11633 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
11634 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
11635 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
11636 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
11637 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
11638 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
11639 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
11640 v4si __builtin_ia32_vphaddbd (v16qi)
11641 v2di __builtin_ia32_vphaddbq (v16qi)
11642 v8hi __builtin_ia32_vphaddbw (v16qi)
11643 v2di __builtin_ia32_vphadddq (v4si)
11644 v4si __builtin_ia32_vphaddubd (v16qi)
11645 v2di __builtin_ia32_vphaddubq (v16qi)
11646 v8hi __builtin_ia32_vphaddubw (v16qi)
11647 v2di __builtin_ia32_vphaddudq (v4si)
11648 v4si __builtin_ia32_vphadduwd (v8hi)
11649 v2di __builtin_ia32_vphadduwq (v8hi)
11650 v4si __builtin_ia32_vphaddwd (v8hi)
11651 v2di __builtin_ia32_vphaddwq (v8hi)
11652 v8hi __builtin_ia32_vphsubbw (v16qi)
11653 v2di __builtin_ia32_vphsubdq (v4si)
11654 v4si __builtin_ia32_vphsubwd (v8hi)
11655 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
11656 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
11657 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
11658 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
11659 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
11660 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
11661 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
11662 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
11663 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
11664 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
11665 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
11666 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
11667 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
11668 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
11669 v4si __builtin_ia32_vprotd (v4si, v4si)
11670 v2di __builtin_ia32_vprotq (v2di, v2di)
11671 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
11672 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
11673 v4si __builtin_ia32_vpshad (v4si, v4si)
11674 v2di __builtin_ia32_vpshaq (v2di, v2di)
11675 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
11676 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
11677 v4si __builtin_ia32_vpshld (v4si, v4si)
11678 v2di __builtin_ia32_vpshlq (v2di, v2di)
11679 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
11682 The following built-in functions are available when @option{-mfma4} is used.
11683 All of them generate the machine instruction that is part of the name.
11686 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
11687 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
11688 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
11689 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
11690 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
11691 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
11692 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
11693 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
11694 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
11695 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
11696 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
11697 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
11698 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
11699 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
11700 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
11701 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
11702 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
11703 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
11704 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
11705 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
11706 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
11707 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
11708 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
11709 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
11710 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
11711 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
11712 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
11713 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
11714 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
11715 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
11716 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
11717 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
11721 The following built-in functions are available when @option{-mlwp} is used.
11724 void __builtin_ia32_llwpcb16 (void *);
11725 void __builtin_ia32_llwpcb32 (void *);
11726 void __builtin_ia32_llwpcb64 (void *);
11727 void * __builtin_ia32_llwpcb16 (void);
11728 void * __builtin_ia32_llwpcb32 (void);
11729 void * __builtin_ia32_llwpcb64 (void);
11730 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
11731 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
11732 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
11733 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
11734 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
11735 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
11738 The following built-in functions are available when @option{-mbmi} is used.
11739 All of them generate the machine instruction that is part of the name.
11741 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
11742 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
11745 The following built-in functions are available when @option{-mbmi2} is used.
11746 All of them generate the machine instruction that is part of the name.
11748 unsigned int _bzhi_u32 (unsigned int, unsigned int)
11749 unsigned int _pdep_u32 (unsigned int, unsigned int)
11750 unsigned int _pext_u32 (unsigned int, unsigned int)
11751 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
11752 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
11753 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
11756 The following built-in functions are available when @option{-mlzcnt} is used.
11757 All of them generate the machine instruction that is part of the name.
11759 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
11760 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
11761 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
11764 The following built-in functions are available when @option{-mfxsr} is used.
11765 All of them generate the machine instruction that is part of the name.
11767 void __builtin_ia32_fxsave (void *)
11768 void __builtin_ia32_fxrstor (void *)
11769 void __builtin_ia32_fxsave64 (void *)
11770 void __builtin_ia32_fxrstor64 (void *)
11773 The following built-in functions are available when @option{-mxsave} is used.
11774 All of them generate the machine instruction that is part of the name.
11776 void __builtin_ia32_xsave (void *, long long)
11777 void __builtin_ia32_xrstor (void *, long long)
11778 void __builtin_ia32_xsave64 (void *, long long)
11779 void __builtin_ia32_xrstor64 (void *, long long)
11782 The following built-in functions are available when @option{-mxsaveopt} is used.
11783 All of them generate the machine instruction that is part of the name.
11785 void __builtin_ia32_xsaveopt (void *, long long)
11786 void __builtin_ia32_xsaveopt64 (void *, long long)
11789 The following built-in functions are available when @option{-mtbm} is used.
11790 Both of them generate the immediate form of the bextr machine instruction.
11792 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
11793 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
11797 The following built-in functions are available when @option{-m3dnow} is used.
11798 All of them generate the machine instruction that is part of the name.
11801 void __builtin_ia32_femms (void)
11802 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
11803 v2si __builtin_ia32_pf2id (v2sf)
11804 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
11805 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
11806 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
11807 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
11808 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
11809 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
11810 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
11811 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
11812 v2sf __builtin_ia32_pfrcp (v2sf)
11813 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
11814 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
11815 v2sf __builtin_ia32_pfrsqrt (v2sf)
11816 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
11817 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
11818 v2sf __builtin_ia32_pi2fd (v2si)
11819 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
11822 The following built-in functions are available when both @option{-m3dnow}
11823 and @option{-march=athlon} are used. All of them generate the machine
11824 instruction that is part of the name.
11827 v2si __builtin_ia32_pf2iw (v2sf)
11828 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
11829 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
11830 v2sf __builtin_ia32_pi2fw (v2si)
11831 v2sf __builtin_ia32_pswapdsf (v2sf)
11832 v2si __builtin_ia32_pswapdsi (v2si)
11835 The following built-in functions are available when @option{-mrtm} is used
11836 They are used for restricted transactional memory. These are the internal
11837 low level functions. Normally the functions in
11838 @ref{X86 transactional memory intrinsics} should be used instead.
11841 int __builtin_ia32_xbegin ()
11842 void __builtin_ia32_xend ()
11843 void __builtin_ia32_xabort (status)
11844 int __builtin_ia32_xtest ()
11847 @node X86 transactional memory intrinsics
11848 @subsection X86 transaction memory intrinsics
11850 Hardware transactional memory intrinsics for i386. These allow to use
11851 memory transactions with RTM (Restricted Transactional Memory).
11852 For using HLE (Hardware Lock Elision) see @ref{x86 specific memory model extensions for transactional memory} instead.
11853 This support is enabled with the @option{-mrtm} option.
11855 A memory transaction commits all changes to memory in an atomic way,
11856 as visible to other threads. If the transaction fails it is rolled back
11857 and all side effects discarded.
11859 Generally there is no guarantee that a memory transaction ever succeeds
11860 and suitable fallback code always needs to be supplied.
11862 @deftypefn {RTM Function} {unsigned} _xbegin ()
11863 Start a RTM (Restricted Transactional Memory) transaction.
11864 Returns _XBEGIN_STARTED when the transaction
11865 started successfully (note this is not 0, so the constant has to be
11866 explicitely tested). When the transaction aborts all side effects
11867 are undone and an abort code is returned. There is no guarantee
11868 any transaction ever succeeds, so there always needs to be a valid
11869 tested fallback path.
11873 #include <immintrin.h>
11875 if ((status = _xbegin ()) == _XBEGIN_STARTED) @{
11876 ... transaction code...
11879 ... non transactional fallback path...
11883 Valid abort status bits (when the value is not @code{_XBEGIN_STARTED}) are:
11886 @item _XABORT_EXPLICIT
11887 Transaction explicitely aborted with @code{_xabort}. The parameter passed
11888 to @code{_xabort} is available with @code{_XABORT_CODE(status)}
11889 @item _XABORT_RETRY
11890 Transaction retry is possible.
11891 @item _XABORT_CONFLICT
11892 Transaction abort due to a memory conflict with another thread
11893 @item _XABORT_CAPACITY
11894 Transaction abort due to the transaction using too much memory
11895 @item _XABORT_DEBUG
11896 Transaction abort due to a debug trap
11897 @item _XABORT_NESTED
11898 Transaction abort in a inner nested transaction
11901 @deftypefn {RTM Function} {void} _xend ()
11902 Commit the current transaction. When no transaction is active this will
11903 fault. All memory side effects of the transactions will become visible
11904 to other threads in an atomic matter.
11907 @deftypefn {RTM Function} {int} _xtest ()
11908 Return a value not zero when a transaction is currently active, otherwise 0.
11911 @deftypefn {RTM Function} {void} _xabort (status)
11912 Abort the current transaction. When no transaction is active this is a no-op.
11913 status must be a 8bit constant, that is included in the status code returned
11917 @node MIPS DSP Built-in Functions
11918 @subsection MIPS DSP Built-in Functions
11920 The MIPS DSP Application-Specific Extension (ASE) includes new
11921 instructions that are designed to improve the performance of DSP and
11922 media applications. It provides instructions that operate on packed
11923 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
11925 GCC supports MIPS DSP operations using both the generic
11926 vector extensions (@pxref{Vector Extensions}) and a collection of
11927 MIPS-specific built-in functions. Both kinds of support are
11928 enabled by the @option{-mdsp} command-line option.
11930 Revision 2 of the ASE was introduced in the second half of 2006.
11931 This revision adds extra instructions to the original ASE, but is
11932 otherwise backwards-compatible with it. You can select revision 2
11933 using the command-line option @option{-mdspr2}; this option implies
11936 The SCOUNT and POS bits of the DSP control register are global. The
11937 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
11938 POS bits. During optimization, the compiler does not delete these
11939 instructions and it does not delete calls to functions containing
11940 these instructions.
11942 At present, GCC only provides support for operations on 32-bit
11943 vectors. The vector type associated with 8-bit integer data is
11944 usually called @code{v4i8}, the vector type associated with Q7
11945 is usually called @code{v4q7}, the vector type associated with 16-bit
11946 integer data is usually called @code{v2i16}, and the vector type
11947 associated with Q15 is usually called @code{v2q15}. They can be
11948 defined in C as follows:
11951 typedef signed char v4i8 __attribute__ ((vector_size(4)));
11952 typedef signed char v4q7 __attribute__ ((vector_size(4)));
11953 typedef short v2i16 __attribute__ ((vector_size(4)));
11954 typedef short v2q15 __attribute__ ((vector_size(4)));
11957 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
11958 initialized in the same way as aggregates. For example:
11961 v4i8 a = @{1, 2, 3, 4@};
11963 b = (v4i8) @{5, 6, 7, 8@};
11965 v2q15 c = @{0x0fcb, 0x3a75@};
11967 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
11970 @emph{Note:} The CPU's endianness determines the order in which values
11971 are packed. On little-endian targets, the first value is the least
11972 significant and the last value is the most significant. The opposite
11973 order applies to big-endian targets. For example, the code above
11974 sets the lowest byte of @code{a} to @code{1} on little-endian targets
11975 and @code{4} on big-endian targets.
11977 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
11978 representation. As shown in this example, the integer representation
11979 of a Q7 value can be obtained by multiplying the fractional value by
11980 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
11981 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
11984 The table below lists the @code{v4i8} and @code{v2q15} operations for which
11985 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
11986 and @code{c} and @code{d} are @code{v2q15} values.
11988 @multitable @columnfractions .50 .50
11989 @item C code @tab MIPS instruction
11990 @item @code{a + b} @tab @code{addu.qb}
11991 @item @code{c + d} @tab @code{addq.ph}
11992 @item @code{a - b} @tab @code{subu.qb}
11993 @item @code{c - d} @tab @code{subq.ph}
11996 The table below lists the @code{v2i16} operation for which
11997 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
11998 @code{v2i16} values.
12000 @multitable @columnfractions .50 .50
12001 @item C code @tab MIPS instruction
12002 @item @code{e * f} @tab @code{mul.ph}
12005 It is easier to describe the DSP built-in functions if we first define
12006 the following types:
12011 typedef unsigned int ui32;
12012 typedef long long a64;
12015 @code{q31} and @code{i32} are actually the same as @code{int}, but we
12016 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
12017 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
12018 @code{long long}, but we use @code{a64} to indicate values that are
12019 placed in one of the four DSP accumulators (@code{$ac0},
12020 @code{$ac1}, @code{$ac2} or @code{$ac3}).
12022 Also, some built-in functions prefer or require immediate numbers as
12023 parameters, because the corresponding DSP instructions accept both immediate
12024 numbers and register operands, or accept immediate numbers only. The
12025 immediate parameters are listed as follows.
12033 imm0_255: 0 to 255.
12034 imm_n32_31: -32 to 31.
12035 imm_n512_511: -512 to 511.
12038 The following built-in functions map directly to a particular MIPS DSP
12039 instruction. Please refer to the architecture specification
12040 for details on what each instruction does.
12043 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
12044 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
12045 q31 __builtin_mips_addq_s_w (q31, q31)
12046 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
12047 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
12048 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
12049 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
12050 q31 __builtin_mips_subq_s_w (q31, q31)
12051 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
12052 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
12053 i32 __builtin_mips_addsc (i32, i32)
12054 i32 __builtin_mips_addwc (i32, i32)
12055 i32 __builtin_mips_modsub (i32, i32)
12056 i32 __builtin_mips_raddu_w_qb (v4i8)
12057 v2q15 __builtin_mips_absq_s_ph (v2q15)
12058 q31 __builtin_mips_absq_s_w (q31)
12059 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
12060 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
12061 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
12062 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
12063 q31 __builtin_mips_preceq_w_phl (v2q15)
12064 q31 __builtin_mips_preceq_w_phr (v2q15)
12065 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
12066 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
12067 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
12068 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
12069 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
12070 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
12071 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
12072 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
12073 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
12074 v4i8 __builtin_mips_shll_qb (v4i8, i32)
12075 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
12076 v2q15 __builtin_mips_shll_ph (v2q15, i32)
12077 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
12078 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
12079 q31 __builtin_mips_shll_s_w (q31, imm0_31)
12080 q31 __builtin_mips_shll_s_w (q31, i32)
12081 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
12082 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
12083 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
12084 v2q15 __builtin_mips_shra_ph (v2q15, i32)
12085 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
12086 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
12087 q31 __builtin_mips_shra_r_w (q31, imm0_31)
12088 q31 __builtin_mips_shra_r_w (q31, i32)
12089 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
12090 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
12091 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
12092 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
12093 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
12094 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
12095 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
12096 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
12097 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
12098 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
12099 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
12100 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
12101 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
12102 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
12103 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
12104 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
12105 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
12106 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
12107 i32 __builtin_mips_bitrev (i32)
12108 i32 __builtin_mips_insv (i32, i32)
12109 v4i8 __builtin_mips_repl_qb (imm0_255)
12110 v4i8 __builtin_mips_repl_qb (i32)
12111 v2q15 __builtin_mips_repl_ph (imm_n512_511)
12112 v2q15 __builtin_mips_repl_ph (i32)
12113 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
12114 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
12115 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
12116 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
12117 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
12118 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
12119 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
12120 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
12121 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
12122 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
12123 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
12124 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
12125 i32 __builtin_mips_extr_w (a64, imm0_31)
12126 i32 __builtin_mips_extr_w (a64, i32)
12127 i32 __builtin_mips_extr_r_w (a64, imm0_31)
12128 i32 __builtin_mips_extr_s_h (a64, i32)
12129 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
12130 i32 __builtin_mips_extr_rs_w (a64, i32)
12131 i32 __builtin_mips_extr_s_h (a64, imm0_31)
12132 i32 __builtin_mips_extr_r_w (a64, i32)
12133 i32 __builtin_mips_extp (a64, imm0_31)
12134 i32 __builtin_mips_extp (a64, i32)
12135 i32 __builtin_mips_extpdp (a64, imm0_31)
12136 i32 __builtin_mips_extpdp (a64, i32)
12137 a64 __builtin_mips_shilo (a64, imm_n32_31)
12138 a64 __builtin_mips_shilo (a64, i32)
12139 a64 __builtin_mips_mthlip (a64, i32)
12140 void __builtin_mips_wrdsp (i32, imm0_63)
12141 i32 __builtin_mips_rddsp (imm0_63)
12142 i32 __builtin_mips_lbux (void *, i32)
12143 i32 __builtin_mips_lhx (void *, i32)
12144 i32 __builtin_mips_lwx (void *, i32)
12145 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
12146 i32 __builtin_mips_bposge32 (void)
12147 a64 __builtin_mips_madd (a64, i32, i32);
12148 a64 __builtin_mips_maddu (a64, ui32, ui32);
12149 a64 __builtin_mips_msub (a64, i32, i32);
12150 a64 __builtin_mips_msubu (a64, ui32, ui32);
12151 a64 __builtin_mips_mult (i32, i32);
12152 a64 __builtin_mips_multu (ui32, ui32);
12155 The following built-in functions map directly to a particular MIPS DSP REV 2
12156 instruction. Please refer to the architecture specification
12157 for details on what each instruction does.
12160 v4q7 __builtin_mips_absq_s_qb (v4q7);
12161 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
12162 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
12163 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
12164 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
12165 i32 __builtin_mips_append (i32, i32, imm0_31);
12166 i32 __builtin_mips_balign (i32, i32, imm0_3);
12167 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
12168 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
12169 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
12170 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
12171 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
12172 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
12173 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
12174 q31 __builtin_mips_mulq_rs_w (q31, q31);
12175 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
12176 q31 __builtin_mips_mulq_s_w (q31, q31);
12177 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
12178 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
12179 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
12180 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
12181 i32 __builtin_mips_prepend (i32, i32, imm0_31);
12182 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
12183 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
12184 v4i8 __builtin_mips_shra_qb (v4i8, i32);
12185 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
12186 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
12187 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
12188 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
12189 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
12190 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
12191 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
12192 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
12193 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
12194 q31 __builtin_mips_addqh_w (q31, q31);
12195 q31 __builtin_mips_addqh_r_w (q31, q31);
12196 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
12197 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
12198 q31 __builtin_mips_subqh_w (q31, q31);
12199 q31 __builtin_mips_subqh_r_w (q31, q31);
12200 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
12201 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
12202 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
12203 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
12204 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
12205 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
12209 @node MIPS Paired-Single Support
12210 @subsection MIPS Paired-Single Support
12212 The MIPS64 architecture includes a number of instructions that
12213 operate on pairs of single-precision floating-point values.
12214 Each pair is packed into a 64-bit floating-point register,
12215 with one element being designated the ``upper half'' and
12216 the other being designated the ``lower half''.
12218 GCC supports paired-single operations using both the generic
12219 vector extensions (@pxref{Vector Extensions}) and a collection of
12220 MIPS-specific built-in functions. Both kinds of support are
12221 enabled by the @option{-mpaired-single} command-line option.
12223 The vector type associated with paired-single values is usually
12224 called @code{v2sf}. It can be defined in C as follows:
12227 typedef float v2sf __attribute__ ((vector_size (8)));
12230 @code{v2sf} values are initialized in the same way as aggregates.
12234 v2sf a = @{1.5, 9.1@};
12237 b = (v2sf) @{e, f@};
12240 @emph{Note:} The CPU's endianness determines which value is stored in
12241 the upper half of a register and which value is stored in the lower half.
12242 On little-endian targets, the first value is the lower one and the second
12243 value is the upper one. The opposite order applies to big-endian targets.
12244 For example, the code above sets the lower half of @code{a} to
12245 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
12247 @node MIPS Loongson Built-in Functions
12248 @subsection MIPS Loongson Built-in Functions
12250 GCC provides intrinsics to access the SIMD instructions provided by the
12251 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
12252 available after inclusion of the @code{loongson.h} header file,
12253 operate on the following 64-bit vector types:
12256 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
12257 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
12258 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
12259 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
12260 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
12261 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
12264 The intrinsics provided are listed below; each is named after the
12265 machine instruction to which it corresponds, with suffixes added as
12266 appropriate to distinguish intrinsics that expand to the same machine
12267 instruction yet have different argument types. Refer to the architecture
12268 documentation for a description of the functionality of each
12272 int16x4_t packsswh (int32x2_t s, int32x2_t t);
12273 int8x8_t packsshb (int16x4_t s, int16x4_t t);
12274 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
12275 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
12276 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
12277 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
12278 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
12279 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
12280 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
12281 uint64_t paddd_u (uint64_t s, uint64_t t);
12282 int64_t paddd_s (int64_t s, int64_t t);
12283 int16x4_t paddsh (int16x4_t s, int16x4_t t);
12284 int8x8_t paddsb (int8x8_t s, int8x8_t t);
12285 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
12286 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
12287 uint64_t pandn_ud (uint64_t s, uint64_t t);
12288 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
12289 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
12290 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
12291 int64_t pandn_sd (int64_t s, int64_t t);
12292 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
12293 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
12294 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
12295 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
12296 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
12297 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
12298 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
12299 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
12300 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
12301 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
12302 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
12303 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
12304 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
12305 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
12306 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
12307 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
12308 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
12309 uint16x4_t pextrh_u (uint16x4_t s, int field);
12310 int16x4_t pextrh_s (int16x4_t s, int field);
12311 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
12312 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
12313 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
12314 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
12315 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
12316 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
12317 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
12318 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
12319 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
12320 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
12321 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
12322 int16x4_t pminsh (int16x4_t s, int16x4_t t);
12323 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
12324 uint8x8_t pmovmskb_u (uint8x8_t s);
12325 int8x8_t pmovmskb_s (int8x8_t s);
12326 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
12327 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
12328 int16x4_t pmullh (int16x4_t s, int16x4_t t);
12329 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
12330 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
12331 uint16x4_t biadd (uint8x8_t s);
12332 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
12333 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
12334 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
12335 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
12336 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
12337 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
12338 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
12339 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
12340 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
12341 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
12342 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
12343 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
12344 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
12345 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
12346 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
12347 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
12348 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
12349 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
12350 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
12351 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
12352 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
12353 uint64_t psubd_u (uint64_t s, uint64_t t);
12354 int64_t psubd_s (int64_t s, int64_t t);
12355 int16x4_t psubsh (int16x4_t s, int16x4_t t);
12356 int8x8_t psubsb (int8x8_t s, int8x8_t t);
12357 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
12358 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
12359 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
12360 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
12361 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
12362 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
12363 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
12364 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
12365 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
12366 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
12367 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
12368 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
12369 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
12370 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
12374 * Paired-Single Arithmetic::
12375 * Paired-Single Built-in Functions::
12376 * MIPS-3D Built-in Functions::
12379 @node Paired-Single Arithmetic
12380 @subsubsection Paired-Single Arithmetic
12382 The table below lists the @code{v2sf} operations for which hardware
12383 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
12384 values and @code{x} is an integral value.
12386 @multitable @columnfractions .50 .50
12387 @item C code @tab MIPS instruction
12388 @item @code{a + b} @tab @code{add.ps}
12389 @item @code{a - b} @tab @code{sub.ps}
12390 @item @code{-a} @tab @code{neg.ps}
12391 @item @code{a * b} @tab @code{mul.ps}
12392 @item @code{a * b + c} @tab @code{madd.ps}
12393 @item @code{a * b - c} @tab @code{msub.ps}
12394 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
12395 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
12396 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
12399 Note that the multiply-accumulate instructions can be disabled
12400 using the command-line option @code{-mno-fused-madd}.
12402 @node Paired-Single Built-in Functions
12403 @subsubsection Paired-Single Built-in Functions
12405 The following paired-single functions map directly to a particular
12406 MIPS instruction. Please refer to the architecture specification
12407 for details on what each instruction does.
12410 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
12411 Pair lower lower (@code{pll.ps}).
12413 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
12414 Pair upper lower (@code{pul.ps}).
12416 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
12417 Pair lower upper (@code{plu.ps}).
12419 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
12420 Pair upper upper (@code{puu.ps}).
12422 @item v2sf __builtin_mips_cvt_ps_s (float, float)
12423 Convert pair to paired single (@code{cvt.ps.s}).
12425 @item float __builtin_mips_cvt_s_pl (v2sf)
12426 Convert pair lower to single (@code{cvt.s.pl}).
12428 @item float __builtin_mips_cvt_s_pu (v2sf)
12429 Convert pair upper to single (@code{cvt.s.pu}).
12431 @item v2sf __builtin_mips_abs_ps (v2sf)
12432 Absolute value (@code{abs.ps}).
12434 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
12435 Align variable (@code{alnv.ps}).
12437 @emph{Note:} The value of the third parameter must be 0 or 4
12438 modulo 8, otherwise the result is unpredictable. Please read the
12439 instruction description for details.
12442 The following multi-instruction functions are also available.
12443 In each case, @var{cond} can be any of the 16 floating-point conditions:
12444 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
12445 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
12446 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
12449 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12450 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12451 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
12452 @code{movt.ps}/@code{movf.ps}).
12454 The @code{movt} functions return the value @var{x} computed by:
12457 c.@var{cond}.ps @var{cc},@var{a},@var{b}
12458 mov.ps @var{x},@var{c}
12459 movt.ps @var{x},@var{d},@var{cc}
12462 The @code{movf} functions are similar but use @code{movf.ps} instead
12465 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12466 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12467 Comparison of two paired-single values (@code{c.@var{cond}.ps},
12468 @code{bc1t}/@code{bc1f}).
12470 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
12471 and return either the upper or lower half of the result. For example:
12475 if (__builtin_mips_upper_c_eq_ps (a, b))
12476 upper_halves_are_equal ();
12478 upper_halves_are_unequal ();
12480 if (__builtin_mips_lower_c_eq_ps (a, b))
12481 lower_halves_are_equal ();
12483 lower_halves_are_unequal ();
12487 @node MIPS-3D Built-in Functions
12488 @subsubsection MIPS-3D Built-in Functions
12490 The MIPS-3D Application-Specific Extension (ASE) includes additional
12491 paired-single instructions that are designed to improve the performance
12492 of 3D graphics operations. Support for these instructions is controlled
12493 by the @option{-mips3d} command-line option.
12495 The functions listed below map directly to a particular MIPS-3D
12496 instruction. Please refer to the architecture specification for
12497 more details on what each instruction does.
12500 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
12501 Reduction add (@code{addr.ps}).
12503 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
12504 Reduction multiply (@code{mulr.ps}).
12506 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
12507 Convert paired single to paired word (@code{cvt.pw.ps}).
12509 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
12510 Convert paired word to paired single (@code{cvt.ps.pw}).
12512 @item float __builtin_mips_recip1_s (float)
12513 @itemx double __builtin_mips_recip1_d (double)
12514 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
12515 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
12517 @item float __builtin_mips_recip2_s (float, float)
12518 @itemx double __builtin_mips_recip2_d (double, double)
12519 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
12520 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
12522 @item float __builtin_mips_rsqrt1_s (float)
12523 @itemx double __builtin_mips_rsqrt1_d (double)
12524 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
12525 Reduced-precision reciprocal square root (sequence step 1)
12526 (@code{rsqrt1.@var{fmt}}).
12528 @item float __builtin_mips_rsqrt2_s (float, float)
12529 @itemx double __builtin_mips_rsqrt2_d (double, double)
12530 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
12531 Reduced-precision reciprocal square root (sequence step 2)
12532 (@code{rsqrt2.@var{fmt}}).
12535 The following multi-instruction functions are also available.
12536 In each case, @var{cond} can be any of the 16 floating-point conditions:
12537 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
12538 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
12539 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
12542 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
12543 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
12544 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
12545 @code{bc1t}/@code{bc1f}).
12547 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
12548 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
12553 if (__builtin_mips_cabs_eq_s (a, b))
12559 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12560 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12561 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
12562 @code{bc1t}/@code{bc1f}).
12564 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
12565 and return either the upper or lower half of the result. For example:
12569 if (__builtin_mips_upper_cabs_eq_ps (a, b))
12570 upper_halves_are_equal ();
12572 upper_halves_are_unequal ();
12574 if (__builtin_mips_lower_cabs_eq_ps (a, b))
12575 lower_halves_are_equal ();
12577 lower_halves_are_unequal ();
12580 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12581 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12582 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
12583 @code{movt.ps}/@code{movf.ps}).
12585 The @code{movt} functions return the value @var{x} computed by:
12588 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
12589 mov.ps @var{x},@var{c}
12590 movt.ps @var{x},@var{d},@var{cc}
12593 The @code{movf} functions are similar but use @code{movf.ps} instead
12596 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12597 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12598 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12599 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
12600 Comparison of two paired-single values
12601 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
12602 @code{bc1any2t}/@code{bc1any2f}).
12604 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
12605 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
12606 result is true and the @code{all} forms return true if both results are true.
12611 if (__builtin_mips_any_c_eq_ps (a, b))
12616 if (__builtin_mips_all_c_eq_ps (a, b))
12622 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12623 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12624 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12625 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
12626 Comparison of four paired-single values
12627 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
12628 @code{bc1any4t}/@code{bc1any4f}).
12630 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
12631 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
12632 The @code{any} forms return true if any of the four results are true
12633 and the @code{all} forms return true if all four results are true.
12638 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
12643 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
12650 @node Other MIPS Built-in Functions
12651 @subsection Other MIPS Built-in Functions
12653 GCC provides other MIPS-specific built-in functions:
12656 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
12657 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
12658 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
12659 when this function is available.
12661 @item unsigned int __builtin_mips_get_fcsr (void)
12662 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
12663 Get and set the contents of the floating-point control and status register
12664 (FPU control register 31). These functions are only available in hard-float
12665 code but can be called in both MIPS16 and non-MIPS16 contexts.
12667 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
12668 register except the condition codes, which GCC assumes are preserved.
12671 @node MSP430 Built-in Functions
12672 @subsection MSP430 Built-in Functions
12674 GCC provides a couple of special builtin functions to aid in the
12675 writing of interrupt handlers in C.
12678 @item __bic_SR_register_on_exit (int @var{mask})
12679 This clears the indicated bits in the saved copy of the status register
12680 currently residing on the stack. This only works inside interrupt
12681 handlers and the changes to the status register will only take affect
12682 once the handler returns.
12684 @item __bis_SR_register_on_exit (int @var{mask})
12685 This sets the indicated bits in the saved copy of the status register
12686 currently residing on the stack. This only works inside interrupt
12687 handlers and the changes to the status register will only take affect
12688 once the handler returns.
12691 @node NDS32 Built-in Functions
12692 @subsection NDS32 Built-in Functions
12694 These built-in functions are available for the NDS32 target:
12696 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
12697 Insert an ISYNC instruction into the instruction stream where
12698 @var{addr} is an instruction address for serialization.
12701 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
12702 Insert an ISB instruction into the instruction stream.
12705 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
12706 Return the content of a system register which is mapped by @var{sr}.
12709 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
12710 Return the content of a user space register which is mapped by @var{usr}.
12713 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
12714 Move the @var{value} to a system register which is mapped by @var{sr}.
12717 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
12718 Move the @var{value} to a user space register which is mapped by @var{usr}.
12721 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
12722 Enable global interrupt.
12725 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
12726 Disable global interrupt.
12729 @node picoChip Built-in Functions
12730 @subsection picoChip Built-in Functions
12732 GCC provides an interface to selected machine instructions from the
12733 picoChip instruction set.
12736 @item int __builtin_sbc (int @var{value})
12737 Sign bit count. Return the number of consecutive bits in @var{value}
12738 that have the same value as the sign bit. The result is the number of
12739 leading sign bits minus one, giving the number of redundant sign bits in
12742 @item int __builtin_byteswap (int @var{value})
12743 Byte swap. Return the result of swapping the upper and lower bytes of
12746 @item int __builtin_brev (int @var{value})
12747 Bit reversal. Return the result of reversing the bits in
12748 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
12751 @item int __builtin_adds (int @var{x}, int @var{y})
12752 Saturating addition. Return the result of adding @var{x} and @var{y},
12753 storing the value 32767 if the result overflows.
12755 @item int __builtin_subs (int @var{x}, int @var{y})
12756 Saturating subtraction. Return the result of subtracting @var{y} from
12757 @var{x}, storing the value @minus{}32768 if the result overflows.
12759 @item void __builtin_halt (void)
12760 Halt. The processor stops execution. This built-in is useful for
12761 implementing assertions.
12765 @node PowerPC Built-in Functions
12766 @subsection PowerPC Built-in Functions
12768 These built-in functions are available for the PowerPC family of
12771 float __builtin_recipdivf (float, float);
12772 float __builtin_rsqrtf (float);
12773 double __builtin_recipdiv (double, double);
12774 double __builtin_rsqrt (double);
12775 long __builtin_bpermd (long, long);
12776 uint64_t __builtin_ppc_get_timebase ();
12777 unsigned long __builtin_ppc_mftb ();
12780 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
12781 @code{__builtin_rsqrtf} functions generate multiple instructions to
12782 implement the reciprocal sqrt functionality using reciprocal sqrt
12783 estimate instructions.
12785 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
12786 functions generate multiple instructions to implement division using
12787 the reciprocal estimate instructions.
12789 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
12790 functions generate instructions to read the Time Base Register. The
12791 @code{__builtin_ppc_get_timebase} function may generate multiple
12792 instructions and always returns the 64 bits of the Time Base Register.
12793 The @code{__builtin_ppc_mftb} function always generates one instruction and
12794 returns the Time Base Register value as an unsigned long, throwing away
12795 the most significant word on 32-bit environments.
12797 @node PowerPC AltiVec/VSX Built-in Functions
12798 @subsection PowerPC AltiVec Built-in Functions
12800 GCC provides an interface for the PowerPC family of processors to access
12801 the AltiVec operations described in Motorola's AltiVec Programming
12802 Interface Manual. The interface is made available by including
12803 @code{<altivec.h>} and using @option{-maltivec} and
12804 @option{-mabi=altivec}. The interface supports the following vector
12808 vector unsigned char
12812 vector unsigned short
12813 vector signed short
12817 vector unsigned int
12823 If @option{-mvsx} is used the following additional vector types are
12827 vector unsigned long
12832 The long types are only implemented for 64-bit code generation, and
12833 the long type is only used in the floating point/integer conversion
12836 GCC's implementation of the high-level language interface available from
12837 C and C++ code differs from Motorola's documentation in several ways.
12842 A vector constant is a list of constant expressions within curly braces.
12845 A vector initializer requires no cast if the vector constant is of the
12846 same type as the variable it is initializing.
12849 If @code{signed} or @code{unsigned} is omitted, the signedness of the
12850 vector type is the default signedness of the base type. The default
12851 varies depending on the operating system, so a portable program should
12852 always specify the signedness.
12855 Compiling with @option{-maltivec} adds keywords @code{__vector},
12856 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
12857 @code{bool}. When compiling ISO C, the context-sensitive substitution
12858 of the keywords @code{vector}, @code{pixel} and @code{bool} is
12859 disabled. To use them, you must include @code{<altivec.h>} instead.
12862 GCC allows using a @code{typedef} name as the type specifier for a
12866 For C, overloaded functions are implemented with macros so the following
12870 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
12874 Since @code{vec_add} is a macro, the vector constant in the example
12875 is treated as four separate arguments. Wrap the entire argument in
12876 parentheses for this to work.
12879 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
12880 Internally, GCC uses built-in functions to achieve the functionality in
12881 the aforementioned header file, but they are not supported and are
12882 subject to change without notice.
12884 The following interfaces are supported for the generic and specific
12885 AltiVec operations and the AltiVec predicates. In cases where there
12886 is a direct mapping between generic and specific operations, only the
12887 generic names are shown here, although the specific operations can also
12890 Arguments that are documented as @code{const int} require literal
12891 integral values within the range required for that operation.
12894 vector signed char vec_abs (vector signed char);
12895 vector signed short vec_abs (vector signed short);
12896 vector signed int vec_abs (vector signed int);
12897 vector float vec_abs (vector float);
12899 vector signed char vec_abss (vector signed char);
12900 vector signed short vec_abss (vector signed short);
12901 vector signed int vec_abss (vector signed int);
12903 vector signed char vec_add (vector bool char, vector signed char);
12904 vector signed char vec_add (vector signed char, vector bool char);
12905 vector signed char vec_add (vector signed char, vector signed char);
12906 vector unsigned char vec_add (vector bool char, vector unsigned char);
12907 vector unsigned char vec_add (vector unsigned char, vector bool char);
12908 vector unsigned char vec_add (vector unsigned char,
12909 vector unsigned char);
12910 vector signed short vec_add (vector bool short, vector signed short);
12911 vector signed short vec_add (vector signed short, vector bool short);
12912 vector signed short vec_add (vector signed short, vector signed short);
12913 vector unsigned short vec_add (vector bool short,
12914 vector unsigned short);
12915 vector unsigned short vec_add (vector unsigned short,
12916 vector bool short);
12917 vector unsigned short vec_add (vector unsigned short,
12918 vector unsigned short);
12919 vector signed int vec_add (vector bool int, vector signed int);
12920 vector signed int vec_add (vector signed int, vector bool int);
12921 vector signed int vec_add (vector signed int, vector signed int);
12922 vector unsigned int vec_add (vector bool int, vector unsigned int);
12923 vector unsigned int vec_add (vector unsigned int, vector bool int);
12924 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
12925 vector float vec_add (vector float, vector float);
12927 vector float vec_vaddfp (vector float, vector float);
12929 vector signed int vec_vadduwm (vector bool int, vector signed int);
12930 vector signed int vec_vadduwm (vector signed int, vector bool int);
12931 vector signed int vec_vadduwm (vector signed int, vector signed int);
12932 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
12933 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
12934 vector unsigned int vec_vadduwm (vector unsigned int,
12935 vector unsigned int);
12937 vector signed short vec_vadduhm (vector bool short,
12938 vector signed short);
12939 vector signed short vec_vadduhm (vector signed short,
12940 vector bool short);
12941 vector signed short vec_vadduhm (vector signed short,
12942 vector signed short);
12943 vector unsigned short vec_vadduhm (vector bool short,
12944 vector unsigned short);
12945 vector unsigned short vec_vadduhm (vector unsigned short,
12946 vector bool short);
12947 vector unsigned short vec_vadduhm (vector unsigned short,
12948 vector unsigned short);
12950 vector signed char vec_vaddubm (vector bool char, vector signed char);
12951 vector signed char vec_vaddubm (vector signed char, vector bool char);
12952 vector signed char vec_vaddubm (vector signed char, vector signed char);
12953 vector unsigned char vec_vaddubm (vector bool char,
12954 vector unsigned char);
12955 vector unsigned char vec_vaddubm (vector unsigned char,
12957 vector unsigned char vec_vaddubm (vector unsigned char,
12958 vector unsigned char);
12960 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
12962 vector unsigned char vec_adds (vector bool char, vector unsigned char);
12963 vector unsigned char vec_adds (vector unsigned char, vector bool char);
12964 vector unsigned char vec_adds (vector unsigned char,
12965 vector unsigned char);
12966 vector signed char vec_adds (vector bool char, vector signed char);
12967 vector signed char vec_adds (vector signed char, vector bool char);
12968 vector signed char vec_adds (vector signed char, vector signed char);
12969 vector unsigned short vec_adds (vector bool short,
12970 vector unsigned short);
12971 vector unsigned short vec_adds (vector unsigned short,
12972 vector bool short);
12973 vector unsigned short vec_adds (vector unsigned short,
12974 vector unsigned short);
12975 vector signed short vec_adds (vector bool short, vector signed short);
12976 vector signed short vec_adds (vector signed short, vector bool short);
12977 vector signed short vec_adds (vector signed short, vector signed short);
12978 vector unsigned int vec_adds (vector bool int, vector unsigned int);
12979 vector unsigned int vec_adds (vector unsigned int, vector bool int);
12980 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
12981 vector signed int vec_adds (vector bool int, vector signed int);
12982 vector signed int vec_adds (vector signed int, vector bool int);
12983 vector signed int vec_adds (vector signed int, vector signed int);
12985 vector signed int vec_vaddsws (vector bool int, vector signed int);
12986 vector signed int vec_vaddsws (vector signed int, vector bool int);
12987 vector signed int vec_vaddsws (vector signed int, vector signed int);
12989 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
12990 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
12991 vector unsigned int vec_vadduws (vector unsigned int,
12992 vector unsigned int);
12994 vector signed short vec_vaddshs (vector bool short,
12995 vector signed short);
12996 vector signed short vec_vaddshs (vector signed short,
12997 vector bool short);
12998 vector signed short vec_vaddshs (vector signed short,
12999 vector signed short);
13001 vector unsigned short vec_vadduhs (vector bool short,
13002 vector unsigned short);
13003 vector unsigned short vec_vadduhs (vector unsigned short,
13004 vector bool short);
13005 vector unsigned short vec_vadduhs (vector unsigned short,
13006 vector unsigned short);
13008 vector signed char vec_vaddsbs (vector bool char, vector signed char);
13009 vector signed char vec_vaddsbs (vector signed char, vector bool char);
13010 vector signed char vec_vaddsbs (vector signed char, vector signed char);
13012 vector unsigned char vec_vaddubs (vector bool char,
13013 vector unsigned char);
13014 vector unsigned char vec_vaddubs (vector unsigned char,
13016 vector unsigned char vec_vaddubs (vector unsigned char,
13017 vector unsigned char);
13019 vector float vec_and (vector float, vector float);
13020 vector float vec_and (vector float, vector bool int);
13021 vector float vec_and (vector bool int, vector float);
13022 vector bool int vec_and (vector bool int, vector bool int);
13023 vector signed int vec_and (vector bool int, vector signed int);
13024 vector signed int vec_and (vector signed int, vector bool int);
13025 vector signed int vec_and (vector signed int, vector signed int);
13026 vector unsigned int vec_and (vector bool int, vector unsigned int);
13027 vector unsigned int vec_and (vector unsigned int, vector bool int);
13028 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
13029 vector bool short vec_and (vector bool short, vector bool short);
13030 vector signed short vec_and (vector bool short, vector signed short);
13031 vector signed short vec_and (vector signed short, vector bool short);
13032 vector signed short vec_and (vector signed short, vector signed short);
13033 vector unsigned short vec_and (vector bool short,
13034 vector unsigned short);
13035 vector unsigned short vec_and (vector unsigned short,
13036 vector bool short);
13037 vector unsigned short vec_and (vector unsigned short,
13038 vector unsigned short);
13039 vector signed char vec_and (vector bool char, vector signed char);
13040 vector bool char vec_and (vector bool char, vector bool char);
13041 vector signed char vec_and (vector signed char, vector bool char);
13042 vector signed char vec_and (vector signed char, vector signed char);
13043 vector unsigned char vec_and (vector bool char, vector unsigned char);
13044 vector unsigned char vec_and (vector unsigned char, vector bool char);
13045 vector unsigned char vec_and (vector unsigned char,
13046 vector unsigned char);
13048 vector float vec_andc (vector float, vector float);
13049 vector float vec_andc (vector float, vector bool int);
13050 vector float vec_andc (vector bool int, vector float);
13051 vector bool int vec_andc (vector bool int, vector bool int);
13052 vector signed int vec_andc (vector bool int, vector signed int);
13053 vector signed int vec_andc (vector signed int, vector bool int);
13054 vector signed int vec_andc (vector signed int, vector signed int);
13055 vector unsigned int vec_andc (vector bool int, vector unsigned int);
13056 vector unsigned int vec_andc (vector unsigned int, vector bool int);
13057 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
13058 vector bool short vec_andc (vector bool short, vector bool short);
13059 vector signed short vec_andc (vector bool short, vector signed short);
13060 vector signed short vec_andc (vector signed short, vector bool short);
13061 vector signed short vec_andc (vector signed short, vector signed short);
13062 vector unsigned short vec_andc (vector bool short,
13063 vector unsigned short);
13064 vector unsigned short vec_andc (vector unsigned short,
13065 vector bool short);
13066 vector unsigned short vec_andc (vector unsigned short,
13067 vector unsigned short);
13068 vector signed char vec_andc (vector bool char, vector signed char);
13069 vector bool char vec_andc (vector bool char, vector bool char);
13070 vector signed char vec_andc (vector signed char, vector bool char);
13071 vector signed char vec_andc (vector signed char, vector signed char);
13072 vector unsigned char vec_andc (vector bool char, vector unsigned char);
13073 vector unsigned char vec_andc (vector unsigned char, vector bool char);
13074 vector unsigned char vec_andc (vector unsigned char,
13075 vector unsigned char);
13077 vector unsigned char vec_avg (vector unsigned char,
13078 vector unsigned char);
13079 vector signed char vec_avg (vector signed char, vector signed char);
13080 vector unsigned short vec_avg (vector unsigned short,
13081 vector unsigned short);
13082 vector signed short vec_avg (vector signed short, vector signed short);
13083 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
13084 vector signed int vec_avg (vector signed int, vector signed int);
13086 vector signed int vec_vavgsw (vector signed int, vector signed int);
13088 vector unsigned int vec_vavguw (vector unsigned int,
13089 vector unsigned int);
13091 vector signed short vec_vavgsh (vector signed short,
13092 vector signed short);
13094 vector unsigned short vec_vavguh (vector unsigned short,
13095 vector unsigned short);
13097 vector signed char vec_vavgsb (vector signed char, vector signed char);
13099 vector unsigned char vec_vavgub (vector unsigned char,
13100 vector unsigned char);
13102 vector float vec_copysign (vector float);
13104 vector float vec_ceil (vector float);
13106 vector signed int vec_cmpb (vector float, vector float);
13108 vector bool char vec_cmpeq (vector signed char, vector signed char);
13109 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
13110 vector bool short vec_cmpeq (vector signed short, vector signed short);
13111 vector bool short vec_cmpeq (vector unsigned short,
13112 vector unsigned short);
13113 vector bool int vec_cmpeq (vector signed int, vector signed int);
13114 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
13115 vector bool int vec_cmpeq (vector float, vector float);
13117 vector bool int vec_vcmpeqfp (vector float, vector float);
13119 vector bool int vec_vcmpequw (vector signed int, vector signed int);
13120 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
13122 vector bool short vec_vcmpequh (vector signed short,
13123 vector signed short);
13124 vector bool short vec_vcmpequh (vector unsigned short,
13125 vector unsigned short);
13127 vector bool char vec_vcmpequb (vector signed char, vector signed char);
13128 vector bool char vec_vcmpequb (vector unsigned char,
13129 vector unsigned char);
13131 vector bool int vec_cmpge (vector float, vector float);
13133 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
13134 vector bool char vec_cmpgt (vector signed char, vector signed char);
13135 vector bool short vec_cmpgt (vector unsigned short,
13136 vector unsigned short);
13137 vector bool short vec_cmpgt (vector signed short, vector signed short);
13138 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
13139 vector bool int vec_cmpgt (vector signed int, vector signed int);
13140 vector bool int vec_cmpgt (vector float, vector float);
13142 vector bool int vec_vcmpgtfp (vector float, vector float);
13144 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
13146 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
13148 vector bool short vec_vcmpgtsh (vector signed short,
13149 vector signed short);
13151 vector bool short vec_vcmpgtuh (vector unsigned short,
13152 vector unsigned short);
13154 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
13156 vector bool char vec_vcmpgtub (vector unsigned char,
13157 vector unsigned char);
13159 vector bool int vec_cmple (vector float, vector float);
13161 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
13162 vector bool char vec_cmplt (vector signed char, vector signed char);
13163 vector bool short vec_cmplt (vector unsigned short,
13164 vector unsigned short);
13165 vector bool short vec_cmplt (vector signed short, vector signed short);
13166 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
13167 vector bool int vec_cmplt (vector signed int, vector signed int);
13168 vector bool int vec_cmplt (vector float, vector float);
13170 vector float vec_ctf (vector unsigned int, const int);
13171 vector float vec_ctf (vector signed int, const int);
13173 vector float vec_vcfsx (vector signed int, const int);
13175 vector float vec_vcfux (vector unsigned int, const int);
13177 vector signed int vec_cts (vector float, const int);
13179 vector unsigned int vec_ctu (vector float, const int);
13181 void vec_dss (const int);
13183 void vec_dssall (void);
13185 void vec_dst (const vector unsigned char *, int, const int);
13186 void vec_dst (const vector signed char *, int, const int);
13187 void vec_dst (const vector bool char *, int, const int);
13188 void vec_dst (const vector unsigned short *, int, const int);
13189 void vec_dst (const vector signed short *, int, const int);
13190 void vec_dst (const vector bool short *, int, const int);
13191 void vec_dst (const vector pixel *, int, const int);
13192 void vec_dst (const vector unsigned int *, int, const int);
13193 void vec_dst (const vector signed int *, int, const int);
13194 void vec_dst (const vector bool int *, int, const int);
13195 void vec_dst (const vector float *, int, const int);
13196 void vec_dst (const unsigned char *, int, const int);
13197 void vec_dst (const signed char *, int, const int);
13198 void vec_dst (const unsigned short *, int, const int);
13199 void vec_dst (const short *, int, const int);
13200 void vec_dst (const unsigned int *, int, const int);
13201 void vec_dst (const int *, int, const int);
13202 void vec_dst (const unsigned long *, int, const int);
13203 void vec_dst (const long *, int, const int);
13204 void vec_dst (const float *, int, const int);
13206 void vec_dstst (const vector unsigned char *, int, const int);
13207 void vec_dstst (const vector signed char *, int, const int);
13208 void vec_dstst (const vector bool char *, int, const int);
13209 void vec_dstst (const vector unsigned short *, int, const int);
13210 void vec_dstst (const vector signed short *, int, const int);
13211 void vec_dstst (const vector bool short *, int, const int);
13212 void vec_dstst (const vector pixel *, int, const int);
13213 void vec_dstst (const vector unsigned int *, int, const int);
13214 void vec_dstst (const vector signed int *, int, const int);
13215 void vec_dstst (const vector bool int *, int, const int);
13216 void vec_dstst (const vector float *, int, const int);
13217 void vec_dstst (const unsigned char *, int, const int);
13218 void vec_dstst (const signed char *, int, const int);
13219 void vec_dstst (const unsigned short *, int, const int);
13220 void vec_dstst (const short *, int, const int);
13221 void vec_dstst (const unsigned int *, int, const int);
13222 void vec_dstst (const int *, int, const int);
13223 void vec_dstst (const unsigned long *, int, const int);
13224 void vec_dstst (const long *, int, const int);
13225 void vec_dstst (const float *, int, const int);
13227 void vec_dststt (const vector unsigned char *, int, const int);
13228 void vec_dststt (const vector signed char *, int, const int);
13229 void vec_dststt (const vector bool char *, int, const int);
13230 void vec_dststt (const vector unsigned short *, int, const int);
13231 void vec_dststt (const vector signed short *, int, const int);
13232 void vec_dststt (const vector bool short *, int, const int);
13233 void vec_dststt (const vector pixel *, int, const int);
13234 void vec_dststt (const vector unsigned int *, int, const int);
13235 void vec_dststt (const vector signed int *, int, const int);
13236 void vec_dststt (const vector bool int *, int, const int);
13237 void vec_dststt (const vector float *, int, const int);
13238 void vec_dststt (const unsigned char *, int, const int);
13239 void vec_dststt (const signed char *, int, const int);
13240 void vec_dststt (const unsigned short *, int, const int);
13241 void vec_dststt (const short *, int, const int);
13242 void vec_dststt (const unsigned int *, int, const int);
13243 void vec_dststt (const int *, int, const int);
13244 void vec_dststt (const unsigned long *, int, const int);
13245 void vec_dststt (const long *, int, const int);
13246 void vec_dststt (const float *, int, const int);
13248 void vec_dstt (const vector unsigned char *, int, const int);
13249 void vec_dstt (const vector signed char *, int, const int);
13250 void vec_dstt (const vector bool char *, int, const int);
13251 void vec_dstt (const vector unsigned short *, int, const int);
13252 void vec_dstt (const vector signed short *, int, const int);
13253 void vec_dstt (const vector bool short *, int, const int);
13254 void vec_dstt (const vector pixel *, int, const int);
13255 void vec_dstt (const vector unsigned int *, int, const int);
13256 void vec_dstt (const vector signed int *, int, const int);
13257 void vec_dstt (const vector bool int *, int, const int);
13258 void vec_dstt (const vector float *, int, const int);
13259 void vec_dstt (const unsigned char *, int, const int);
13260 void vec_dstt (const signed char *, int, const int);
13261 void vec_dstt (const unsigned short *, int, const int);
13262 void vec_dstt (const short *, int, const int);
13263 void vec_dstt (const unsigned int *, int, const int);
13264 void vec_dstt (const int *, int, const int);
13265 void vec_dstt (const unsigned long *, int, const int);
13266 void vec_dstt (const long *, int, const int);
13267 void vec_dstt (const float *, int, const int);
13269 vector float vec_expte (vector float);
13271 vector float vec_floor (vector float);
13273 vector float vec_ld (int, const vector float *);
13274 vector float vec_ld (int, const float *);
13275 vector bool int vec_ld (int, const vector bool int *);
13276 vector signed int vec_ld (int, const vector signed int *);
13277 vector signed int vec_ld (int, const int *);
13278 vector signed int vec_ld (int, const long *);
13279 vector unsigned int vec_ld (int, const vector unsigned int *);
13280 vector unsigned int vec_ld (int, const unsigned int *);
13281 vector unsigned int vec_ld (int, const unsigned long *);
13282 vector bool short vec_ld (int, const vector bool short *);
13283 vector pixel vec_ld (int, const vector pixel *);
13284 vector signed short vec_ld (int, const vector signed short *);
13285 vector signed short vec_ld (int, const short *);
13286 vector unsigned short vec_ld (int, const vector unsigned short *);
13287 vector unsigned short vec_ld (int, const unsigned short *);
13288 vector bool char vec_ld (int, const vector bool char *);
13289 vector signed char vec_ld (int, const vector signed char *);
13290 vector signed char vec_ld (int, const signed char *);
13291 vector unsigned char vec_ld (int, const vector unsigned char *);
13292 vector unsigned char vec_ld (int, const unsigned char *);
13294 vector signed char vec_lde (int, const signed char *);
13295 vector unsigned char vec_lde (int, const unsigned char *);
13296 vector signed short vec_lde (int, const short *);
13297 vector unsigned short vec_lde (int, const unsigned short *);
13298 vector float vec_lde (int, const float *);
13299 vector signed int vec_lde (int, const int *);
13300 vector unsigned int vec_lde (int, const unsigned int *);
13301 vector signed int vec_lde (int, const long *);
13302 vector unsigned int vec_lde (int, const unsigned long *);
13304 vector float vec_lvewx (int, float *);
13305 vector signed int vec_lvewx (int, int *);
13306 vector unsigned int vec_lvewx (int, unsigned int *);
13307 vector signed int vec_lvewx (int, long *);
13308 vector unsigned int vec_lvewx (int, unsigned long *);
13310 vector signed short vec_lvehx (int, short *);
13311 vector unsigned short vec_lvehx (int, unsigned short *);
13313 vector signed char vec_lvebx (int, char *);
13314 vector unsigned char vec_lvebx (int, unsigned char *);
13316 vector float vec_ldl (int, const vector float *);
13317 vector float vec_ldl (int, const float *);
13318 vector bool int vec_ldl (int, const vector bool int *);
13319 vector signed int vec_ldl (int, const vector signed int *);
13320 vector signed int vec_ldl (int, const int *);
13321 vector signed int vec_ldl (int, const long *);
13322 vector unsigned int vec_ldl (int, const vector unsigned int *);
13323 vector unsigned int vec_ldl (int, const unsigned int *);
13324 vector unsigned int vec_ldl (int, const unsigned long *);
13325 vector bool short vec_ldl (int, const vector bool short *);
13326 vector pixel vec_ldl (int, const vector pixel *);
13327 vector signed short vec_ldl (int, const vector signed short *);
13328 vector signed short vec_ldl (int, const short *);
13329 vector unsigned short vec_ldl (int, const vector unsigned short *);
13330 vector unsigned short vec_ldl (int, const unsigned short *);
13331 vector bool char vec_ldl (int, const vector bool char *);
13332 vector signed char vec_ldl (int, const vector signed char *);
13333 vector signed char vec_ldl (int, const signed char *);
13334 vector unsigned char vec_ldl (int, const vector unsigned char *);
13335 vector unsigned char vec_ldl (int, const unsigned char *);
13337 vector float vec_loge (vector float);
13339 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
13340 vector unsigned char vec_lvsl (int, const volatile signed char *);
13341 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
13342 vector unsigned char vec_lvsl (int, const volatile short *);
13343 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
13344 vector unsigned char vec_lvsl (int, const volatile int *);
13345 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
13346 vector unsigned char vec_lvsl (int, const volatile long *);
13347 vector unsigned char vec_lvsl (int, const volatile float *);
13349 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
13350 vector unsigned char vec_lvsr (int, const volatile signed char *);
13351 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
13352 vector unsigned char vec_lvsr (int, const volatile short *);
13353 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
13354 vector unsigned char vec_lvsr (int, const volatile int *);
13355 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
13356 vector unsigned char vec_lvsr (int, const volatile long *);
13357 vector unsigned char vec_lvsr (int, const volatile float *);
13359 vector float vec_madd (vector float, vector float, vector float);
13361 vector signed short vec_madds (vector signed short,
13362 vector signed short,
13363 vector signed short);
13365 vector unsigned char vec_max (vector bool char, vector unsigned char);
13366 vector unsigned char vec_max (vector unsigned char, vector bool char);
13367 vector unsigned char vec_max (vector unsigned char,
13368 vector unsigned char);
13369 vector signed char vec_max (vector bool char, vector signed char);
13370 vector signed char vec_max (vector signed char, vector bool char);
13371 vector signed char vec_max (vector signed char, vector signed char);
13372 vector unsigned short vec_max (vector bool short,
13373 vector unsigned short);
13374 vector unsigned short vec_max (vector unsigned short,
13375 vector bool short);
13376 vector unsigned short vec_max (vector unsigned short,
13377 vector unsigned short);
13378 vector signed short vec_max (vector bool short, vector signed short);
13379 vector signed short vec_max (vector signed short, vector bool short);
13380 vector signed short vec_max (vector signed short, vector signed short);
13381 vector unsigned int vec_max (vector bool int, vector unsigned int);
13382 vector unsigned int vec_max (vector unsigned int, vector bool int);
13383 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
13384 vector signed int vec_max (vector bool int, vector signed int);
13385 vector signed int vec_max (vector signed int, vector bool int);
13386 vector signed int vec_max (vector signed int, vector signed int);
13387 vector float vec_max (vector float, vector float);
13389 vector float vec_vmaxfp (vector float, vector float);
13391 vector signed int vec_vmaxsw (vector bool int, vector signed int);
13392 vector signed int vec_vmaxsw (vector signed int, vector bool int);
13393 vector signed int vec_vmaxsw (vector signed int, vector signed int);
13395 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
13396 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
13397 vector unsigned int vec_vmaxuw (vector unsigned int,
13398 vector unsigned int);
13400 vector signed short vec_vmaxsh (vector bool short, vector signed short);
13401 vector signed short vec_vmaxsh (vector signed short, vector bool short);
13402 vector signed short vec_vmaxsh (vector signed short,
13403 vector signed short);
13405 vector unsigned short vec_vmaxuh (vector bool short,
13406 vector unsigned short);
13407 vector unsigned short vec_vmaxuh (vector unsigned short,
13408 vector bool short);
13409 vector unsigned short vec_vmaxuh (vector unsigned short,
13410 vector unsigned short);
13412 vector signed char vec_vmaxsb (vector bool char, vector signed char);
13413 vector signed char vec_vmaxsb (vector signed char, vector bool char);
13414 vector signed char vec_vmaxsb (vector signed char, vector signed char);
13416 vector unsigned char vec_vmaxub (vector bool char,
13417 vector unsigned char);
13418 vector unsigned char vec_vmaxub (vector unsigned char,
13420 vector unsigned char vec_vmaxub (vector unsigned char,
13421 vector unsigned char);
13423 vector bool char vec_mergeh (vector bool char, vector bool char);
13424 vector signed char vec_mergeh (vector signed char, vector signed char);
13425 vector unsigned char vec_mergeh (vector unsigned char,
13426 vector unsigned char);
13427 vector bool short vec_mergeh (vector bool short, vector bool short);
13428 vector pixel vec_mergeh (vector pixel, vector pixel);
13429 vector signed short vec_mergeh (vector signed short,
13430 vector signed short);
13431 vector unsigned short vec_mergeh (vector unsigned short,
13432 vector unsigned short);
13433 vector float vec_mergeh (vector float, vector float);
13434 vector bool int vec_mergeh (vector bool int, vector bool int);
13435 vector signed int vec_mergeh (vector signed int, vector signed int);
13436 vector unsigned int vec_mergeh (vector unsigned int,
13437 vector unsigned int);
13439 vector float vec_vmrghw (vector float, vector float);
13440 vector bool int vec_vmrghw (vector bool int, vector bool int);
13441 vector signed int vec_vmrghw (vector signed int, vector signed int);
13442 vector unsigned int vec_vmrghw (vector unsigned int,
13443 vector unsigned int);
13445 vector bool short vec_vmrghh (vector bool short, vector bool short);
13446 vector signed short vec_vmrghh (vector signed short,
13447 vector signed short);
13448 vector unsigned short vec_vmrghh (vector unsigned short,
13449 vector unsigned short);
13450 vector pixel vec_vmrghh (vector pixel, vector pixel);
13452 vector bool char vec_vmrghb (vector bool char, vector bool char);
13453 vector signed char vec_vmrghb (vector signed char, vector signed char);
13454 vector unsigned char vec_vmrghb (vector unsigned char,
13455 vector unsigned char);
13457 vector bool char vec_mergel (vector bool char, vector bool char);
13458 vector signed char vec_mergel (vector signed char, vector signed char);
13459 vector unsigned char vec_mergel (vector unsigned char,
13460 vector unsigned char);
13461 vector bool short vec_mergel (vector bool short, vector bool short);
13462 vector pixel vec_mergel (vector pixel, vector pixel);
13463 vector signed short vec_mergel (vector signed short,
13464 vector signed short);
13465 vector unsigned short vec_mergel (vector unsigned short,
13466 vector unsigned short);
13467 vector float vec_mergel (vector float, vector float);
13468 vector bool int vec_mergel (vector bool int, vector bool int);
13469 vector signed int vec_mergel (vector signed int, vector signed int);
13470 vector unsigned int vec_mergel (vector unsigned int,
13471 vector unsigned int);
13473 vector float vec_vmrglw (vector float, vector float);
13474 vector signed int vec_vmrglw (vector signed int, vector signed int);
13475 vector unsigned int vec_vmrglw (vector unsigned int,
13476 vector unsigned int);
13477 vector bool int vec_vmrglw (vector bool int, vector bool int);
13479 vector bool short vec_vmrglh (vector bool short, vector bool short);
13480 vector signed short vec_vmrglh (vector signed short,
13481 vector signed short);
13482 vector unsigned short vec_vmrglh (vector unsigned short,
13483 vector unsigned short);
13484 vector pixel vec_vmrglh (vector pixel, vector pixel);
13486 vector bool char vec_vmrglb (vector bool char, vector bool char);
13487 vector signed char vec_vmrglb (vector signed char, vector signed char);
13488 vector unsigned char vec_vmrglb (vector unsigned char,
13489 vector unsigned char);
13491 vector unsigned short vec_mfvscr (void);
13493 vector unsigned char vec_min (vector bool char, vector unsigned char);
13494 vector unsigned char vec_min (vector unsigned char, vector bool char);
13495 vector unsigned char vec_min (vector unsigned char,
13496 vector unsigned char);
13497 vector signed char vec_min (vector bool char, vector signed char);
13498 vector signed char vec_min (vector signed char, vector bool char);
13499 vector signed char vec_min (vector signed char, vector signed char);
13500 vector unsigned short vec_min (vector bool short,
13501 vector unsigned short);
13502 vector unsigned short vec_min (vector unsigned short,
13503 vector bool short);
13504 vector unsigned short vec_min (vector unsigned short,
13505 vector unsigned short);
13506 vector signed short vec_min (vector bool short, vector signed short);
13507 vector signed short vec_min (vector signed short, vector bool short);
13508 vector signed short vec_min (vector signed short, vector signed short);
13509 vector unsigned int vec_min (vector bool int, vector unsigned int);
13510 vector unsigned int vec_min (vector unsigned int, vector bool int);
13511 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
13512 vector signed int vec_min (vector bool int, vector signed int);
13513 vector signed int vec_min (vector signed int, vector bool int);
13514 vector signed int vec_min (vector signed int, vector signed int);
13515 vector float vec_min (vector float, vector float);
13517 vector float vec_vminfp (vector float, vector float);
13519 vector signed int vec_vminsw (vector bool int, vector signed int);
13520 vector signed int vec_vminsw (vector signed int, vector bool int);
13521 vector signed int vec_vminsw (vector signed int, vector signed int);
13523 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
13524 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
13525 vector unsigned int vec_vminuw (vector unsigned int,
13526 vector unsigned int);
13528 vector signed short vec_vminsh (vector bool short, vector signed short);
13529 vector signed short vec_vminsh (vector signed short, vector bool short);
13530 vector signed short vec_vminsh (vector signed short,
13531 vector signed short);
13533 vector unsigned short vec_vminuh (vector bool short,
13534 vector unsigned short);
13535 vector unsigned short vec_vminuh (vector unsigned short,
13536 vector bool short);
13537 vector unsigned short vec_vminuh (vector unsigned short,
13538 vector unsigned short);
13540 vector signed char vec_vminsb (vector bool char, vector signed char);
13541 vector signed char vec_vminsb (vector signed char, vector bool char);
13542 vector signed char vec_vminsb (vector signed char, vector signed char);
13544 vector unsigned char vec_vminub (vector bool char,
13545 vector unsigned char);
13546 vector unsigned char vec_vminub (vector unsigned char,
13548 vector unsigned char vec_vminub (vector unsigned char,
13549 vector unsigned char);
13551 vector signed short vec_mladd (vector signed short,
13552 vector signed short,
13553 vector signed short);
13554 vector signed short vec_mladd (vector signed short,
13555 vector unsigned short,
13556 vector unsigned short);
13557 vector signed short vec_mladd (vector unsigned short,
13558 vector signed short,
13559 vector signed short);
13560 vector unsigned short vec_mladd (vector unsigned short,
13561 vector unsigned short,
13562 vector unsigned short);
13564 vector signed short vec_mradds (vector signed short,
13565 vector signed short,
13566 vector signed short);
13568 vector unsigned int vec_msum (vector unsigned char,
13569 vector unsigned char,
13570 vector unsigned int);
13571 vector signed int vec_msum (vector signed char,
13572 vector unsigned char,
13573 vector signed int);
13574 vector unsigned int vec_msum (vector unsigned short,
13575 vector unsigned short,
13576 vector unsigned int);
13577 vector signed int vec_msum (vector signed short,
13578 vector signed short,
13579 vector signed int);
13581 vector signed int vec_vmsumshm (vector signed short,
13582 vector signed short,
13583 vector signed int);
13585 vector unsigned int vec_vmsumuhm (vector unsigned short,
13586 vector unsigned short,
13587 vector unsigned int);
13589 vector signed int vec_vmsummbm (vector signed char,
13590 vector unsigned char,
13591 vector signed int);
13593 vector unsigned int vec_vmsumubm (vector unsigned char,
13594 vector unsigned char,
13595 vector unsigned int);
13597 vector unsigned int vec_msums (vector unsigned short,
13598 vector unsigned short,
13599 vector unsigned int);
13600 vector signed int vec_msums (vector signed short,
13601 vector signed short,
13602 vector signed int);
13604 vector signed int vec_vmsumshs (vector signed short,
13605 vector signed short,
13606 vector signed int);
13608 vector unsigned int vec_vmsumuhs (vector unsigned short,
13609 vector unsigned short,
13610 vector unsigned int);
13612 void vec_mtvscr (vector signed int);
13613 void vec_mtvscr (vector unsigned int);
13614 void vec_mtvscr (vector bool int);
13615 void vec_mtvscr (vector signed short);
13616 void vec_mtvscr (vector unsigned short);
13617 void vec_mtvscr (vector bool short);
13618 void vec_mtvscr (vector pixel);
13619 void vec_mtvscr (vector signed char);
13620 void vec_mtvscr (vector unsigned char);
13621 void vec_mtvscr (vector bool char);
13623 vector unsigned short vec_mule (vector unsigned char,
13624 vector unsigned char);
13625 vector signed short vec_mule (vector signed char,
13626 vector signed char);
13627 vector unsigned int vec_mule (vector unsigned short,
13628 vector unsigned short);
13629 vector signed int vec_mule (vector signed short, vector signed short);
13631 vector signed int vec_vmulesh (vector signed short,
13632 vector signed short);
13634 vector unsigned int vec_vmuleuh (vector unsigned short,
13635 vector unsigned short);
13637 vector signed short vec_vmulesb (vector signed char,
13638 vector signed char);
13640 vector unsigned short vec_vmuleub (vector unsigned char,
13641 vector unsigned char);
13643 vector unsigned short vec_mulo (vector unsigned char,
13644 vector unsigned char);
13645 vector signed short vec_mulo (vector signed char, vector signed char);
13646 vector unsigned int vec_mulo (vector unsigned short,
13647 vector unsigned short);
13648 vector signed int vec_mulo (vector signed short, vector signed short);
13650 vector signed int vec_vmulosh (vector signed short,
13651 vector signed short);
13653 vector unsigned int vec_vmulouh (vector unsigned short,
13654 vector unsigned short);
13656 vector signed short vec_vmulosb (vector signed char,
13657 vector signed char);
13659 vector unsigned short vec_vmuloub (vector unsigned char,
13660 vector unsigned char);
13662 vector float vec_nmsub (vector float, vector float, vector float);
13664 vector float vec_nor (vector float, vector float);
13665 vector signed int vec_nor (vector signed int, vector signed int);
13666 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
13667 vector bool int vec_nor (vector bool int, vector bool int);
13668 vector signed short vec_nor (vector signed short, vector signed short);
13669 vector unsigned short vec_nor (vector unsigned short,
13670 vector unsigned short);
13671 vector bool short vec_nor (vector bool short, vector bool short);
13672 vector signed char vec_nor (vector signed char, vector signed char);
13673 vector unsigned char vec_nor (vector unsigned char,
13674 vector unsigned char);
13675 vector bool char vec_nor (vector bool char, vector bool char);
13677 vector float vec_or (vector float, vector float);
13678 vector float vec_or (vector float, vector bool int);
13679 vector float vec_or (vector bool int, vector float);
13680 vector bool int vec_or (vector bool int, vector bool int);
13681 vector signed int vec_or (vector bool int, vector signed int);
13682 vector signed int vec_or (vector signed int, vector bool int);
13683 vector signed int vec_or (vector signed int, vector signed int);
13684 vector unsigned int vec_or (vector bool int, vector unsigned int);
13685 vector unsigned int vec_or (vector unsigned int, vector bool int);
13686 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
13687 vector bool short vec_or (vector bool short, vector bool short);
13688 vector signed short vec_or (vector bool short, vector signed short);
13689 vector signed short vec_or (vector signed short, vector bool short);
13690 vector signed short vec_or (vector signed short, vector signed short);
13691 vector unsigned short vec_or (vector bool short, vector unsigned short);
13692 vector unsigned short vec_or (vector unsigned short, vector bool short);
13693 vector unsigned short vec_or (vector unsigned short,
13694 vector unsigned short);
13695 vector signed char vec_or (vector bool char, vector signed char);
13696 vector bool char vec_or (vector bool char, vector bool char);
13697 vector signed char vec_or (vector signed char, vector bool char);
13698 vector signed char vec_or (vector signed char, vector signed char);
13699 vector unsigned char vec_or (vector bool char, vector unsigned char);
13700 vector unsigned char vec_or (vector unsigned char, vector bool char);
13701 vector unsigned char vec_or (vector unsigned char,
13702 vector unsigned char);
13704 vector signed char vec_pack (vector signed short, vector signed short);
13705 vector unsigned char vec_pack (vector unsigned short,
13706 vector unsigned short);
13707 vector bool char vec_pack (vector bool short, vector bool short);
13708 vector signed short vec_pack (vector signed int, vector signed int);
13709 vector unsigned short vec_pack (vector unsigned int,
13710 vector unsigned int);
13711 vector bool short vec_pack (vector bool int, vector bool int);
13713 vector bool short vec_vpkuwum (vector bool int, vector bool int);
13714 vector signed short vec_vpkuwum (vector signed int, vector signed int);
13715 vector unsigned short vec_vpkuwum (vector unsigned int,
13716 vector unsigned int);
13718 vector bool char vec_vpkuhum (vector bool short, vector bool short);
13719 vector signed char vec_vpkuhum (vector signed short,
13720 vector signed short);
13721 vector unsigned char vec_vpkuhum (vector unsigned short,
13722 vector unsigned short);
13724 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
13726 vector unsigned char vec_packs (vector unsigned short,
13727 vector unsigned short);
13728 vector signed char vec_packs (vector signed short, vector signed short);
13729 vector unsigned short vec_packs (vector unsigned int,
13730 vector unsigned int);
13731 vector signed short vec_packs (vector signed int, vector signed int);
13733 vector signed short vec_vpkswss (vector signed int, vector signed int);
13735 vector unsigned short vec_vpkuwus (vector unsigned int,
13736 vector unsigned int);
13738 vector signed char vec_vpkshss (vector signed short,
13739 vector signed short);
13741 vector unsigned char vec_vpkuhus (vector unsigned short,
13742 vector unsigned short);
13744 vector unsigned char vec_packsu (vector unsigned short,
13745 vector unsigned short);
13746 vector unsigned char vec_packsu (vector signed short,
13747 vector signed short);
13748 vector unsigned short vec_packsu (vector unsigned int,
13749 vector unsigned int);
13750 vector unsigned short vec_packsu (vector signed int, vector signed int);
13752 vector unsigned short vec_vpkswus (vector signed int,
13753 vector signed int);
13755 vector unsigned char vec_vpkshus (vector signed short,
13756 vector signed short);
13758 vector float vec_perm (vector float,
13760 vector unsigned char);
13761 vector signed int vec_perm (vector signed int,
13763 vector unsigned char);
13764 vector unsigned int vec_perm (vector unsigned int,
13765 vector unsigned int,
13766 vector unsigned char);
13767 vector bool int vec_perm (vector bool int,
13769 vector unsigned char);
13770 vector signed short vec_perm (vector signed short,
13771 vector signed short,
13772 vector unsigned char);
13773 vector unsigned short vec_perm (vector unsigned short,
13774 vector unsigned short,
13775 vector unsigned char);
13776 vector bool short vec_perm (vector bool short,
13778 vector unsigned char);
13779 vector pixel vec_perm (vector pixel,
13781 vector unsigned char);
13782 vector signed char vec_perm (vector signed char,
13783 vector signed char,
13784 vector unsigned char);
13785 vector unsigned char vec_perm (vector unsigned char,
13786 vector unsigned char,
13787 vector unsigned char);
13788 vector bool char vec_perm (vector bool char,
13790 vector unsigned char);
13792 vector float vec_re (vector float);
13794 vector signed char vec_rl (vector signed char,
13795 vector unsigned char);
13796 vector unsigned char vec_rl (vector unsigned char,
13797 vector unsigned char);
13798 vector signed short vec_rl (vector signed short, vector unsigned short);
13799 vector unsigned short vec_rl (vector unsigned short,
13800 vector unsigned short);
13801 vector signed int vec_rl (vector signed int, vector unsigned int);
13802 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
13804 vector signed int vec_vrlw (vector signed int, vector unsigned int);
13805 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
13807 vector signed short vec_vrlh (vector signed short,
13808 vector unsigned short);
13809 vector unsigned short vec_vrlh (vector unsigned short,
13810 vector unsigned short);
13812 vector signed char vec_vrlb (vector signed char, vector unsigned char);
13813 vector unsigned char vec_vrlb (vector unsigned char,
13814 vector unsigned char);
13816 vector float vec_round (vector float);
13818 vector float vec_recip (vector float, vector float);
13820 vector float vec_rsqrt (vector float);
13822 vector float vec_rsqrte (vector float);
13824 vector float vec_sel (vector float, vector float, vector bool int);
13825 vector float vec_sel (vector float, vector float, vector unsigned int);
13826 vector signed int vec_sel (vector signed int,
13829 vector signed int vec_sel (vector signed int,
13831 vector unsigned int);
13832 vector unsigned int vec_sel (vector unsigned int,
13833 vector unsigned int,
13835 vector unsigned int vec_sel (vector unsigned int,
13836 vector unsigned int,
13837 vector unsigned int);
13838 vector bool int vec_sel (vector bool int,
13841 vector bool int vec_sel (vector bool int,
13843 vector unsigned int);
13844 vector signed short vec_sel (vector signed short,
13845 vector signed short,
13846 vector bool short);
13847 vector signed short vec_sel (vector signed short,
13848 vector signed short,
13849 vector unsigned short);
13850 vector unsigned short vec_sel (vector unsigned short,
13851 vector unsigned short,
13852 vector bool short);
13853 vector unsigned short vec_sel (vector unsigned short,
13854 vector unsigned short,
13855 vector unsigned short);
13856 vector bool short vec_sel (vector bool short,
13858 vector bool short);
13859 vector bool short vec_sel (vector bool short,
13861 vector unsigned short);
13862 vector signed char vec_sel (vector signed char,
13863 vector signed char,
13865 vector signed char vec_sel (vector signed char,
13866 vector signed char,
13867 vector unsigned char);
13868 vector unsigned char vec_sel (vector unsigned char,
13869 vector unsigned char,
13871 vector unsigned char vec_sel (vector unsigned char,
13872 vector unsigned char,
13873 vector unsigned char);
13874 vector bool char vec_sel (vector bool char,
13877 vector bool char vec_sel (vector bool char,
13879 vector unsigned char);
13881 vector signed char vec_sl (vector signed char,
13882 vector unsigned char);
13883 vector unsigned char vec_sl (vector unsigned char,
13884 vector unsigned char);
13885 vector signed short vec_sl (vector signed short, vector unsigned short);
13886 vector unsigned short vec_sl (vector unsigned short,
13887 vector unsigned short);
13888 vector signed int vec_sl (vector signed int, vector unsigned int);
13889 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
13891 vector signed int vec_vslw (vector signed int, vector unsigned int);
13892 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
13894 vector signed short vec_vslh (vector signed short,
13895 vector unsigned short);
13896 vector unsigned short vec_vslh (vector unsigned short,
13897 vector unsigned short);
13899 vector signed char vec_vslb (vector signed char, vector unsigned char);
13900 vector unsigned char vec_vslb (vector unsigned char,
13901 vector unsigned char);
13903 vector float vec_sld (vector float, vector float, const int);
13904 vector signed int vec_sld (vector signed int,
13907 vector unsigned int vec_sld (vector unsigned int,
13908 vector unsigned int,
13910 vector bool int vec_sld (vector bool int,
13913 vector signed short vec_sld (vector signed short,
13914 vector signed short,
13916 vector unsigned short vec_sld (vector unsigned short,
13917 vector unsigned short,
13919 vector bool short vec_sld (vector bool short,
13922 vector pixel vec_sld (vector pixel,
13925 vector signed char vec_sld (vector signed char,
13926 vector signed char,
13928 vector unsigned char vec_sld (vector unsigned char,
13929 vector unsigned char,
13931 vector bool char vec_sld (vector bool char,
13935 vector signed int vec_sll (vector signed int,
13936 vector unsigned int);
13937 vector signed int vec_sll (vector signed int,
13938 vector unsigned short);
13939 vector signed int vec_sll (vector signed int,
13940 vector unsigned char);
13941 vector unsigned int vec_sll (vector unsigned int,
13942 vector unsigned int);
13943 vector unsigned int vec_sll (vector unsigned int,
13944 vector unsigned short);
13945 vector unsigned int vec_sll (vector unsigned int,
13946 vector unsigned char);
13947 vector bool int vec_sll (vector bool int,
13948 vector unsigned int);
13949 vector bool int vec_sll (vector bool int,
13950 vector unsigned short);
13951 vector bool int vec_sll (vector bool int,
13952 vector unsigned char);
13953 vector signed short vec_sll (vector signed short,
13954 vector unsigned int);
13955 vector signed short vec_sll (vector signed short,
13956 vector unsigned short);
13957 vector signed short vec_sll (vector signed short,
13958 vector unsigned char);
13959 vector unsigned short vec_sll (vector unsigned short,
13960 vector unsigned int);
13961 vector unsigned short vec_sll (vector unsigned short,
13962 vector unsigned short);
13963 vector unsigned short vec_sll (vector unsigned short,
13964 vector unsigned char);
13965 vector bool short vec_sll (vector bool short, vector unsigned int);
13966 vector bool short vec_sll (vector bool short, vector unsigned short);
13967 vector bool short vec_sll (vector bool short, vector unsigned char);
13968 vector pixel vec_sll (vector pixel, vector unsigned int);
13969 vector pixel vec_sll (vector pixel, vector unsigned short);
13970 vector pixel vec_sll (vector pixel, vector unsigned char);
13971 vector signed char vec_sll (vector signed char, vector unsigned int);
13972 vector signed char vec_sll (vector signed char, vector unsigned short);
13973 vector signed char vec_sll (vector signed char, vector unsigned char);
13974 vector unsigned char vec_sll (vector unsigned char,
13975 vector unsigned int);
13976 vector unsigned char vec_sll (vector unsigned char,
13977 vector unsigned short);
13978 vector unsigned char vec_sll (vector unsigned char,
13979 vector unsigned char);
13980 vector bool char vec_sll (vector bool char, vector unsigned int);
13981 vector bool char vec_sll (vector bool char, vector unsigned short);
13982 vector bool char vec_sll (vector bool char, vector unsigned char);
13984 vector float vec_slo (vector float, vector signed char);
13985 vector float vec_slo (vector float, vector unsigned char);
13986 vector signed int vec_slo (vector signed int, vector signed char);
13987 vector signed int vec_slo (vector signed int, vector unsigned char);
13988 vector unsigned int vec_slo (vector unsigned int, vector signed char);
13989 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
13990 vector signed short vec_slo (vector signed short, vector signed char);
13991 vector signed short vec_slo (vector signed short, vector unsigned char);
13992 vector unsigned short vec_slo (vector unsigned short,
13993 vector signed char);
13994 vector unsigned short vec_slo (vector unsigned short,
13995 vector unsigned char);
13996 vector pixel vec_slo (vector pixel, vector signed char);
13997 vector pixel vec_slo (vector pixel, vector unsigned char);
13998 vector signed char vec_slo (vector signed char, vector signed char);
13999 vector signed char vec_slo (vector signed char, vector unsigned char);
14000 vector unsigned char vec_slo (vector unsigned char, vector signed char);
14001 vector unsigned char vec_slo (vector unsigned char,
14002 vector unsigned char);
14004 vector signed char vec_splat (vector signed char, const int);
14005 vector unsigned char vec_splat (vector unsigned char, const int);
14006 vector bool char vec_splat (vector bool char, const int);
14007 vector signed short vec_splat (vector signed short, const int);
14008 vector unsigned short vec_splat (vector unsigned short, const int);
14009 vector bool short vec_splat (vector bool short, const int);
14010 vector pixel vec_splat (vector pixel, const int);
14011 vector float vec_splat (vector float, const int);
14012 vector signed int vec_splat (vector signed int, const int);
14013 vector unsigned int vec_splat (vector unsigned int, const int);
14014 vector bool int vec_splat (vector bool int, const int);
14016 vector float vec_vspltw (vector float, const int);
14017 vector signed int vec_vspltw (vector signed int, const int);
14018 vector unsigned int vec_vspltw (vector unsigned int, const int);
14019 vector bool int vec_vspltw (vector bool int, const int);
14021 vector bool short vec_vsplth (vector bool short, const int);
14022 vector signed short vec_vsplth (vector signed short, const int);
14023 vector unsigned short vec_vsplth (vector unsigned short, const int);
14024 vector pixel vec_vsplth (vector pixel, const int);
14026 vector signed char vec_vspltb (vector signed char, const int);
14027 vector unsigned char vec_vspltb (vector unsigned char, const int);
14028 vector bool char vec_vspltb (vector bool char, const int);
14030 vector signed char vec_splat_s8 (const int);
14032 vector signed short vec_splat_s16 (const int);
14034 vector signed int vec_splat_s32 (const int);
14036 vector unsigned char vec_splat_u8 (const int);
14038 vector unsigned short vec_splat_u16 (const int);
14040 vector unsigned int vec_splat_u32 (const int);
14042 vector signed char vec_sr (vector signed char, vector unsigned char);
14043 vector unsigned char vec_sr (vector unsigned char,
14044 vector unsigned char);
14045 vector signed short vec_sr (vector signed short,
14046 vector unsigned short);
14047 vector unsigned short vec_sr (vector unsigned short,
14048 vector unsigned short);
14049 vector signed int vec_sr (vector signed int, vector unsigned int);
14050 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
14052 vector signed int vec_vsrw (vector signed int, vector unsigned int);
14053 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
14055 vector signed short vec_vsrh (vector signed short,
14056 vector unsigned short);
14057 vector unsigned short vec_vsrh (vector unsigned short,
14058 vector unsigned short);
14060 vector signed char vec_vsrb (vector signed char, vector unsigned char);
14061 vector unsigned char vec_vsrb (vector unsigned char,
14062 vector unsigned char);
14064 vector signed char vec_sra (vector signed char, vector unsigned char);
14065 vector unsigned char vec_sra (vector unsigned char,
14066 vector unsigned char);
14067 vector signed short vec_sra (vector signed short,
14068 vector unsigned short);
14069 vector unsigned short vec_sra (vector unsigned short,
14070 vector unsigned short);
14071 vector signed int vec_sra (vector signed int, vector unsigned int);
14072 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
14074 vector signed int vec_vsraw (vector signed int, vector unsigned int);
14075 vector unsigned int vec_vsraw (vector unsigned int,
14076 vector unsigned int);
14078 vector signed short vec_vsrah (vector signed short,
14079 vector unsigned short);
14080 vector unsigned short vec_vsrah (vector unsigned short,
14081 vector unsigned short);
14083 vector signed char vec_vsrab (vector signed char, vector unsigned char);
14084 vector unsigned char vec_vsrab (vector unsigned char,
14085 vector unsigned char);
14087 vector signed int vec_srl (vector signed int, vector unsigned int);
14088 vector signed int vec_srl (vector signed int, vector unsigned short);
14089 vector signed int vec_srl (vector signed int, vector unsigned char);
14090 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
14091 vector unsigned int vec_srl (vector unsigned int,
14092 vector unsigned short);
14093 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
14094 vector bool int vec_srl (vector bool int, vector unsigned int);
14095 vector bool int vec_srl (vector bool int, vector unsigned short);
14096 vector bool int vec_srl (vector bool int, vector unsigned char);
14097 vector signed short vec_srl (vector signed short, vector unsigned int);
14098 vector signed short vec_srl (vector signed short,
14099 vector unsigned short);
14100 vector signed short vec_srl (vector signed short, vector unsigned char);
14101 vector unsigned short vec_srl (vector unsigned short,
14102 vector unsigned int);
14103 vector unsigned short vec_srl (vector unsigned short,
14104 vector unsigned short);
14105 vector unsigned short vec_srl (vector unsigned short,
14106 vector unsigned char);
14107 vector bool short vec_srl (vector bool short, vector unsigned int);
14108 vector bool short vec_srl (vector bool short, vector unsigned short);
14109 vector bool short vec_srl (vector bool short, vector unsigned char);
14110 vector pixel vec_srl (vector pixel, vector unsigned int);
14111 vector pixel vec_srl (vector pixel, vector unsigned short);
14112 vector pixel vec_srl (vector pixel, vector unsigned char);
14113 vector signed char vec_srl (vector signed char, vector unsigned int);
14114 vector signed char vec_srl (vector signed char, vector unsigned short);
14115 vector signed char vec_srl (vector signed char, vector unsigned char);
14116 vector unsigned char vec_srl (vector unsigned char,
14117 vector unsigned int);
14118 vector unsigned char vec_srl (vector unsigned char,
14119 vector unsigned short);
14120 vector unsigned char vec_srl (vector unsigned char,
14121 vector unsigned char);
14122 vector bool char vec_srl (vector bool char, vector unsigned int);
14123 vector bool char vec_srl (vector bool char, vector unsigned short);
14124 vector bool char vec_srl (vector bool char, vector unsigned char);
14126 vector float vec_sro (vector float, vector signed char);
14127 vector float vec_sro (vector float, vector unsigned char);
14128 vector signed int vec_sro (vector signed int, vector signed char);
14129 vector signed int vec_sro (vector signed int, vector unsigned char);
14130 vector unsigned int vec_sro (vector unsigned int, vector signed char);
14131 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
14132 vector signed short vec_sro (vector signed short, vector signed char);
14133 vector signed short vec_sro (vector signed short, vector unsigned char);
14134 vector unsigned short vec_sro (vector unsigned short,
14135 vector signed char);
14136 vector unsigned short vec_sro (vector unsigned short,
14137 vector unsigned char);
14138 vector pixel vec_sro (vector pixel, vector signed char);
14139 vector pixel vec_sro (vector pixel, vector unsigned char);
14140 vector signed char vec_sro (vector signed char, vector signed char);
14141 vector signed char vec_sro (vector signed char, vector unsigned char);
14142 vector unsigned char vec_sro (vector unsigned char, vector signed char);
14143 vector unsigned char vec_sro (vector unsigned char,
14144 vector unsigned char);
14146 void vec_st (vector float, int, vector float *);
14147 void vec_st (vector float, int, float *);
14148 void vec_st (vector signed int, int, vector signed int *);
14149 void vec_st (vector signed int, int, int *);
14150 void vec_st (vector unsigned int, int, vector unsigned int *);
14151 void vec_st (vector unsigned int, int, unsigned int *);
14152 void vec_st (vector bool int, int, vector bool int *);
14153 void vec_st (vector bool int, int, unsigned int *);
14154 void vec_st (vector bool int, int, int *);
14155 void vec_st (vector signed short, int, vector signed short *);
14156 void vec_st (vector signed short, int, short *);
14157 void vec_st (vector unsigned short, int, vector unsigned short *);
14158 void vec_st (vector unsigned short, int, unsigned short *);
14159 void vec_st (vector bool short, int, vector bool short *);
14160 void vec_st (vector bool short, int, unsigned short *);
14161 void vec_st (vector pixel, int, vector pixel *);
14162 void vec_st (vector pixel, int, unsigned short *);
14163 void vec_st (vector pixel, int, short *);
14164 void vec_st (vector bool short, int, short *);
14165 void vec_st (vector signed char, int, vector signed char *);
14166 void vec_st (vector signed char, int, signed char *);
14167 void vec_st (vector unsigned char, int, vector unsigned char *);
14168 void vec_st (vector unsigned char, int, unsigned char *);
14169 void vec_st (vector bool char, int, vector bool char *);
14170 void vec_st (vector bool char, int, unsigned char *);
14171 void vec_st (vector bool char, int, signed char *);
14173 void vec_ste (vector signed char, int, signed char *);
14174 void vec_ste (vector unsigned char, int, unsigned char *);
14175 void vec_ste (vector bool char, int, signed char *);
14176 void vec_ste (vector bool char, int, unsigned char *);
14177 void vec_ste (vector signed short, int, short *);
14178 void vec_ste (vector unsigned short, int, unsigned short *);
14179 void vec_ste (vector bool short, int, short *);
14180 void vec_ste (vector bool short, int, unsigned short *);
14181 void vec_ste (vector pixel, int, short *);
14182 void vec_ste (vector pixel, int, unsigned short *);
14183 void vec_ste (vector float, int, float *);
14184 void vec_ste (vector signed int, int, int *);
14185 void vec_ste (vector unsigned int, int, unsigned int *);
14186 void vec_ste (vector bool int, int, int *);
14187 void vec_ste (vector bool int, int, unsigned int *);
14189 void vec_stvewx (vector float, int, float *);
14190 void vec_stvewx (vector signed int, int, int *);
14191 void vec_stvewx (vector unsigned int, int, unsigned int *);
14192 void vec_stvewx (vector bool int, int, int *);
14193 void vec_stvewx (vector bool int, int, unsigned int *);
14195 void vec_stvehx (vector signed short, int, short *);
14196 void vec_stvehx (vector unsigned short, int, unsigned short *);
14197 void vec_stvehx (vector bool short, int, short *);
14198 void vec_stvehx (vector bool short, int, unsigned short *);
14199 void vec_stvehx (vector pixel, int, short *);
14200 void vec_stvehx (vector pixel, int, unsigned short *);
14202 void vec_stvebx (vector signed char, int, signed char *);
14203 void vec_stvebx (vector unsigned char, int, unsigned char *);
14204 void vec_stvebx (vector bool char, int, signed char *);
14205 void vec_stvebx (vector bool char, int, unsigned char *);
14207 void vec_stl (vector float, int, vector float *);
14208 void vec_stl (vector float, int, float *);
14209 void vec_stl (vector signed int, int, vector signed int *);
14210 void vec_stl (vector signed int, int, int *);
14211 void vec_stl (vector unsigned int, int, vector unsigned int *);
14212 void vec_stl (vector unsigned int, int, unsigned int *);
14213 void vec_stl (vector bool int, int, vector bool int *);
14214 void vec_stl (vector bool int, int, unsigned int *);
14215 void vec_stl (vector bool int, int, int *);
14216 void vec_stl (vector signed short, int, vector signed short *);
14217 void vec_stl (vector signed short, int, short *);
14218 void vec_stl (vector unsigned short, int, vector unsigned short *);
14219 void vec_stl (vector unsigned short, int, unsigned short *);
14220 void vec_stl (vector bool short, int, vector bool short *);
14221 void vec_stl (vector bool short, int, unsigned short *);
14222 void vec_stl (vector bool short, int, short *);
14223 void vec_stl (vector pixel, int, vector pixel *);
14224 void vec_stl (vector pixel, int, unsigned short *);
14225 void vec_stl (vector pixel, int, short *);
14226 void vec_stl (vector signed char, int, vector signed char *);
14227 void vec_stl (vector signed char, int, signed char *);
14228 void vec_stl (vector unsigned char, int, vector unsigned char *);
14229 void vec_stl (vector unsigned char, int, unsigned char *);
14230 void vec_stl (vector bool char, int, vector bool char *);
14231 void vec_stl (vector bool char, int, unsigned char *);
14232 void vec_stl (vector bool char, int, signed char *);
14234 vector signed char vec_sub (vector bool char, vector signed char);
14235 vector signed char vec_sub (vector signed char, vector bool char);
14236 vector signed char vec_sub (vector signed char, vector signed char);
14237 vector unsigned char vec_sub (vector bool char, vector unsigned char);
14238 vector unsigned char vec_sub (vector unsigned char, vector bool char);
14239 vector unsigned char vec_sub (vector unsigned char,
14240 vector unsigned char);
14241 vector signed short vec_sub (vector bool short, vector signed short);
14242 vector signed short vec_sub (vector signed short, vector bool short);
14243 vector signed short vec_sub (vector signed short, vector signed short);
14244 vector unsigned short vec_sub (vector bool short,
14245 vector unsigned short);
14246 vector unsigned short vec_sub (vector unsigned short,
14247 vector bool short);
14248 vector unsigned short vec_sub (vector unsigned short,
14249 vector unsigned short);
14250 vector signed int vec_sub (vector bool int, vector signed int);
14251 vector signed int vec_sub (vector signed int, vector bool int);
14252 vector signed int vec_sub (vector signed int, vector signed int);
14253 vector unsigned int vec_sub (vector bool int, vector unsigned int);
14254 vector unsigned int vec_sub (vector unsigned int, vector bool int);
14255 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
14256 vector float vec_sub (vector float, vector float);
14258 vector float vec_vsubfp (vector float, vector float);
14260 vector signed int vec_vsubuwm (vector bool int, vector signed int);
14261 vector signed int vec_vsubuwm (vector signed int, vector bool int);
14262 vector signed int vec_vsubuwm (vector signed int, vector signed int);
14263 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
14264 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
14265 vector unsigned int vec_vsubuwm (vector unsigned int,
14266 vector unsigned int);
14268 vector signed short vec_vsubuhm (vector bool short,
14269 vector signed short);
14270 vector signed short vec_vsubuhm (vector signed short,
14271 vector bool short);
14272 vector signed short vec_vsubuhm (vector signed short,
14273 vector signed short);
14274 vector unsigned short vec_vsubuhm (vector bool short,
14275 vector unsigned short);
14276 vector unsigned short vec_vsubuhm (vector unsigned short,
14277 vector bool short);
14278 vector unsigned short vec_vsubuhm (vector unsigned short,
14279 vector unsigned short);
14281 vector signed char vec_vsububm (vector bool char, vector signed char);
14282 vector signed char vec_vsububm (vector signed char, vector bool char);
14283 vector signed char vec_vsububm (vector signed char, vector signed char);
14284 vector unsigned char vec_vsububm (vector bool char,
14285 vector unsigned char);
14286 vector unsigned char vec_vsububm (vector unsigned char,
14288 vector unsigned char vec_vsububm (vector unsigned char,
14289 vector unsigned char);
14291 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
14293 vector unsigned char vec_subs (vector bool char, vector unsigned char);
14294 vector unsigned char vec_subs (vector unsigned char, vector bool char);
14295 vector unsigned char vec_subs (vector unsigned char,
14296 vector unsigned char);
14297 vector signed char vec_subs (vector bool char, vector signed char);
14298 vector signed char vec_subs (vector signed char, vector bool char);
14299 vector signed char vec_subs (vector signed char, vector signed char);
14300 vector unsigned short vec_subs (vector bool short,
14301 vector unsigned short);
14302 vector unsigned short vec_subs (vector unsigned short,
14303 vector bool short);
14304 vector unsigned short vec_subs (vector unsigned short,
14305 vector unsigned short);
14306 vector signed short vec_subs (vector bool short, vector signed short);
14307 vector signed short vec_subs (vector signed short, vector bool short);
14308 vector signed short vec_subs (vector signed short, vector signed short);
14309 vector unsigned int vec_subs (vector bool int, vector unsigned int);
14310 vector unsigned int vec_subs (vector unsigned int, vector bool int);
14311 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
14312 vector signed int vec_subs (vector bool int, vector signed int);
14313 vector signed int vec_subs (vector signed int, vector bool int);
14314 vector signed int vec_subs (vector signed int, vector signed int);
14316 vector signed int vec_vsubsws (vector bool int, vector signed int);
14317 vector signed int vec_vsubsws (vector signed int, vector bool int);
14318 vector signed int vec_vsubsws (vector signed int, vector signed int);
14320 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
14321 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
14322 vector unsigned int vec_vsubuws (vector unsigned int,
14323 vector unsigned int);
14325 vector signed short vec_vsubshs (vector bool short,
14326 vector signed short);
14327 vector signed short vec_vsubshs (vector signed short,
14328 vector bool short);
14329 vector signed short vec_vsubshs (vector signed short,
14330 vector signed short);
14332 vector unsigned short vec_vsubuhs (vector bool short,
14333 vector unsigned short);
14334 vector unsigned short vec_vsubuhs (vector unsigned short,
14335 vector bool short);
14336 vector unsigned short vec_vsubuhs (vector unsigned short,
14337 vector unsigned short);
14339 vector signed char vec_vsubsbs (vector bool char, vector signed char);
14340 vector signed char vec_vsubsbs (vector signed char, vector bool char);
14341 vector signed char vec_vsubsbs (vector signed char, vector signed char);
14343 vector unsigned char vec_vsububs (vector bool char,
14344 vector unsigned char);
14345 vector unsigned char vec_vsububs (vector unsigned char,
14347 vector unsigned char vec_vsububs (vector unsigned char,
14348 vector unsigned char);
14350 vector unsigned int vec_sum4s (vector unsigned char,
14351 vector unsigned int);
14352 vector signed int vec_sum4s (vector signed char, vector signed int);
14353 vector signed int vec_sum4s (vector signed short, vector signed int);
14355 vector signed int vec_vsum4shs (vector signed short, vector signed int);
14357 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
14359 vector unsigned int vec_vsum4ubs (vector unsigned char,
14360 vector unsigned int);
14362 vector signed int vec_sum2s (vector signed int, vector signed int);
14364 vector signed int vec_sums (vector signed int, vector signed int);
14366 vector float vec_trunc (vector float);
14368 vector signed short vec_unpackh (vector signed char);
14369 vector bool short vec_unpackh (vector bool char);
14370 vector signed int vec_unpackh (vector signed short);
14371 vector bool int vec_unpackh (vector bool short);
14372 vector unsigned int vec_unpackh (vector pixel);
14374 vector bool int vec_vupkhsh (vector bool short);
14375 vector signed int vec_vupkhsh (vector signed short);
14377 vector unsigned int vec_vupkhpx (vector pixel);
14379 vector bool short vec_vupkhsb (vector bool char);
14380 vector signed short vec_vupkhsb (vector signed char);
14382 vector signed short vec_unpackl (vector signed char);
14383 vector bool short vec_unpackl (vector bool char);
14384 vector unsigned int vec_unpackl (vector pixel);
14385 vector signed int vec_unpackl (vector signed short);
14386 vector bool int vec_unpackl (vector bool short);
14388 vector unsigned int vec_vupklpx (vector pixel);
14390 vector bool int vec_vupklsh (vector bool short);
14391 vector signed int vec_vupklsh (vector signed short);
14393 vector bool short vec_vupklsb (vector bool char);
14394 vector signed short vec_vupklsb (vector signed char);
14396 vector float vec_xor (vector float, vector float);
14397 vector float vec_xor (vector float, vector bool int);
14398 vector float vec_xor (vector bool int, vector float);
14399 vector bool int vec_xor (vector bool int, vector bool int);
14400 vector signed int vec_xor (vector bool int, vector signed int);
14401 vector signed int vec_xor (vector signed int, vector bool int);
14402 vector signed int vec_xor (vector signed int, vector signed int);
14403 vector unsigned int vec_xor (vector bool int, vector unsigned int);
14404 vector unsigned int vec_xor (vector unsigned int, vector bool int);
14405 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
14406 vector bool short vec_xor (vector bool short, vector bool short);
14407 vector signed short vec_xor (vector bool short, vector signed short);
14408 vector signed short vec_xor (vector signed short, vector bool short);
14409 vector signed short vec_xor (vector signed short, vector signed short);
14410 vector unsigned short vec_xor (vector bool short,
14411 vector unsigned short);
14412 vector unsigned short vec_xor (vector unsigned short,
14413 vector bool short);
14414 vector unsigned short vec_xor (vector unsigned short,
14415 vector unsigned short);
14416 vector signed char vec_xor (vector bool char, vector signed char);
14417 vector bool char vec_xor (vector bool char, vector bool char);
14418 vector signed char vec_xor (vector signed char, vector bool char);
14419 vector signed char vec_xor (vector signed char, vector signed char);
14420 vector unsigned char vec_xor (vector bool char, vector unsigned char);
14421 vector unsigned char vec_xor (vector unsigned char, vector bool char);
14422 vector unsigned char vec_xor (vector unsigned char,
14423 vector unsigned char);
14425 int vec_all_eq (vector signed char, vector bool char);
14426 int vec_all_eq (vector signed char, vector signed char);
14427 int vec_all_eq (vector unsigned char, vector bool char);
14428 int vec_all_eq (vector unsigned char, vector unsigned char);
14429 int vec_all_eq (vector bool char, vector bool char);
14430 int vec_all_eq (vector bool char, vector unsigned char);
14431 int vec_all_eq (vector bool char, vector signed char);
14432 int vec_all_eq (vector signed short, vector bool short);
14433 int vec_all_eq (vector signed short, vector signed short);
14434 int vec_all_eq (vector unsigned short, vector bool short);
14435 int vec_all_eq (vector unsigned short, vector unsigned short);
14436 int vec_all_eq (vector bool short, vector bool short);
14437 int vec_all_eq (vector bool short, vector unsigned short);
14438 int vec_all_eq (vector bool short, vector signed short);
14439 int vec_all_eq (vector pixel, vector pixel);
14440 int vec_all_eq (vector signed int, vector bool int);
14441 int vec_all_eq (vector signed int, vector signed int);
14442 int vec_all_eq (vector unsigned int, vector bool int);
14443 int vec_all_eq (vector unsigned int, vector unsigned int);
14444 int vec_all_eq (vector bool int, vector bool int);
14445 int vec_all_eq (vector bool int, vector unsigned int);
14446 int vec_all_eq (vector bool int, vector signed int);
14447 int vec_all_eq (vector float, vector float);
14449 int vec_all_ge (vector bool char, vector unsigned char);
14450 int vec_all_ge (vector unsigned char, vector bool char);
14451 int vec_all_ge (vector unsigned char, vector unsigned char);
14452 int vec_all_ge (vector bool char, vector signed char);
14453 int vec_all_ge (vector signed char, vector bool char);
14454 int vec_all_ge (vector signed char, vector signed char);
14455 int vec_all_ge (vector bool short, vector unsigned short);
14456 int vec_all_ge (vector unsigned short, vector bool short);
14457 int vec_all_ge (vector unsigned short, vector unsigned short);
14458 int vec_all_ge (vector signed short, vector signed short);
14459 int vec_all_ge (vector bool short, vector signed short);
14460 int vec_all_ge (vector signed short, vector bool short);
14461 int vec_all_ge (vector bool int, vector unsigned int);
14462 int vec_all_ge (vector unsigned int, vector bool int);
14463 int vec_all_ge (vector unsigned int, vector unsigned int);
14464 int vec_all_ge (vector bool int, vector signed int);
14465 int vec_all_ge (vector signed int, vector bool int);
14466 int vec_all_ge (vector signed int, vector signed int);
14467 int vec_all_ge (vector float, vector float);
14469 int vec_all_gt (vector bool char, vector unsigned char);
14470 int vec_all_gt (vector unsigned char, vector bool char);
14471 int vec_all_gt (vector unsigned char, vector unsigned char);
14472 int vec_all_gt (vector bool char, vector signed char);
14473 int vec_all_gt (vector signed char, vector bool char);
14474 int vec_all_gt (vector signed char, vector signed char);
14475 int vec_all_gt (vector bool short, vector unsigned short);
14476 int vec_all_gt (vector unsigned short, vector bool short);
14477 int vec_all_gt (vector unsigned short, vector unsigned short);
14478 int vec_all_gt (vector bool short, vector signed short);
14479 int vec_all_gt (vector signed short, vector bool short);
14480 int vec_all_gt (vector signed short, vector signed short);
14481 int vec_all_gt (vector bool int, vector unsigned int);
14482 int vec_all_gt (vector unsigned int, vector bool int);
14483 int vec_all_gt (vector unsigned int, vector unsigned int);
14484 int vec_all_gt (vector bool int, vector signed int);
14485 int vec_all_gt (vector signed int, vector bool int);
14486 int vec_all_gt (vector signed int, vector signed int);
14487 int vec_all_gt (vector float, vector float);
14489 int vec_all_in (vector float, vector float);
14491 int vec_all_le (vector bool char, vector unsigned char);
14492 int vec_all_le (vector unsigned char, vector bool char);
14493 int vec_all_le (vector unsigned char, vector unsigned char);
14494 int vec_all_le (vector bool char, vector signed char);
14495 int vec_all_le (vector signed char, vector bool char);
14496 int vec_all_le (vector signed char, vector signed char);
14497 int vec_all_le (vector bool short, vector unsigned short);
14498 int vec_all_le (vector unsigned short, vector bool short);
14499 int vec_all_le (vector unsigned short, vector unsigned short);
14500 int vec_all_le (vector bool short, vector signed short);
14501 int vec_all_le (vector signed short, vector bool short);
14502 int vec_all_le (vector signed short, vector signed short);
14503 int vec_all_le (vector bool int, vector unsigned int);
14504 int vec_all_le (vector unsigned int, vector bool int);
14505 int vec_all_le (vector unsigned int, vector unsigned int);
14506 int vec_all_le (vector bool int, vector signed int);
14507 int vec_all_le (vector signed int, vector bool int);
14508 int vec_all_le (vector signed int, vector signed int);
14509 int vec_all_le (vector float, vector float);
14511 int vec_all_lt (vector bool char, vector unsigned char);
14512 int vec_all_lt (vector unsigned char, vector bool char);
14513 int vec_all_lt (vector unsigned char, vector unsigned char);
14514 int vec_all_lt (vector bool char, vector signed char);
14515 int vec_all_lt (vector signed char, vector bool char);
14516 int vec_all_lt (vector signed char, vector signed char);
14517 int vec_all_lt (vector bool short, vector unsigned short);
14518 int vec_all_lt (vector unsigned short, vector bool short);
14519 int vec_all_lt (vector unsigned short, vector unsigned short);
14520 int vec_all_lt (vector bool short, vector signed short);
14521 int vec_all_lt (vector signed short, vector bool short);
14522 int vec_all_lt (vector signed short, vector signed short);
14523 int vec_all_lt (vector bool int, vector unsigned int);
14524 int vec_all_lt (vector unsigned int, vector bool int);
14525 int vec_all_lt (vector unsigned int, vector unsigned int);
14526 int vec_all_lt (vector bool int, vector signed int);
14527 int vec_all_lt (vector signed int, vector bool int);
14528 int vec_all_lt (vector signed int, vector signed int);
14529 int vec_all_lt (vector float, vector float);
14531 int vec_all_nan (vector float);
14533 int vec_all_ne (vector signed char, vector bool char);
14534 int vec_all_ne (vector signed char, vector signed char);
14535 int vec_all_ne (vector unsigned char, vector bool char);
14536 int vec_all_ne (vector unsigned char, vector unsigned char);
14537 int vec_all_ne (vector bool char, vector bool char);
14538 int vec_all_ne (vector bool char, vector unsigned char);
14539 int vec_all_ne (vector bool char, vector signed char);
14540 int vec_all_ne (vector signed short, vector bool short);
14541 int vec_all_ne (vector signed short, vector signed short);
14542 int vec_all_ne (vector unsigned short, vector bool short);
14543 int vec_all_ne (vector unsigned short, vector unsigned short);
14544 int vec_all_ne (vector bool short, vector bool short);
14545 int vec_all_ne (vector bool short, vector unsigned short);
14546 int vec_all_ne (vector bool short, vector signed short);
14547 int vec_all_ne (vector pixel, vector pixel);
14548 int vec_all_ne (vector signed int, vector bool int);
14549 int vec_all_ne (vector signed int, vector signed int);
14550 int vec_all_ne (vector unsigned int, vector bool int);
14551 int vec_all_ne (vector unsigned int, vector unsigned int);
14552 int vec_all_ne (vector bool int, vector bool int);
14553 int vec_all_ne (vector bool int, vector unsigned int);
14554 int vec_all_ne (vector bool int, vector signed int);
14555 int vec_all_ne (vector float, vector float);
14557 int vec_all_nge (vector float, vector float);
14559 int vec_all_ngt (vector float, vector float);
14561 int vec_all_nle (vector float, vector float);
14563 int vec_all_nlt (vector float, vector float);
14565 int vec_all_numeric (vector float);
14567 int vec_any_eq (vector signed char, vector bool char);
14568 int vec_any_eq (vector signed char, vector signed char);
14569 int vec_any_eq (vector unsigned char, vector bool char);
14570 int vec_any_eq (vector unsigned char, vector unsigned char);
14571 int vec_any_eq (vector bool char, vector bool char);
14572 int vec_any_eq (vector bool char, vector unsigned char);
14573 int vec_any_eq (vector bool char, vector signed char);
14574 int vec_any_eq (vector signed short, vector bool short);
14575 int vec_any_eq (vector signed short, vector signed short);
14576 int vec_any_eq (vector unsigned short, vector bool short);
14577 int vec_any_eq (vector unsigned short, vector unsigned short);
14578 int vec_any_eq (vector bool short, vector bool short);
14579 int vec_any_eq (vector bool short, vector unsigned short);
14580 int vec_any_eq (vector bool short, vector signed short);
14581 int vec_any_eq (vector pixel, vector pixel);
14582 int vec_any_eq (vector signed int, vector bool int);
14583 int vec_any_eq (vector signed int, vector signed int);
14584 int vec_any_eq (vector unsigned int, vector bool int);
14585 int vec_any_eq (vector unsigned int, vector unsigned int);
14586 int vec_any_eq (vector bool int, vector bool int);
14587 int vec_any_eq (vector bool int, vector unsigned int);
14588 int vec_any_eq (vector bool int, vector signed int);
14589 int vec_any_eq (vector float, vector float);
14591 int vec_any_ge (vector signed char, vector bool char);
14592 int vec_any_ge (vector unsigned char, vector bool char);
14593 int vec_any_ge (vector unsigned char, vector unsigned char);
14594 int vec_any_ge (vector signed char, vector signed char);
14595 int vec_any_ge (vector bool char, vector unsigned char);
14596 int vec_any_ge (vector bool char, vector signed char);
14597 int vec_any_ge (vector unsigned short, vector bool short);
14598 int vec_any_ge (vector unsigned short, vector unsigned short);
14599 int vec_any_ge (vector signed short, vector signed short);
14600 int vec_any_ge (vector signed short, vector bool short);
14601 int vec_any_ge (vector bool short, vector unsigned short);
14602 int vec_any_ge (vector bool short, vector signed short);
14603 int vec_any_ge (vector signed int, vector bool int);
14604 int vec_any_ge (vector unsigned int, vector bool int);
14605 int vec_any_ge (vector unsigned int, vector unsigned int);
14606 int vec_any_ge (vector signed int, vector signed int);
14607 int vec_any_ge (vector bool int, vector unsigned int);
14608 int vec_any_ge (vector bool int, vector signed int);
14609 int vec_any_ge (vector float, vector float);
14611 int vec_any_gt (vector bool char, vector unsigned char);
14612 int vec_any_gt (vector unsigned char, vector bool char);
14613 int vec_any_gt (vector unsigned char, vector unsigned char);
14614 int vec_any_gt (vector bool char, vector signed char);
14615 int vec_any_gt (vector signed char, vector bool char);
14616 int vec_any_gt (vector signed char, vector signed char);
14617 int vec_any_gt (vector bool short, vector unsigned short);
14618 int vec_any_gt (vector unsigned short, vector bool short);
14619 int vec_any_gt (vector unsigned short, vector unsigned short);
14620 int vec_any_gt (vector bool short, vector signed short);
14621 int vec_any_gt (vector signed short, vector bool short);
14622 int vec_any_gt (vector signed short, vector signed short);
14623 int vec_any_gt (vector bool int, vector unsigned int);
14624 int vec_any_gt (vector unsigned int, vector bool int);
14625 int vec_any_gt (vector unsigned int, vector unsigned int);
14626 int vec_any_gt (vector bool int, vector signed int);
14627 int vec_any_gt (vector signed int, vector bool int);
14628 int vec_any_gt (vector signed int, vector signed int);
14629 int vec_any_gt (vector float, vector float);
14631 int vec_any_le (vector bool char, vector unsigned char);
14632 int vec_any_le (vector unsigned char, vector bool char);
14633 int vec_any_le (vector unsigned char, vector unsigned char);
14634 int vec_any_le (vector bool char, vector signed char);
14635 int vec_any_le (vector signed char, vector bool char);
14636 int vec_any_le (vector signed char, vector signed char);
14637 int vec_any_le (vector bool short, vector unsigned short);
14638 int vec_any_le (vector unsigned short, vector bool short);
14639 int vec_any_le (vector unsigned short, vector unsigned short);
14640 int vec_any_le (vector bool short, vector signed short);
14641 int vec_any_le (vector signed short, vector bool short);
14642 int vec_any_le (vector signed short, vector signed short);
14643 int vec_any_le (vector bool int, vector unsigned int);
14644 int vec_any_le (vector unsigned int, vector bool int);
14645 int vec_any_le (vector unsigned int, vector unsigned int);
14646 int vec_any_le (vector bool int, vector signed int);
14647 int vec_any_le (vector signed int, vector bool int);
14648 int vec_any_le (vector signed int, vector signed int);
14649 int vec_any_le (vector float, vector float);
14651 int vec_any_lt (vector bool char, vector unsigned char);
14652 int vec_any_lt (vector unsigned char, vector bool char);
14653 int vec_any_lt (vector unsigned char, vector unsigned char);
14654 int vec_any_lt (vector bool char, vector signed char);
14655 int vec_any_lt (vector signed char, vector bool char);
14656 int vec_any_lt (vector signed char, vector signed char);
14657 int vec_any_lt (vector bool short, vector unsigned short);
14658 int vec_any_lt (vector unsigned short, vector bool short);
14659 int vec_any_lt (vector unsigned short, vector unsigned short);
14660 int vec_any_lt (vector bool short, vector signed short);
14661 int vec_any_lt (vector signed short, vector bool short);
14662 int vec_any_lt (vector signed short, vector signed short);
14663 int vec_any_lt (vector bool int, vector unsigned int);
14664 int vec_any_lt (vector unsigned int, vector bool int);
14665 int vec_any_lt (vector unsigned int, vector unsigned int);
14666 int vec_any_lt (vector bool int, vector signed int);
14667 int vec_any_lt (vector signed int, vector bool int);
14668 int vec_any_lt (vector signed int, vector signed int);
14669 int vec_any_lt (vector float, vector float);
14671 int vec_any_nan (vector float);
14673 int vec_any_ne (vector signed char, vector bool char);
14674 int vec_any_ne (vector signed char, vector signed char);
14675 int vec_any_ne (vector unsigned char, vector bool char);
14676 int vec_any_ne (vector unsigned char, vector unsigned char);
14677 int vec_any_ne (vector bool char, vector bool char);
14678 int vec_any_ne (vector bool char, vector unsigned char);
14679 int vec_any_ne (vector bool char, vector signed char);
14680 int vec_any_ne (vector signed short, vector bool short);
14681 int vec_any_ne (vector signed short, vector signed short);
14682 int vec_any_ne (vector unsigned short, vector bool short);
14683 int vec_any_ne (vector unsigned short, vector unsigned short);
14684 int vec_any_ne (vector bool short, vector bool short);
14685 int vec_any_ne (vector bool short, vector unsigned short);
14686 int vec_any_ne (vector bool short, vector signed short);
14687 int vec_any_ne (vector pixel, vector pixel);
14688 int vec_any_ne (vector signed int, vector bool int);
14689 int vec_any_ne (vector signed int, vector signed int);
14690 int vec_any_ne (vector unsigned int, vector bool int);
14691 int vec_any_ne (vector unsigned int, vector unsigned int);
14692 int vec_any_ne (vector bool int, vector bool int);
14693 int vec_any_ne (vector bool int, vector unsigned int);
14694 int vec_any_ne (vector bool int, vector signed int);
14695 int vec_any_ne (vector float, vector float);
14697 int vec_any_nge (vector float, vector float);
14699 int vec_any_ngt (vector float, vector float);
14701 int vec_any_nle (vector float, vector float);
14703 int vec_any_nlt (vector float, vector float);
14705 int vec_any_numeric (vector float);
14707 int vec_any_out (vector float, vector float);
14710 If the vector/scalar (VSX) instruction set is available, the following
14711 additional functions are available:
14714 vector double vec_abs (vector double);
14715 vector double vec_add (vector double, vector double);
14716 vector double vec_and (vector double, vector double);
14717 vector double vec_and (vector double, vector bool long);
14718 vector double vec_and (vector bool long, vector double);
14719 vector double vec_andc (vector double, vector double);
14720 vector double vec_andc (vector double, vector bool long);
14721 vector double vec_andc (vector bool long, vector double);
14722 vector double vec_ceil (vector double);
14723 vector bool long vec_cmpeq (vector double, vector double);
14724 vector bool long vec_cmpge (vector double, vector double);
14725 vector bool long vec_cmpgt (vector double, vector double);
14726 vector bool long vec_cmple (vector double, vector double);
14727 vector bool long vec_cmplt (vector double, vector double);
14728 vector float vec_div (vector float, vector float);
14729 vector double vec_div (vector double, vector double);
14730 vector double vec_floor (vector double);
14731 vector double vec_ld (int, const vector double *);
14732 vector double vec_ld (int, const double *);
14733 vector double vec_ldl (int, const vector double *);
14734 vector double vec_ldl (int, const double *);
14735 vector unsigned char vec_lvsl (int, const volatile double *);
14736 vector unsigned char vec_lvsr (int, const volatile double *);
14737 vector double vec_madd (vector double, vector double, vector double);
14738 vector double vec_max (vector double, vector double);
14739 vector double vec_min (vector double, vector double);
14740 vector float vec_msub (vector float, vector float, vector float);
14741 vector double vec_msub (vector double, vector double, vector double);
14742 vector float vec_mul (vector float, vector float);
14743 vector double vec_mul (vector double, vector double);
14744 vector float vec_nearbyint (vector float);
14745 vector double vec_nearbyint (vector double);
14746 vector float vec_nmadd (vector float, vector float, vector float);
14747 vector double vec_nmadd (vector double, vector double, vector double);
14748 vector double vec_nmsub (vector double, vector double, vector double);
14749 vector double vec_nor (vector double, vector double);
14750 vector double vec_or (vector double, vector double);
14751 vector double vec_or (vector double, vector bool long);
14752 vector double vec_or (vector bool long, vector double);
14753 vector double vec_perm (vector double,
14755 vector unsigned char);
14756 vector double vec_rint (vector double);
14757 vector double vec_recip (vector double, vector double);
14758 vector double vec_rsqrt (vector double);
14759 vector double vec_rsqrte (vector double);
14760 vector double vec_sel (vector double, vector double, vector bool long);
14761 vector double vec_sel (vector double, vector double, vector unsigned long);
14762 vector double vec_sub (vector double, vector double);
14763 vector float vec_sqrt (vector float);
14764 vector double vec_sqrt (vector double);
14765 void vec_st (vector double, int, vector double *);
14766 void vec_st (vector double, int, double *);
14767 vector double vec_trunc (vector double);
14768 vector double vec_xor (vector double, vector double);
14769 vector double vec_xor (vector double, vector bool long);
14770 vector double vec_xor (vector bool long, vector double);
14771 int vec_all_eq (vector double, vector double);
14772 int vec_all_ge (vector double, vector double);
14773 int vec_all_gt (vector double, vector double);
14774 int vec_all_le (vector double, vector double);
14775 int vec_all_lt (vector double, vector double);
14776 int vec_all_nan (vector double);
14777 int vec_all_ne (vector double, vector double);
14778 int vec_all_nge (vector double, vector double);
14779 int vec_all_ngt (vector double, vector double);
14780 int vec_all_nle (vector double, vector double);
14781 int vec_all_nlt (vector double, vector double);
14782 int vec_all_numeric (vector double);
14783 int vec_any_eq (vector double, vector double);
14784 int vec_any_ge (vector double, vector double);
14785 int vec_any_gt (vector double, vector double);
14786 int vec_any_le (vector double, vector double);
14787 int vec_any_lt (vector double, vector double);
14788 int vec_any_nan (vector double);
14789 int vec_any_ne (vector double, vector double);
14790 int vec_any_nge (vector double, vector double);
14791 int vec_any_ngt (vector double, vector double);
14792 int vec_any_nle (vector double, vector double);
14793 int vec_any_nlt (vector double, vector double);
14794 int vec_any_numeric (vector double);
14796 vector double vec_vsx_ld (int, const vector double *);
14797 vector double vec_vsx_ld (int, const double *);
14798 vector float vec_vsx_ld (int, const vector float *);
14799 vector float vec_vsx_ld (int, const float *);
14800 vector bool int vec_vsx_ld (int, const vector bool int *);
14801 vector signed int vec_vsx_ld (int, const vector signed int *);
14802 vector signed int vec_vsx_ld (int, const int *);
14803 vector signed int vec_vsx_ld (int, const long *);
14804 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
14805 vector unsigned int vec_vsx_ld (int, const unsigned int *);
14806 vector unsigned int vec_vsx_ld (int, const unsigned long *);
14807 vector bool short vec_vsx_ld (int, const vector bool short *);
14808 vector pixel vec_vsx_ld (int, const vector pixel *);
14809 vector signed short vec_vsx_ld (int, const vector signed short *);
14810 vector signed short vec_vsx_ld (int, const short *);
14811 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
14812 vector unsigned short vec_vsx_ld (int, const unsigned short *);
14813 vector bool char vec_vsx_ld (int, const vector bool char *);
14814 vector signed char vec_vsx_ld (int, const vector signed char *);
14815 vector signed char vec_vsx_ld (int, const signed char *);
14816 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
14817 vector unsigned char vec_vsx_ld (int, const unsigned char *);
14819 void vec_vsx_st (vector double, int, vector double *);
14820 void vec_vsx_st (vector double, int, double *);
14821 void vec_vsx_st (vector float, int, vector float *);
14822 void vec_vsx_st (vector float, int, float *);
14823 void vec_vsx_st (vector signed int, int, vector signed int *);
14824 void vec_vsx_st (vector signed int, int, int *);
14825 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
14826 void vec_vsx_st (vector unsigned int, int, unsigned int *);
14827 void vec_vsx_st (vector bool int, int, vector bool int *);
14828 void vec_vsx_st (vector bool int, int, unsigned int *);
14829 void vec_vsx_st (vector bool int, int, int *);
14830 void vec_vsx_st (vector signed short, int, vector signed short *);
14831 void vec_vsx_st (vector signed short, int, short *);
14832 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
14833 void vec_vsx_st (vector unsigned short, int, unsigned short *);
14834 void vec_vsx_st (vector bool short, int, vector bool short *);
14835 void vec_vsx_st (vector bool short, int, unsigned short *);
14836 void vec_vsx_st (vector pixel, int, vector pixel *);
14837 void vec_vsx_st (vector pixel, int, unsigned short *);
14838 void vec_vsx_st (vector pixel, int, short *);
14839 void vec_vsx_st (vector bool short, int, short *);
14840 void vec_vsx_st (vector signed char, int, vector signed char *);
14841 void vec_vsx_st (vector signed char, int, signed char *);
14842 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
14843 void vec_vsx_st (vector unsigned char, int, unsigned char *);
14844 void vec_vsx_st (vector bool char, int, vector bool char *);
14845 void vec_vsx_st (vector bool char, int, unsigned char *);
14846 void vec_vsx_st (vector bool char, int, signed char *);
14849 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
14850 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
14851 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
14852 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
14853 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
14855 If the ISA 2.07 additions to the vector/scalar (power8-vector)
14856 instruction set is available, the following additional functions are
14857 available for both 32-bit and 64-bit targets. For 64-bit targets, you
14858 can use @var{vector long} instead of @var{vector long long},
14859 @var{vector bool long} instead of @var{vector bool long long}, and
14860 @var{vector unsigned long} instead of @var{vector unsigned long long}.
14863 vector long long vec_abs (vector long long);
14865 vector long long vec_add (vector long long, vector long long);
14866 vector unsigned long long vec_add (vector unsigned long long,
14867 vector unsigned long long);
14869 int vec_all_eq (vector long long, vector long long);
14870 int vec_all_ge (vector long long, vector long long);
14871 int vec_all_gt (vector long long, vector long long);
14872 int vec_all_le (vector long long, vector long long);
14873 int vec_all_lt (vector long long, vector long long);
14874 int vec_all_ne (vector long long, vector long long);
14875 int vec_any_eq (vector long long, vector long long);
14876 int vec_any_ge (vector long long, vector long long);
14877 int vec_any_gt (vector long long, vector long long);
14878 int vec_any_le (vector long long, vector long long);
14879 int vec_any_lt (vector long long, vector long long);
14880 int vec_any_ne (vector long long, vector long long);
14882 vector long long vec_eqv (vector long long, vector long long);
14883 vector long long vec_eqv (vector bool long long, vector long long);
14884 vector long long vec_eqv (vector long long, vector bool long long);
14885 vector unsigned long long vec_eqv (vector unsigned long long,
14886 vector unsigned long long);
14887 vector unsigned long long vec_eqv (vector bool long long,
14888 vector unsigned long long);
14889 vector unsigned long long vec_eqv (vector unsigned long long,
14890 vector bool long long);
14891 vector int vec_eqv (vector int, vector int);
14892 vector int vec_eqv (vector bool int, vector int);
14893 vector int vec_eqv (vector int, vector bool int);
14894 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
14895 vector unsigned int vec_eqv (vector bool unsigned int,
14896 vector unsigned int);
14897 vector unsigned int vec_eqv (vector unsigned int,
14898 vector bool unsigned int);
14899 vector short vec_eqv (vector short, vector short);
14900 vector short vec_eqv (vector bool short, vector short);
14901 vector short vec_eqv (vector short, vector bool short);
14902 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
14903 vector unsigned short vec_eqv (vector bool unsigned short,
14904 vector unsigned short);
14905 vector unsigned short vec_eqv (vector unsigned short,
14906 vector bool unsigned short);
14907 vector signed char vec_eqv (vector signed char, vector signed char);
14908 vector signed char vec_eqv (vector bool signed char, vector signed char);
14909 vector signed char vec_eqv (vector signed char, vector bool signed char);
14910 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
14911 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
14912 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
14914 vector long long vec_max (vector long long, vector long long);
14915 vector unsigned long long vec_max (vector unsigned long long,
14916 vector unsigned long long);
14918 vector long long vec_min (vector long long, vector long long);
14919 vector unsigned long long vec_min (vector unsigned long long,
14920 vector unsigned long long);
14922 vector long long vec_nand (vector long long, vector long long);
14923 vector long long vec_nand (vector bool long long, vector long long);
14924 vector long long vec_nand (vector long long, vector bool long long);
14925 vector unsigned long long vec_nand (vector unsigned long long,
14926 vector unsigned long long);
14927 vector unsigned long long vec_nand (vector bool long long,
14928 vector unsigned long long);
14929 vector unsigned long long vec_nand (vector unsigned long long,
14930 vector bool long long);
14931 vector int vec_nand (vector int, vector int);
14932 vector int vec_nand (vector bool int, vector int);
14933 vector int vec_nand (vector int, vector bool int);
14934 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
14935 vector unsigned int vec_nand (vector bool unsigned int,
14936 vector unsigned int);
14937 vector unsigned int vec_nand (vector unsigned int,
14938 vector bool unsigned int);
14939 vector short vec_nand (vector short, vector short);
14940 vector short vec_nand (vector bool short, vector short);
14941 vector short vec_nand (vector short, vector bool short);
14942 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
14943 vector unsigned short vec_nand (vector bool unsigned short,
14944 vector unsigned short);
14945 vector unsigned short vec_nand (vector unsigned short,
14946 vector bool unsigned short);
14947 vector signed char vec_nand (vector signed char, vector signed char);
14948 vector signed char vec_nand (vector bool signed char, vector signed char);
14949 vector signed char vec_nand (vector signed char, vector bool signed char);
14950 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
14951 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
14952 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
14954 vector long long vec_orc (vector long long, vector long long);
14955 vector long long vec_orc (vector bool long long, vector long long);
14956 vector long long vec_orc (vector long long, vector bool long long);
14957 vector unsigned long long vec_orc (vector unsigned long long,
14958 vector unsigned long long);
14959 vector unsigned long long vec_orc (vector bool long long,
14960 vector unsigned long long);
14961 vector unsigned long long vec_orc (vector unsigned long long,
14962 vector bool long long);
14963 vector int vec_orc (vector int, vector int);
14964 vector int vec_orc (vector bool int, vector int);
14965 vector int vec_orc (vector int, vector bool int);
14966 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
14967 vector unsigned int vec_orc (vector bool unsigned int,
14968 vector unsigned int);
14969 vector unsigned int vec_orc (vector unsigned int,
14970 vector bool unsigned int);
14971 vector short vec_orc (vector short, vector short);
14972 vector short vec_orc (vector bool short, vector short);
14973 vector short vec_orc (vector short, vector bool short);
14974 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
14975 vector unsigned short vec_orc (vector bool unsigned short,
14976 vector unsigned short);
14977 vector unsigned short vec_orc (vector unsigned short,
14978 vector bool unsigned short);
14979 vector signed char vec_orc (vector signed char, vector signed char);
14980 vector signed char vec_orc (vector bool signed char, vector signed char);
14981 vector signed char vec_orc (vector signed char, vector bool signed char);
14982 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
14983 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
14984 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
14986 vector int vec_pack (vector long long, vector long long);
14987 vector unsigned int vec_pack (vector unsigned long long,
14988 vector unsigned long long);
14989 vector bool int vec_pack (vector bool long long, vector bool long long);
14991 vector int vec_packs (vector long long, vector long long);
14992 vector unsigned int vec_packs (vector unsigned long long,
14993 vector unsigned long long);
14995 vector unsigned int vec_packsu (vector long long, vector long long);
14997 vector long long vec_rl (vector long long,
14998 vector unsigned long long);
14999 vector long long vec_rl (vector unsigned long long,
15000 vector unsigned long long);
15002 vector long long vec_sl (vector long long, vector unsigned long long);
15003 vector long long vec_sl (vector unsigned long long,
15004 vector unsigned long long);
15006 vector long long vec_sr (vector long long, vector unsigned long long);
15007 vector unsigned long long char vec_sr (vector unsigned long long,
15008 vector unsigned long long);
15010 vector long long vec_sra (vector long long, vector unsigned long long);
15011 vector unsigned long long vec_sra (vector unsigned long long,
15012 vector unsigned long long);
15014 vector long long vec_sub (vector long long, vector long long);
15015 vector unsigned long long vec_sub (vector unsigned long long,
15016 vector unsigned long long);
15018 vector long long vec_unpackh (vector int);
15019 vector unsigned long long vec_unpackh (vector unsigned int);
15021 vector long long vec_unpackl (vector int);
15022 vector unsigned long long vec_unpackl (vector unsigned int);
15024 vector long long vec_vaddudm (vector long long, vector long long);
15025 vector long long vec_vaddudm (vector bool long long, vector long long);
15026 vector long long vec_vaddudm (vector long long, vector bool long long);
15027 vector unsigned long long vec_vaddudm (vector unsigned long long,
15028 vector unsigned long long);
15029 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
15030 vector unsigned long long);
15031 vector unsigned long long vec_vaddudm (vector unsigned long long,
15032 vector bool unsigned long long);
15034 vector long long vec_vclz (vector long long);
15035 vector unsigned long long vec_vclz (vector unsigned long long);
15036 vector int vec_vclz (vector int);
15037 vector unsigned int vec_vclz (vector int);
15038 vector short vec_vclz (vector short);
15039 vector unsigned short vec_vclz (vector unsigned short);
15040 vector signed char vec_vclz (vector signed char);
15041 vector unsigned char vec_vclz (vector unsigned char);
15043 vector signed char vec_vclzb (vector signed char);
15044 vector unsigned char vec_vclzb (vector unsigned char);
15046 vector long long vec_vclzd (vector long long);
15047 vector unsigned long long vec_vclzd (vector unsigned long long);
15049 vector short vec_vclzh (vector short);
15050 vector unsigned short vec_vclzh (vector unsigned short);
15052 vector int vec_vclzw (vector int);
15053 vector unsigned int vec_vclzw (vector int);
15055 vector long long vec_vmaxsd (vector long long, vector long long);
15057 vector unsigned long long vec_vmaxud (vector unsigned long long,
15058 unsigned vector long long);
15060 vector long long vec_vminsd (vector long long, vector long long);
15062 vector unsigned long long vec_vminud (vector long long,
15065 vector int vec_vpksdss (vector long long, vector long long);
15066 vector unsigned int vec_vpksdss (vector long long, vector long long);
15068 vector unsigned int vec_vpkudus (vector unsigned long long,
15069 vector unsigned long long);
15071 vector int vec_vpkudum (vector long long, vector long long);
15072 vector unsigned int vec_vpkudum (vector unsigned long long,
15073 vector unsigned long long);
15074 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
15076 vector long long vec_vpopcnt (vector long long);
15077 vector unsigned long long vec_vpopcnt (vector unsigned long long);
15078 vector int vec_vpopcnt (vector int);
15079 vector unsigned int vec_vpopcnt (vector int);
15080 vector short vec_vpopcnt (vector short);
15081 vector unsigned short vec_vpopcnt (vector unsigned short);
15082 vector signed char vec_vpopcnt (vector signed char);
15083 vector unsigned char vec_vpopcnt (vector unsigned char);
15085 vector signed char vec_vpopcntb (vector signed char);
15086 vector unsigned char vec_vpopcntb (vector unsigned char);
15088 vector long long vec_vpopcntd (vector long long);
15089 vector unsigned long long vec_vpopcntd (vector unsigned long long);
15091 vector short vec_vpopcnth (vector short);
15092 vector unsigned short vec_vpopcnth (vector unsigned short);
15094 vector int vec_vpopcntw (vector int);
15095 vector unsigned int vec_vpopcntw (vector int);
15097 vector long long vec_vrld (vector long long, vector unsigned long long);
15098 vector unsigned long long vec_vrld (vector unsigned long long,
15099 vector unsigned long long);
15101 vector long long vec_vsld (vector long long, vector unsigned long long);
15102 vector long long vec_vsld (vector unsigned long long,
15103 vector unsigned long long);
15105 vector long long vec_vsrad (vector long long, vector unsigned long long);
15106 vector unsigned long long vec_vsrad (vector unsigned long long,
15107 vector unsigned long long);
15109 vector long long vec_vsrd (vector long long, vector unsigned long long);
15110 vector unsigned long long char vec_vsrd (vector unsigned long long,
15111 vector unsigned long long);
15113 vector long long vec_vsubudm (vector long long, vector long long);
15114 vector long long vec_vsubudm (vector bool long long, vector long long);
15115 vector long long vec_vsubudm (vector long long, vector bool long long);
15116 vector unsigned long long vec_vsubudm (vector unsigned long long,
15117 vector unsigned long long);
15118 vector unsigned long long vec_vsubudm (vector bool long long,
15119 vector unsigned long long);
15120 vector unsigned long long vec_vsubudm (vector unsigned long long,
15121 vector bool long long);
15123 vector long long vec_vupkhsw (vector int);
15124 vector unsigned long long vec_vupkhsw (vector unsigned int);
15126 vector long long vec_vupklsw (vector int);
15127 vector unsigned long long vec_vupklsw (vector int);
15130 If the cryptographic instructions are enabled (@option{-mcrypto} or
15131 @option{-mcpu=power8}), the following builtins are enabled.
15134 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
15136 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
15137 vector unsigned long long);
15139 vector unsigned long long __builtin_crypto_vcipherlast
15140 (vector unsigned long long,
15141 vector unsigned long long);
15143 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
15144 vector unsigned long long);
15146 vector unsigned long long __builtin_crypto_vncipherlast
15147 (vector unsigned long long,
15148 vector unsigned long long);
15150 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
15151 vector unsigned char,
15152 vector unsigned char);
15154 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
15155 vector unsigned short,
15156 vector unsigned short);
15158 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
15159 vector unsigned int,
15160 vector unsigned int);
15162 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
15163 vector unsigned long long,
15164 vector unsigned long long);
15166 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
15167 vector unsigned char);
15169 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
15170 vector unsigned short);
15172 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
15173 vector unsigned int);
15175 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
15176 vector unsigned long long);
15178 vector unsigned long long __builtin_crypto_vshasigmad
15179 (vector unsigned long long, int, int);
15181 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
15185 The second argument to the @var{__builtin_crypto_vshasigmad} and
15186 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
15187 integer that is 0 or 1. The third argument to these builtin functions
15188 must be a constant integer in the range of 0 to 15.
15190 @node PowerPC Hardware Transactional Memory Built-in Functions
15191 @subsection PowerPC Hardware Transactional Memory Built-in Functions
15192 GCC provides two interfaces for accessing the Hardware Transactional
15193 Memory (HTM) instructions available on some of the PowerPC family
15194 of prcoessors (eg, POWER8). The two interfaces come in a low level
15195 interface, consisting of built-in functions specific to PowerPC and a
15196 higher level interface consisting of inline functions that are common
15197 between PowerPC and S/390.
15199 @subsubsection PowerPC HTM Low Level Built-in Functions
15201 The following low level built-in functions are available with
15202 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
15203 They all generate the machine instruction that is part of the name.
15205 The HTM built-ins return true or false depending on their success and
15206 their arguments match exactly the type and order of the associated
15207 hardware instruction's operands. Refer to the ISA manual for a
15208 description of each instruction's operands.
15211 unsigned int __builtin_tbegin (unsigned int)
15212 unsigned int __builtin_tend (unsigned int)
15214 unsigned int __builtin_tabort (unsigned int)
15215 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
15216 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
15217 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
15218 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
15220 unsigned int __builtin_tcheck (unsigned int)
15221 unsigned int __builtin_treclaim (unsigned int)
15222 unsigned int __builtin_trechkpt (void)
15223 unsigned int __builtin_tsr (unsigned int)
15226 In addition to the above HTM built-ins, we have added built-ins for
15227 some common extended mnemonics of the HTM instructions:
15230 unsigned int __builtin_tendall (void)
15231 unsigned int __builtin_tresume (void)
15232 unsigned int __builtin_tsuspend (void)
15235 The following set of built-in functions are available to gain access
15236 to the HTM specific special purpose registers.
15239 unsigned long __builtin_get_texasr (void)
15240 unsigned long __builtin_get_texasru (void)
15241 unsigned long __builtin_get_tfhar (void)
15242 unsigned long __builtin_get_tfiar (void)
15244 void __builtin_set_texasr (unsigned long);
15245 void __builtin_set_texasru (unsigned long);
15246 void __builtin_set_tfhar (unsigned long);
15247 void __builtin_set_tfiar (unsigned long);
15250 Example usage of these low level built-in functions may look like:
15253 #include <htmintrin.h>
15255 int num_retries = 10;
15259 if (__builtin_tbegin (0))
15261 /* Transaction State Initiated. */
15262 if (is_locked (lock))
15263 __builtin_tabort (0);
15264 ... transaction code...
15265 __builtin_tend (0);
15270 /* Transaction State Failed. Use locks if the transaction
15271 failure is "persistent" or we've tried too many times. */
15272 if (num_retries-- <= 0
15273 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
15275 acquire_lock (lock);
15276 ... non transactional fallback path...
15277 release_lock (lock);
15284 One final built-in function has been added that returns the value of
15285 the 2-bit Transaction State field of the Machine Status Register (MSR)
15286 as stored in @code{CR0}.
15289 unsigned long __builtin_ttest (void)
15292 This built-in can be used to determine the current transaction state
15293 using the following code example:
15296 #include <htmintrin.h>
15298 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
15300 if (tx_state == _HTM_TRANSACTIONAL)
15302 /* Code to use in transactional state. */
15304 else if (tx_state == _HTM_NONTRANSACTIONAL)
15306 /* Code to use in non-transactional state. */
15308 else if (tx_state == _HTM_SUSPENDED)
15310 /* Code to use in transaction suspended state. */
15314 @subsubsection PowerPC HTM High Level Inline Functions
15316 The following high level HTM interface is made available by including
15317 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
15318 where CPU is `power8' or later. This interface is common between PowerPC
15319 and S/390, allowing users to write one HTM source implementation that
15320 can be compiled and executed on either system.
15323 long __TM_simple_begin (void)
15324 long __TM_begin (void* const TM_buff)
15325 long __TM_end (void)
15326 void __TM_abort (void)
15327 void __TM_named_abort (unsigned char const code)
15328 void __TM_resume (void)
15329 void __TM_suspend (void)
15331 long __TM_is_user_abort (void* const TM_buff)
15332 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
15333 long __TM_is_illegal (void* const TM_buff)
15334 long __TM_is_footprint_exceeded (void* const TM_buff)
15335 long __TM_nesting_depth (void* const TM_buff)
15336 long __TM_is_nested_too_deep(void* const TM_buff)
15337 long __TM_is_conflict(void* const TM_buff)
15338 long __TM_is_failure_persistent(void* const TM_buff)
15339 long __TM_failure_address(void* const TM_buff)
15340 long long __TM_failure_code(void* const TM_buff)
15343 Using these common set of HTM inline functions, we can create
15344 a more portable version of the HTM example in the previous
15345 section that will work on either PowerPC or S/390:
15348 #include <htmxlintrin.h>
15350 int num_retries = 10;
15351 TM_buff_type TM_buff;
15355 if (__TM_begin (TM_buff))
15357 /* Transaction State Initiated. */
15358 if (is_locked (lock))
15360 ... transaction code...
15366 /* Transaction State Failed. Use locks if the transaction
15367 failure is "persistent" or we've tried too many times. */
15368 if (num_retries-- <= 0
15369 || __TM_is_failure_persistent (TM_buff))
15371 acquire_lock (lock);
15372 ... non transactional fallback path...
15373 release_lock (lock);
15380 @node RX Built-in Functions
15381 @subsection RX Built-in Functions
15382 GCC supports some of the RX instructions which cannot be expressed in
15383 the C programming language via the use of built-in functions. The
15384 following functions are supported:
15386 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
15387 Generates the @code{brk} machine instruction.
15390 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
15391 Generates the @code{clrpsw} machine instruction to clear the specified
15392 bit in the processor status word.
15395 @deftypefn {Built-in Function} void __builtin_rx_int (int)
15396 Generates the @code{int} machine instruction to generate an interrupt
15397 with the specified value.
15400 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
15401 Generates the @code{machi} machine instruction to add the result of
15402 multiplying the top 16 bits of the two arguments into the
15406 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
15407 Generates the @code{maclo} machine instruction to add the result of
15408 multiplying the bottom 16 bits of the two arguments into the
15412 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
15413 Generates the @code{mulhi} machine instruction to place the result of
15414 multiplying the top 16 bits of the two arguments into the
15418 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
15419 Generates the @code{mullo} machine instruction to place the result of
15420 multiplying the bottom 16 bits of the two arguments into the
15424 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
15425 Generates the @code{mvfachi} machine instruction to read the top
15426 32 bits of the accumulator.
15429 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
15430 Generates the @code{mvfacmi} machine instruction to read the middle
15431 32 bits of the accumulator.
15434 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
15435 Generates the @code{mvfc} machine instruction which reads the control
15436 register specified in its argument and returns its value.
15439 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
15440 Generates the @code{mvtachi} machine instruction to set the top
15441 32 bits of the accumulator.
15444 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
15445 Generates the @code{mvtaclo} machine instruction to set the bottom
15446 32 bits of the accumulator.
15449 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
15450 Generates the @code{mvtc} machine instruction which sets control
15451 register number @code{reg} to @code{val}.
15454 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
15455 Generates the @code{mvtipl} machine instruction set the interrupt
15459 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
15460 Generates the @code{racw} machine instruction to round the accumulator
15461 according to the specified mode.
15464 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
15465 Generates the @code{revw} machine instruction which swaps the bytes in
15466 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
15467 and also bits 16--23 occupy bits 24--31 and vice versa.
15470 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
15471 Generates the @code{rmpa} machine instruction which initiates a
15472 repeated multiply and accumulate sequence.
15475 @deftypefn {Built-in Function} void __builtin_rx_round (float)
15476 Generates the @code{round} machine instruction which returns the
15477 floating-point argument rounded according to the current rounding mode
15478 set in the floating-point status word register.
15481 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
15482 Generates the @code{sat} machine instruction which returns the
15483 saturated value of the argument.
15486 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
15487 Generates the @code{setpsw} machine instruction to set the specified
15488 bit in the processor status word.
15491 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
15492 Generates the @code{wait} machine instruction.
15495 @node S/390 System z Built-in Functions
15496 @subsection S/390 System z Built-in Functions
15497 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
15498 Generates the @code{tbegin} machine instruction starting a
15499 non-constraint hardware transaction. If the parameter is non-NULL the
15500 memory area is used to store the transaction diagnostic buffer and
15501 will be passed as first operand to @code{tbegin}. This buffer can be
15502 defined using the @code{struct __htm_tdb} C struct defined in
15503 @code{htmintrin.h} and must reside on a double-word boundary. The
15504 second tbegin operand is set to @code{0xff0c}. This enables
15505 save/restore of all GPRs and disables aborts for FPR and AR
15506 manipulations inside the transaction body. The condition code set by
15507 the tbegin instruction is returned as integer value. The tbegin
15508 instruction by definition overwrites the content of all FPRs. The
15509 compiler will generate code which saves and restores the FPRs. For
15510 soft-float code it is recommended to used the @code{*_nofloat}
15511 variant. In order to prevent a TDB from being written it is required
15512 to pass an constant zero value as parameter. Passing the zero value
15513 through a variable is not sufficient. Although modifications of
15514 access registers inside the transaction will not trigger an
15515 transaction abort it is not supported to actually modify them. Access
15516 registers do not get saved when entering a transaction. They will have
15517 undefined state when reaching the abort code.
15520 Macros for the possible return codes of tbegin are defined in the
15521 @code{htmintrin.h} header file:
15524 @item _HTM_TBEGIN_STARTED
15525 @code{tbegin} has been executed as part of normal processing. The
15526 transaction body is supposed to be executed.
15527 @item _HTM_TBEGIN_INDETERMINATE
15528 The transaction was aborted due to an indeterminate condition which
15529 might be persistent.
15530 @item _HTM_TBEGIN_TRANSIENT
15531 The transaction aborted due to a transient failure. The transaction
15532 should be re-executed in that case.
15533 @item _HTM_TBEGIN_PERSISTENT
15534 The transaction aborted due to a persistent failure. Re-execution
15535 under same circumstances will not be productive.
15538 @defmac _HTM_FIRST_USER_ABORT_CODE
15539 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
15540 specifies the first abort code which can be used for
15541 @code{__builtin_tabort}. Values below this threshold are reserved for
15545 @deftp {Data type} {struct __htm_tdb}
15546 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
15547 the structure of the transaction diagnostic block as specified in the
15548 Principles of Operation manual chapter 5-91.
15551 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
15552 Same as @code{__builtin_tbegin} but without FPR saves and restores.
15553 Using this variant in code making use of FPRs will leave the FPRs in
15554 undefined state when entering the transaction abort handler code.
15557 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
15558 In addition to @code{__builtin_tbegin} a loop for transient failures
15559 is generated. If tbegin returns a condition code of 2 the transaction
15560 will be retried as often as specified in the second argument. The
15561 perform processor assist instruction is used to tell the CPU about the
15562 number of fails so far.
15565 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
15566 Same as @code{__builtin_tbegin_retry} but without FPR saves and
15567 restores. Using this variant in code making use of FPRs will leave
15568 the FPRs in undefined state when entering the transaction abort
15572 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
15573 Generates the @code{tbeginc} machine instruction starting a constraint
15574 hardware transaction. The second operand is set to @code{0xff08}.
15577 @deftypefn {Built-in Function} int __builtin_tend (void)
15578 Generates the @code{tend} machine instruction finishing a transaction
15579 and making the changes visible to other threads. The condition code
15580 generated by tend is returned as integer value.
15583 @deftypefn {Built-in Function} void __builtin_tabort (int)
15584 Generates the @code{tabort} machine instruction with the specified
15585 abort code. Abort codes from 0 through 255 are reserved and will
15586 result in an error message.
15589 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
15590 Generates the @code{ppa rX,rY,1} machine instruction. Where the
15591 integer parameter is loaded into rX and a value of zero is loaded into
15592 rY. The integer parameter specifies the number of times the
15593 transaction repeatedly aborted.
15596 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
15597 Generates the @code{etnd} machine instruction. The current nesting
15598 depth is returned as integer value. For a nesting depth of 0 the code
15599 is not executed as part of an transaction.
15602 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
15604 Generates the @code{ntstg} machine instruction. The second argument
15605 is written to the first arguments location. The store operation will
15606 not be rolled-back in case of an transaction abort.
15609 @node SH Built-in Functions
15610 @subsection SH Built-in Functions
15611 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
15612 families of processors:
15614 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
15615 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
15616 used by system code that manages threads and execution contexts. The compiler
15617 normally does not generate code that modifies the contents of @samp{GBR} and
15618 thus the value is preserved across function calls. Changing the @samp{GBR}
15619 value in user code must be done with caution, since the compiler might use
15620 @samp{GBR} in order to access thread local variables.
15624 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
15625 Returns the value that is currently set in the @samp{GBR} register.
15626 Memory loads and stores that use the thread pointer as a base address are
15627 turned into @samp{GBR} based displacement loads and stores, if possible.
15635 int get_tcb_value (void)
15637 // Generate @samp{mov.l @@(8,gbr),r0} instruction
15638 return ((my_tcb*)__builtin_thread_pointer ())->c;
15644 @node SPARC VIS Built-in Functions
15645 @subsection SPARC VIS Built-in Functions
15647 GCC supports SIMD operations on the SPARC using both the generic vector
15648 extensions (@pxref{Vector Extensions}) as well as built-in functions for
15649 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
15650 switch, the VIS extension is exposed as the following built-in functions:
15653 typedef int v1si __attribute__ ((vector_size (4)));
15654 typedef int v2si __attribute__ ((vector_size (8)));
15655 typedef short v4hi __attribute__ ((vector_size (8)));
15656 typedef short v2hi __attribute__ ((vector_size (4)));
15657 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
15658 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
15660 void __builtin_vis_write_gsr (int64_t);
15661 int64_t __builtin_vis_read_gsr (void);
15663 void * __builtin_vis_alignaddr (void *, long);
15664 void * __builtin_vis_alignaddrl (void *, long);
15665 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
15666 v2si __builtin_vis_faligndatav2si (v2si, v2si);
15667 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
15668 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
15670 v4hi __builtin_vis_fexpand (v4qi);
15672 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
15673 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
15674 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
15675 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
15676 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
15677 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
15678 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
15680 v4qi __builtin_vis_fpack16 (v4hi);
15681 v8qi __builtin_vis_fpack32 (v2si, v8qi);
15682 v2hi __builtin_vis_fpackfix (v2si);
15683 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
15685 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
15687 long __builtin_vis_edge8 (void *, void *);
15688 long __builtin_vis_edge8l (void *, void *);
15689 long __builtin_vis_edge16 (void *, void *);
15690 long __builtin_vis_edge16l (void *, void *);
15691 long __builtin_vis_edge32 (void *, void *);
15692 long __builtin_vis_edge32l (void *, void *);
15694 long __builtin_vis_fcmple16 (v4hi, v4hi);
15695 long __builtin_vis_fcmple32 (v2si, v2si);
15696 long __builtin_vis_fcmpne16 (v4hi, v4hi);
15697 long __builtin_vis_fcmpne32 (v2si, v2si);
15698 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
15699 long __builtin_vis_fcmpgt32 (v2si, v2si);
15700 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
15701 long __builtin_vis_fcmpeq32 (v2si, v2si);
15703 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
15704 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
15705 v2si __builtin_vis_fpadd32 (v2si, v2si);
15706 v1si __builtin_vis_fpadd32s (v1si, v1si);
15707 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
15708 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
15709 v2si __builtin_vis_fpsub32 (v2si, v2si);
15710 v1si __builtin_vis_fpsub32s (v1si, v1si);
15712 long __builtin_vis_array8 (long, long);
15713 long __builtin_vis_array16 (long, long);
15714 long __builtin_vis_array32 (long, long);
15717 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
15718 functions also become available:
15721 long __builtin_vis_bmask (long, long);
15722 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
15723 v2si __builtin_vis_bshufflev2si (v2si, v2si);
15724 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
15725 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
15727 long __builtin_vis_edge8n (void *, void *);
15728 long __builtin_vis_edge8ln (void *, void *);
15729 long __builtin_vis_edge16n (void *, void *);
15730 long __builtin_vis_edge16ln (void *, void *);
15731 long __builtin_vis_edge32n (void *, void *);
15732 long __builtin_vis_edge32ln (void *, void *);
15735 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
15736 functions also become available:
15739 void __builtin_vis_cmask8 (long);
15740 void __builtin_vis_cmask16 (long);
15741 void __builtin_vis_cmask32 (long);
15743 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
15745 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
15746 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
15747 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
15748 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
15749 v2si __builtin_vis_fsll16 (v2si, v2si);
15750 v2si __builtin_vis_fslas16 (v2si, v2si);
15751 v2si __builtin_vis_fsrl16 (v2si, v2si);
15752 v2si __builtin_vis_fsra16 (v2si, v2si);
15754 long __builtin_vis_pdistn (v8qi, v8qi);
15756 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
15758 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
15759 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
15761 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
15762 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
15763 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
15764 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
15765 v2si __builtin_vis_fpadds32 (v2si, v2si);
15766 v1si __builtin_vis_fpadds32s (v1si, v1si);
15767 v2si __builtin_vis_fpsubs32 (v2si, v2si);
15768 v1si __builtin_vis_fpsubs32s (v1si, v1si);
15770 long __builtin_vis_fucmple8 (v8qi, v8qi);
15771 long __builtin_vis_fucmpne8 (v8qi, v8qi);
15772 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
15773 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
15775 float __builtin_vis_fhadds (float, float);
15776 double __builtin_vis_fhaddd (double, double);
15777 float __builtin_vis_fhsubs (float, float);
15778 double __builtin_vis_fhsubd (double, double);
15779 float __builtin_vis_fnhadds (float, float);
15780 double __builtin_vis_fnhaddd (double, double);
15782 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
15783 int64_t __builtin_vis_xmulx (int64_t, int64_t);
15784 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
15787 @node SPU Built-in Functions
15788 @subsection SPU Built-in Functions
15790 GCC provides extensions for the SPU processor as described in the
15791 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
15792 found at @uref{http://cell.scei.co.jp/} or
15793 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
15794 implementation differs in several ways.
15799 The optional extension of specifying vector constants in parentheses is
15803 A vector initializer requires no cast if the vector constant is of the
15804 same type as the variable it is initializing.
15807 If @code{signed} or @code{unsigned} is omitted, the signedness of the
15808 vector type is the default signedness of the base type. The default
15809 varies depending on the operating system, so a portable program should
15810 always specify the signedness.
15813 By default, the keyword @code{__vector} is added. The macro
15814 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
15818 GCC allows using a @code{typedef} name as the type specifier for a
15822 For C, overloaded functions are implemented with macros so the following
15826 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
15830 Since @code{spu_add} is a macro, the vector constant in the example
15831 is treated as four separate arguments. Wrap the entire argument in
15832 parentheses for this to work.
15835 The extended version of @code{__builtin_expect} is not supported.
15839 @emph{Note:} Only the interface described in the aforementioned
15840 specification is supported. Internally, GCC uses built-in functions to
15841 implement the required functionality, but these are not supported and
15842 are subject to change without notice.
15844 @node TI C6X Built-in Functions
15845 @subsection TI C6X Built-in Functions
15847 GCC provides intrinsics to access certain instructions of the TI C6X
15848 processors. These intrinsics, listed below, are available after
15849 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
15850 to C6X instructions.
15854 int _sadd (int, int)
15855 int _ssub (int, int)
15856 int _sadd2 (int, int)
15857 int _ssub2 (int, int)
15858 long long _mpy2 (int, int)
15859 long long _smpy2 (int, int)
15860 int _add4 (int, int)
15861 int _sub4 (int, int)
15862 int _saddu4 (int, int)
15864 int _smpy (int, int)
15865 int _smpyh (int, int)
15866 int _smpyhl (int, int)
15867 int _smpylh (int, int)
15869 int _sshl (int, int)
15870 int _subc (int, int)
15872 int _avg2 (int, int)
15873 int _avgu4 (int, int)
15875 int _clrr (int, int)
15876 int _extr (int, int)
15877 int _extru (int, int)
15883 @node TILE-Gx Built-in Functions
15884 @subsection TILE-Gx Built-in Functions
15886 GCC provides intrinsics to access every instruction of the TILE-Gx
15887 processor. The intrinsics are of the form:
15891 unsigned long long __insn_@var{op} (...)
15895 Where @var{op} is the name of the instruction. Refer to the ISA manual
15896 for the complete list of instructions.
15898 GCC also provides intrinsics to directly access the network registers.
15899 The intrinsics are:
15903 unsigned long long __tile_idn0_receive (void)
15904 unsigned long long __tile_idn1_receive (void)
15905 unsigned long long __tile_udn0_receive (void)
15906 unsigned long long __tile_udn1_receive (void)
15907 unsigned long long __tile_udn2_receive (void)
15908 unsigned long long __tile_udn3_receive (void)
15909 void __tile_idn_send (unsigned long long)
15910 void __tile_udn_send (unsigned long long)
15914 The intrinsic @code{void __tile_network_barrier (void)} is used to
15915 guarantee that no network operations before it are reordered with
15918 @node TILEPro Built-in Functions
15919 @subsection TILEPro Built-in Functions
15921 GCC provides intrinsics to access every instruction of the TILEPro
15922 processor. The intrinsics are of the form:
15926 unsigned __insn_@var{op} (...)
15931 where @var{op} is the name of the instruction. Refer to the ISA manual
15932 for the complete list of instructions.
15934 GCC also provides intrinsics to directly access the network registers.
15935 The intrinsics are:
15939 unsigned __tile_idn0_receive (void)
15940 unsigned __tile_idn1_receive (void)
15941 unsigned __tile_sn_receive (void)
15942 unsigned __tile_udn0_receive (void)
15943 unsigned __tile_udn1_receive (void)
15944 unsigned __tile_udn2_receive (void)
15945 unsigned __tile_udn3_receive (void)
15946 void __tile_idn_send (unsigned)
15947 void __tile_sn_send (unsigned)
15948 void __tile_udn_send (unsigned)
15952 The intrinsic @code{void __tile_network_barrier (void)} is used to
15953 guarantee that no network operations before it are reordered with
15956 @node Target Format Checks
15957 @section Format Checks Specific to Particular Target Machines
15959 For some target machines, GCC supports additional options to the
15961 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
15964 * Solaris Format Checks::
15965 * Darwin Format Checks::
15968 @node Solaris Format Checks
15969 @subsection Solaris Format Checks
15971 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
15972 check. @code{cmn_err} accepts a subset of the standard @code{printf}
15973 conversions, and the two-argument @code{%b} conversion for displaying
15974 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
15976 @node Darwin Format Checks
15977 @subsection Darwin Format Checks
15979 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
15980 attribute context. Declarations made with such attribution are parsed for correct syntax
15981 and format argument types. However, parsing of the format string itself is currently undefined
15982 and is not carried out by this version of the compiler.
15984 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
15985 also be used as format arguments. Note that the relevant headers are only likely to be
15986 available on Darwin (OSX) installations. On such installations, the XCode and system
15987 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
15988 associated functions.
15991 @section Pragmas Accepted by GCC
15993 @cindex @code{#pragma}
15995 GCC supports several types of pragmas, primarily in order to compile
15996 code originally written for other compilers. Note that in general
15997 we do not recommend the use of pragmas; @xref{Function Attributes},
15998 for further explanation.
16004 * RS/6000 and PowerPC Pragmas::
16006 * Solaris Pragmas::
16007 * Symbol-Renaming Pragmas::
16008 * Structure-Packing Pragmas::
16010 * Diagnostic Pragmas::
16011 * Visibility Pragmas::
16012 * Push/Pop Macro Pragmas::
16013 * Function Specific Option Pragmas::
16014 * Loop-Specific Pragmas::
16018 @subsection ARM Pragmas
16020 The ARM target defines pragmas for controlling the default addition of
16021 @code{long_call} and @code{short_call} attributes to functions.
16022 @xref{Function Attributes}, for information about the effects of these
16027 @cindex pragma, long_calls
16028 Set all subsequent functions to have the @code{long_call} attribute.
16030 @item no_long_calls
16031 @cindex pragma, no_long_calls
16032 Set all subsequent functions to have the @code{short_call} attribute.
16034 @item long_calls_off
16035 @cindex pragma, long_calls_off
16036 Do not affect the @code{long_call} or @code{short_call} attributes of
16037 subsequent functions.
16041 @subsection M32C Pragmas
16044 @item GCC memregs @var{number}
16045 @cindex pragma, memregs
16046 Overrides the command-line option @code{-memregs=} for the current
16047 file. Use with care! This pragma must be before any function in the
16048 file, and mixing different memregs values in different objects may
16049 make them incompatible. This pragma is useful when a
16050 performance-critical function uses a memreg for temporary values,
16051 as it may allow you to reduce the number of memregs used.
16053 @item ADDRESS @var{name} @var{address}
16054 @cindex pragma, address
16055 For any declared symbols matching @var{name}, this does three things
16056 to that symbol: it forces the symbol to be located at the given
16057 address (a number), it forces the symbol to be volatile, and it
16058 changes the symbol's scope to be static. This pragma exists for
16059 compatibility with other compilers, but note that the common
16060 @code{1234H} numeric syntax is not supported (use @code{0x1234}
16064 #pragma ADDRESS port3 0x103
16071 @subsection MeP Pragmas
16075 @item custom io_volatile (on|off)
16076 @cindex pragma, custom io_volatile
16077 Overrides the command-line option @code{-mio-volatile} for the current
16078 file. Note that for compatibility with future GCC releases, this
16079 option should only be used once before any @code{io} variables in each
16082 @item GCC coprocessor available @var{registers}
16083 @cindex pragma, coprocessor available
16084 Specifies which coprocessor registers are available to the register
16085 allocator. @var{registers} may be a single register, register range
16086 separated by ellipses, or comma-separated list of those. Example:
16089 #pragma GCC coprocessor available $c0...$c10, $c28
16092 @item GCC coprocessor call_saved @var{registers}
16093 @cindex pragma, coprocessor call_saved
16094 Specifies which coprocessor registers are to be saved and restored by
16095 any function using them. @var{registers} may be a single register,
16096 register range separated by ellipses, or comma-separated list of
16100 #pragma GCC coprocessor call_saved $c4...$c6, $c31
16103 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
16104 @cindex pragma, coprocessor subclass
16105 Creates and defines a register class. These register classes can be
16106 used by inline @code{asm} constructs. @var{registers} may be a single
16107 register, register range separated by ellipses, or comma-separated
16108 list of those. Example:
16111 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
16113 asm ("cpfoo %0" : "=B" (x));
16116 @item GCC disinterrupt @var{name} , @var{name} @dots{}
16117 @cindex pragma, disinterrupt
16118 For the named functions, the compiler adds code to disable interrupts
16119 for the duration of those functions. If any functions so named
16120 are not encountered in the source, a warning is emitted that the pragma is
16121 not used. Examples:
16124 #pragma disinterrupt foo
16125 #pragma disinterrupt bar, grill
16126 int foo () @{ @dots{} @}
16129 @item GCC call @var{name} , @var{name} @dots{}
16130 @cindex pragma, call
16131 For the named functions, the compiler always uses a register-indirect
16132 call model when calling the named functions. Examples:
16141 @node RS/6000 and PowerPC Pragmas
16142 @subsection RS/6000 and PowerPC Pragmas
16144 The RS/6000 and PowerPC targets define one pragma for controlling
16145 whether or not the @code{longcall} attribute is added to function
16146 declarations by default. This pragma overrides the @option{-mlongcall}
16147 option, but not the @code{longcall} and @code{shortcall} attributes.
16148 @xref{RS/6000 and PowerPC Options}, for more information about when long
16149 calls are and are not necessary.
16153 @cindex pragma, longcall
16154 Apply the @code{longcall} attribute to all subsequent function
16158 Do not apply the @code{longcall} attribute to subsequent function
16162 @c Describe h8300 pragmas here.
16163 @c Describe sh pragmas here.
16164 @c Describe v850 pragmas here.
16166 @node Darwin Pragmas
16167 @subsection Darwin Pragmas
16169 The following pragmas are available for all architectures running the
16170 Darwin operating system. These are useful for compatibility with other
16174 @item mark @var{tokens}@dots{}
16175 @cindex pragma, mark
16176 This pragma is accepted, but has no effect.
16178 @item options align=@var{alignment}
16179 @cindex pragma, options align
16180 This pragma sets the alignment of fields in structures. The values of
16181 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
16182 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
16183 properly; to restore the previous setting, use @code{reset} for the
16186 @item segment @var{tokens}@dots{}
16187 @cindex pragma, segment
16188 This pragma is accepted, but has no effect.
16190 @item unused (@var{var} [, @var{var}]@dots{})
16191 @cindex pragma, unused
16192 This pragma declares variables to be possibly unused. GCC does not
16193 produce warnings for the listed variables. The effect is similar to
16194 that of the @code{unused} attribute, except that this pragma may appear
16195 anywhere within the variables' scopes.
16198 @node Solaris Pragmas
16199 @subsection Solaris Pragmas
16201 The Solaris target supports @code{#pragma redefine_extname}
16202 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
16203 @code{#pragma} directives for compatibility with the system compiler.
16206 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
16207 @cindex pragma, align
16209 Increase the minimum alignment of each @var{variable} to @var{alignment}.
16210 This is the same as GCC's @code{aligned} attribute @pxref{Variable
16211 Attributes}). Macro expansion occurs on the arguments to this pragma
16212 when compiling C and Objective-C@. It does not currently occur when
16213 compiling C++, but this is a bug which may be fixed in a future
16216 @item fini (@var{function} [, @var{function}]...)
16217 @cindex pragma, fini
16219 This pragma causes each listed @var{function} to be called after
16220 main, or during shared module unloading, by adding a call to the
16221 @code{.fini} section.
16223 @item init (@var{function} [, @var{function}]...)
16224 @cindex pragma, init
16226 This pragma causes each listed @var{function} to be called during
16227 initialization (before @code{main}) or during shared module loading, by
16228 adding a call to the @code{.init} section.
16232 @node Symbol-Renaming Pragmas
16233 @subsection Symbol-Renaming Pragmas
16235 For compatibility with the Solaris system headers, GCC
16236 supports two @code{#pragma} directives that change the name used in
16237 assembly for a given declaration. To get this effect
16238 on all platforms supported by GCC, use the asm labels extension (@pxref{Asm
16242 @item redefine_extname @var{oldname} @var{newname}
16243 @cindex pragma, redefine_extname
16245 This pragma gives the C function @var{oldname} the assembly symbol
16246 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
16247 is defined if this pragma is available (currently on all platforms).
16250 This pragma and the asm labels extension interact in a complicated
16251 manner. Here are some corner cases you may want to be aware of.
16254 @item Both pragmas silently apply only to declarations with external
16255 linkage. Asm labels do not have this restriction.
16257 @item In C++, both pragmas silently apply only to declarations with
16258 ``C'' linkage. Again, asm labels do not have this restriction.
16260 @item If any of the three ways of changing the assembly name of a
16261 declaration is applied to a declaration whose assembly name has
16262 already been determined (either by a previous use of one of these
16263 features, or because the compiler needed the assembly name in order to
16264 generate code), and the new name is different, a warning issues and
16265 the name does not change.
16267 @item The @var{oldname} used by @code{#pragma redefine_extname} is
16268 always the C-language name.
16271 @node Structure-Packing Pragmas
16272 @subsection Structure-Packing Pragmas
16274 For compatibility with Microsoft Windows compilers, GCC supports a
16275 set of @code{#pragma} directives that change the maximum alignment of
16276 members of structures (other than zero-width bit-fields), unions, and
16277 classes subsequently defined. The @var{n} value below always is required
16278 to be a small power of two and specifies the new alignment in bytes.
16281 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
16282 @item @code{#pragma pack()} sets the alignment to the one that was in
16283 effect when compilation started (see also command-line option
16284 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
16285 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
16286 setting on an internal stack and then optionally sets the new alignment.
16287 @item @code{#pragma pack(pop)} restores the alignment setting to the one
16288 saved at the top of the internal stack (and removes that stack entry).
16289 Note that @code{#pragma pack([@var{n}])} does not influence this internal
16290 stack; thus it is possible to have @code{#pragma pack(push)} followed by
16291 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
16292 @code{#pragma pack(pop)}.
16295 Some targets, e.g.@: i386 and PowerPC, support the @code{ms_struct}
16296 @code{#pragma} which lays out a structure as the documented
16297 @code{__attribute__ ((ms_struct))}.
16299 @item @code{#pragma ms_struct on} turns on the layout for structures
16301 @item @code{#pragma ms_struct off} turns off the layout for structures
16303 @item @code{#pragma ms_struct reset} goes back to the default layout.
16307 @subsection Weak Pragmas
16309 For compatibility with SVR4, GCC supports a set of @code{#pragma}
16310 directives for declaring symbols to be weak, and defining weak
16314 @item #pragma weak @var{symbol}
16315 @cindex pragma, weak
16316 This pragma declares @var{symbol} to be weak, as if the declaration
16317 had the attribute of the same name. The pragma may appear before
16318 or after the declaration of @var{symbol}. It is not an error for
16319 @var{symbol} to never be defined at all.
16321 @item #pragma weak @var{symbol1} = @var{symbol2}
16322 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
16323 It is an error if @var{symbol2} is not defined in the current
16327 @node Diagnostic Pragmas
16328 @subsection Diagnostic Pragmas
16330 GCC allows the user to selectively enable or disable certain types of
16331 diagnostics, and change the kind of the diagnostic. For example, a
16332 project's policy might require that all sources compile with
16333 @option{-Werror} but certain files might have exceptions allowing
16334 specific types of warnings. Or, a project might selectively enable
16335 diagnostics and treat them as errors depending on which preprocessor
16336 macros are defined.
16339 @item #pragma GCC diagnostic @var{kind} @var{option}
16340 @cindex pragma, diagnostic
16342 Modifies the disposition of a diagnostic. Note that not all
16343 diagnostics are modifiable; at the moment only warnings (normally
16344 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
16345 Use @option{-fdiagnostics-show-option} to determine which diagnostics
16346 are controllable and which option controls them.
16348 @var{kind} is @samp{error} to treat this diagnostic as an error,
16349 @samp{warning} to treat it like a warning (even if @option{-Werror} is
16350 in effect), or @samp{ignored} if the diagnostic is to be ignored.
16351 @var{option} is a double quoted string that matches the command-line
16355 #pragma GCC diagnostic warning "-Wformat"
16356 #pragma GCC diagnostic error "-Wformat"
16357 #pragma GCC diagnostic ignored "-Wformat"
16360 Note that these pragmas override any command-line options. GCC keeps
16361 track of the location of each pragma, and issues diagnostics according
16362 to the state as of that point in the source file. Thus, pragmas occurring
16363 after a line do not affect diagnostics caused by that line.
16365 @item #pragma GCC diagnostic push
16366 @itemx #pragma GCC diagnostic pop
16368 Causes GCC to remember the state of the diagnostics as of each
16369 @code{push}, and restore to that point at each @code{pop}. If a
16370 @code{pop} has no matching @code{push}, the command-line options are
16374 #pragma GCC diagnostic error "-Wuninitialized"
16375 foo(a); /* error is given for this one */
16376 #pragma GCC diagnostic push
16377 #pragma GCC diagnostic ignored "-Wuninitialized"
16378 foo(b); /* no diagnostic for this one */
16379 #pragma GCC diagnostic pop
16380 foo(c); /* error is given for this one */
16381 #pragma GCC diagnostic pop
16382 foo(d); /* depends on command-line options */
16387 GCC also offers a simple mechanism for printing messages during
16391 @item #pragma message @var{string}
16392 @cindex pragma, diagnostic
16394 Prints @var{string} as a compiler message on compilation. The message
16395 is informational only, and is neither a compilation warning nor an error.
16398 #pragma message "Compiling " __FILE__ "..."
16401 @var{string} may be parenthesized, and is printed with location
16402 information. For example,
16405 #define DO_PRAGMA(x) _Pragma (#x)
16406 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
16408 TODO(Remember to fix this)
16412 prints @samp{/tmp/file.c:4: note: #pragma message:
16413 TODO - Remember to fix this}.
16417 @node Visibility Pragmas
16418 @subsection Visibility Pragmas
16421 @item #pragma GCC visibility push(@var{visibility})
16422 @itemx #pragma GCC visibility pop
16423 @cindex pragma, visibility
16425 This pragma allows the user to set the visibility for multiple
16426 declarations without having to give each a visibility attribute
16427 @xref{Function Attributes}, for more information about visibility and
16428 the attribute syntax.
16430 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
16431 declarations. Class members and template specializations are not
16432 affected; if you want to override the visibility for a particular
16433 member or instantiation, you must use an attribute.
16438 @node Push/Pop Macro Pragmas
16439 @subsection Push/Pop Macro Pragmas
16441 For compatibility with Microsoft Windows compilers, GCC supports
16442 @samp{#pragma push_macro(@var{"macro_name"})}
16443 and @samp{#pragma pop_macro(@var{"macro_name"})}.
16446 @item #pragma push_macro(@var{"macro_name"})
16447 @cindex pragma, push_macro
16448 This pragma saves the value of the macro named as @var{macro_name} to
16449 the top of the stack for this macro.
16451 @item #pragma pop_macro(@var{"macro_name"})
16452 @cindex pragma, pop_macro
16453 This pragma sets the value of the macro named as @var{macro_name} to
16454 the value on top of the stack for this macro. If the stack for
16455 @var{macro_name} is empty, the value of the macro remains unchanged.
16462 #pragma push_macro("X")
16465 #pragma pop_macro("X")
16470 In this example, the definition of X as 1 is saved by @code{#pragma
16471 push_macro} and restored by @code{#pragma pop_macro}.
16473 @node Function Specific Option Pragmas
16474 @subsection Function Specific Option Pragmas
16477 @item #pragma GCC target (@var{"string"}...)
16478 @cindex pragma GCC target
16480 This pragma allows you to set target specific options for functions
16481 defined later in the source file. One or more strings can be
16482 specified. Each function that is defined after this point is as
16483 if @code{attribute((target("STRING")))} was specified for that
16484 function. The parenthesis around the options is optional.
16485 @xref{Function Attributes}, for more information about the
16486 @code{target} attribute and the attribute syntax.
16488 The @code{#pragma GCC target} pragma is presently implemented for
16489 i386/x86_64, PowerPC, and Nios II targets only.
16493 @item #pragma GCC optimize (@var{"string"}...)
16494 @cindex pragma GCC optimize
16496 This pragma allows you to set global optimization options for functions
16497 defined later in the source file. One or more strings can be
16498 specified. Each function that is defined after this point is as
16499 if @code{attribute((optimize("STRING")))} was specified for that
16500 function. The parenthesis around the options is optional.
16501 @xref{Function Attributes}, for more information about the
16502 @code{optimize} attribute and the attribute syntax.
16504 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
16505 versions earlier than 4.4.
16509 @item #pragma GCC push_options
16510 @itemx #pragma GCC pop_options
16511 @cindex pragma GCC push_options
16512 @cindex pragma GCC pop_options
16514 These pragmas maintain a stack of the current target and optimization
16515 options. It is intended for include files where you temporarily want
16516 to switch to using a different @samp{#pragma GCC target} or
16517 @samp{#pragma GCC optimize} and then to pop back to the previous
16520 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
16521 pragmas are not implemented in GCC versions earlier than 4.4.
16525 @item #pragma GCC reset_options
16526 @cindex pragma GCC reset_options
16528 This pragma clears the current @code{#pragma GCC target} and
16529 @code{#pragma GCC optimize} to use the default switches as specified
16530 on the command line.
16532 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
16533 versions earlier than 4.4.
16536 @node Loop-Specific Pragmas
16537 @subsection Loop-Specific Pragmas
16540 @item #pragma GCC ivdep
16541 @cindex pragma GCC ivdep
16544 With this pragma, the programmer asserts that there are no loop-carried
16545 dependencies which would prevent that consecutive iterations of
16546 the following loop can be executed concurrently with SIMD
16547 (single instruction multiple data) instructions.
16549 For example, the compiler can only unconditionally vectorize the following
16550 loop with the pragma:
16553 void foo (int n, int *a, int *b, int *c)
16557 for (i = 0; i < n; ++i)
16558 a[i] = b[i] + c[i];
16563 In this example, using the @code{restrict} qualifier had the same
16564 effect. In the following example, that would not be possible. Assume
16565 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
16566 that it can unconditionally vectorize the following loop:
16569 void ignore_vec_dep (int *a, int k, int c, int m)
16572 for (int i = 0; i < m; i++)
16573 a[i] = a[i + k] * c;
16578 @node Unnamed Fields
16579 @section Unnamed struct/union fields within structs/unions
16580 @cindex @code{struct}
16581 @cindex @code{union}
16583 As permitted by ISO C11 and for compatibility with other compilers,
16584 GCC allows you to define
16585 a structure or union that contains, as fields, structures and unions
16586 without names. For example:
16600 In this example, you are able to access members of the unnamed
16601 union with code like @samp{foo.b}. Note that only unnamed structs and
16602 unions are allowed, you may not have, for example, an unnamed
16605 You must never create such structures that cause ambiguous field definitions.
16606 For example, in this structure:
16618 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
16619 The compiler gives errors for such constructs.
16621 @opindex fms-extensions
16622 Unless @option{-fms-extensions} is used, the unnamed field must be a
16623 structure or union definition without a tag (for example, @samp{struct
16624 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
16625 also be a definition with a tag such as @samp{struct foo @{ int a;
16626 @};}, a reference to a previously defined structure or union such as
16627 @samp{struct foo;}, or a reference to a @code{typedef} name for a
16628 previously defined structure or union type.
16630 @opindex fplan9-extensions
16631 The option @option{-fplan9-extensions} enables
16632 @option{-fms-extensions} as well as two other extensions. First, a
16633 pointer to a structure is automatically converted to a pointer to an
16634 anonymous field for assignments and function calls. For example:
16637 struct s1 @{ int a; @};
16638 struct s2 @{ struct s1; @};
16639 extern void f1 (struct s1 *);
16640 void f2 (struct s2 *p) @{ f1 (p); @}
16644 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
16645 converted into a pointer to the anonymous field.
16647 Second, when the type of an anonymous field is a @code{typedef} for a
16648 @code{struct} or @code{union}, code may refer to the field using the
16649 name of the @code{typedef}.
16652 typedef struct @{ int a; @} s1;
16653 struct s2 @{ s1; @};
16654 s1 f1 (struct s2 *p) @{ return p->s1; @}
16657 These usages are only permitted when they are not ambiguous.
16660 @section Thread-Local Storage
16661 @cindex Thread-Local Storage
16662 @cindex @acronym{TLS}
16663 @cindex @code{__thread}
16665 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
16666 are allocated such that there is one instance of the variable per extant
16667 thread. The runtime model GCC uses to implement this originates
16668 in the IA-64 processor-specific ABI, but has since been migrated
16669 to other processors as well. It requires significant support from
16670 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
16671 system libraries (@file{libc.so} and @file{libpthread.so}), so it
16672 is not available everywhere.
16674 At the user level, the extension is visible with a new storage
16675 class keyword: @code{__thread}. For example:
16679 extern __thread struct state s;
16680 static __thread char *p;
16683 The @code{__thread} specifier may be used alone, with the @code{extern}
16684 or @code{static} specifiers, but with no other storage class specifier.
16685 When used with @code{extern} or @code{static}, @code{__thread} must appear
16686 immediately after the other storage class specifier.
16688 The @code{__thread} specifier may be applied to any global, file-scoped
16689 static, function-scoped static, or static data member of a class. It may
16690 not be applied to block-scoped automatic or non-static data member.
16692 When the address-of operator is applied to a thread-local variable, it is
16693 evaluated at run time and returns the address of the current thread's
16694 instance of that variable. An address so obtained may be used by any
16695 thread. When a thread terminates, any pointers to thread-local variables
16696 in that thread become invalid.
16698 No static initialization may refer to the address of a thread-local variable.
16700 In C++, if an initializer is present for a thread-local variable, it must
16701 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
16704 See @uref{http://www.akkadia.org/drepper/tls.pdf,
16705 ELF Handling For Thread-Local Storage} for a detailed explanation of
16706 the four thread-local storage addressing models, and how the runtime
16707 is expected to function.
16710 * C99 Thread-Local Edits::
16711 * C++98 Thread-Local Edits::
16714 @node C99 Thread-Local Edits
16715 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
16717 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
16718 that document the exact semantics of the language extension.
16722 @cite{5.1.2 Execution environments}
16724 Add new text after paragraph 1
16727 Within either execution environment, a @dfn{thread} is a flow of
16728 control within a program. It is implementation defined whether
16729 or not there may be more than one thread associated with a program.
16730 It is implementation defined how threads beyond the first are
16731 created, the name and type of the function called at thread
16732 startup, and how threads may be terminated. However, objects
16733 with thread storage duration shall be initialized before thread
16738 @cite{6.2.4 Storage durations of objects}
16740 Add new text before paragraph 3
16743 An object whose identifier is declared with the storage-class
16744 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
16745 Its lifetime is the entire execution of the thread, and its
16746 stored value is initialized only once, prior to thread startup.
16750 @cite{6.4.1 Keywords}
16752 Add @code{__thread}.
16755 @cite{6.7.1 Storage-class specifiers}
16757 Add @code{__thread} to the list of storage class specifiers in
16760 Change paragraph 2 to
16763 With the exception of @code{__thread}, at most one storage-class
16764 specifier may be given [@dots{}]. The @code{__thread} specifier may
16765 be used alone, or immediately following @code{extern} or
16769 Add new text after paragraph 6
16772 The declaration of an identifier for a variable that has
16773 block scope that specifies @code{__thread} shall also
16774 specify either @code{extern} or @code{static}.
16776 The @code{__thread} specifier shall be used only with
16781 @node C++98 Thread-Local Edits
16782 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
16784 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
16785 that document the exact semantics of the language extension.
16789 @b{[intro.execution]}
16791 New text after paragraph 4
16794 A @dfn{thread} is a flow of control within the abstract machine.
16795 It is implementation defined whether or not there may be more than
16799 New text after paragraph 7
16802 It is unspecified whether additional action must be taken to
16803 ensure when and whether side effects are visible to other threads.
16809 Add @code{__thread}.
16812 @b{[basic.start.main]}
16814 Add after paragraph 5
16817 The thread that begins execution at the @code{main} function is called
16818 the @dfn{main thread}. It is implementation defined how functions
16819 beginning threads other than the main thread are designated or typed.
16820 A function so designated, as well as the @code{main} function, is called
16821 a @dfn{thread startup function}. It is implementation defined what
16822 happens if a thread startup function returns. It is implementation
16823 defined what happens to other threads when any thread calls @code{exit}.
16827 @b{[basic.start.init]}
16829 Add after paragraph 4
16832 The storage for an object of thread storage duration shall be
16833 statically initialized before the first statement of the thread startup
16834 function. An object of thread storage duration shall not require
16835 dynamic initialization.
16839 @b{[basic.start.term]}
16841 Add after paragraph 3
16844 The type of an object with thread storage duration shall not have a
16845 non-trivial destructor, nor shall it be an array type whose elements
16846 (directly or indirectly) have non-trivial destructors.
16852 Add ``thread storage duration'' to the list in paragraph 1.
16857 Thread, static, and automatic storage durations are associated with
16858 objects introduced by declarations [@dots{}].
16861 Add @code{__thread} to the list of specifiers in paragraph 3.
16864 @b{[basic.stc.thread]}
16866 New section before @b{[basic.stc.static]}
16869 The keyword @code{__thread} applied to a non-local object gives the
16870 object thread storage duration.
16872 A local variable or class data member declared both @code{static}
16873 and @code{__thread} gives the variable or member thread storage
16878 @b{[basic.stc.static]}
16883 All objects that have neither thread storage duration, dynamic
16884 storage duration nor are local [@dots{}].
16890 Add @code{__thread} to the list in paragraph 1.
16895 With the exception of @code{__thread}, at most one
16896 @var{storage-class-specifier} shall appear in a given
16897 @var{decl-specifier-seq}. The @code{__thread} specifier may
16898 be used alone, or immediately following the @code{extern} or
16899 @code{static} specifiers. [@dots{}]
16902 Add after paragraph 5
16905 The @code{__thread} specifier can be applied only to the names of objects
16906 and to anonymous unions.
16912 Add after paragraph 6
16915 Non-@code{static} members shall not be @code{__thread}.
16919 @node Binary constants
16920 @section Binary constants using the @samp{0b} prefix
16921 @cindex Binary constants using the @samp{0b} prefix
16923 Integer constants can be written as binary constants, consisting of a
16924 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
16925 @samp{0B}. This is particularly useful in environments that operate a
16926 lot on the bit level (like microcontrollers).
16928 The following statements are identical:
16937 The type of these constants follows the same rules as for octal or
16938 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
16941 @node C++ Extensions
16942 @chapter Extensions to the C++ Language
16943 @cindex extensions, C++ language
16944 @cindex C++ language extensions
16946 The GNU compiler provides these extensions to the C++ language (and you
16947 can also use most of the C language extensions in your C++ programs). If you
16948 want to write code that checks whether these features are available, you can
16949 test for the GNU compiler the same way as for C programs: check for a
16950 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
16951 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
16952 Predefined Macros,cpp,The GNU C Preprocessor}).
16955 * C++ Volatiles:: What constitutes an access to a volatile object.
16956 * Restricted Pointers:: C99 restricted pointers and references.
16957 * Vague Linkage:: Where G++ puts inlines, vtables and such.
16958 * C++ Interface:: You can use a single C++ header file for both
16959 declarations and definitions.
16960 * Template Instantiation:: Methods for ensuring that exactly one copy of
16961 each needed template instantiation is emitted.
16962 * Bound member functions:: You can extract a function pointer to the
16963 method denoted by a @samp{->*} or @samp{.*} expression.
16964 * C++ Attributes:: Variable, function, and type attributes for C++ only.
16965 * Function Multiversioning:: Declaring multiple function versions.
16966 * Namespace Association:: Strong using-directives for namespace association.
16967 * Type Traits:: Compiler support for type traits.
16968 * C++ Concepts:: Improved support for generic programming.
16969 * Java Exceptions:: Tweaking exception handling to work with Java.
16970 * Deprecated Features:: Things will disappear from G++.
16971 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
16974 @node C++ Volatiles
16975 @section When is a Volatile C++ Object Accessed?
16976 @cindex accessing volatiles
16977 @cindex volatile read
16978 @cindex volatile write
16979 @cindex volatile access
16981 The C++ standard differs from the C standard in its treatment of
16982 volatile objects. It fails to specify what constitutes a volatile
16983 access, except to say that C++ should behave in a similar manner to C
16984 with respect to volatiles, where possible. However, the different
16985 lvalueness of expressions between C and C++ complicate the behavior.
16986 G++ behaves the same as GCC for volatile access, @xref{C
16987 Extensions,,Volatiles}, for a description of GCC's behavior.
16989 The C and C++ language specifications differ when an object is
16990 accessed in a void context:
16993 volatile int *src = @var{somevalue};
16997 The C++ standard specifies that such expressions do not undergo lvalue
16998 to rvalue conversion, and that the type of the dereferenced object may
16999 be incomplete. The C++ standard does not specify explicitly that it
17000 is lvalue to rvalue conversion that is responsible for causing an
17001 access. There is reason to believe that it is, because otherwise
17002 certain simple expressions become undefined. However, because it
17003 would surprise most programmers, G++ treats dereferencing a pointer to
17004 volatile object of complete type as GCC would do for an equivalent
17005 type in C@. When the object has incomplete type, G++ issues a
17006 warning; if you wish to force an error, you must force a conversion to
17007 rvalue with, for instance, a static cast.
17009 When using a reference to volatile, G++ does not treat equivalent
17010 expressions as accesses to volatiles, but instead issues a warning that
17011 no volatile is accessed. The rationale for this is that otherwise it
17012 becomes difficult to determine where volatile access occur, and not
17013 possible to ignore the return value from functions returning volatile
17014 references. Again, if you wish to force a read, cast the reference to
17017 G++ implements the same behavior as GCC does when assigning to a
17018 volatile object---there is no reread of the assigned-to object, the
17019 assigned rvalue is reused. Note that in C++ assignment expressions
17020 are lvalues, and if used as an lvalue, the volatile object is
17021 referred to. For instance, @var{vref} refers to @var{vobj}, as
17022 expected, in the following example:
17026 volatile int &vref = vobj = @var{something};
17029 @node Restricted Pointers
17030 @section Restricting Pointer Aliasing
17031 @cindex restricted pointers
17032 @cindex restricted references
17033 @cindex restricted this pointer
17035 As with the C front end, G++ understands the C99 feature of restricted pointers,
17036 specified with the @code{__restrict__}, or @code{__restrict} type
17037 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
17038 language flag, @code{restrict} is not a keyword in C++.
17040 In addition to allowing restricted pointers, you can specify restricted
17041 references, which indicate that the reference is not aliased in the local
17045 void fn (int *__restrict__ rptr, int &__restrict__ rref)
17052 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
17053 @var{rref} refers to a (different) unaliased integer.
17055 You may also specify whether a member function's @var{this} pointer is
17056 unaliased by using @code{__restrict__} as a member function qualifier.
17059 void T::fn () __restrict__
17066 Within the body of @code{T::fn}, @var{this} has the effective
17067 definition @code{T *__restrict__ const this}. Notice that the
17068 interpretation of a @code{__restrict__} member function qualifier is
17069 different to that of @code{const} or @code{volatile} qualifier, in that it
17070 is applied to the pointer rather than the object. This is consistent with
17071 other compilers that implement restricted pointers.
17073 As with all outermost parameter qualifiers, @code{__restrict__} is
17074 ignored in function definition matching. This means you only need to
17075 specify @code{__restrict__} in a function definition, rather than
17076 in a function prototype as well.
17078 @node Vague Linkage
17079 @section Vague Linkage
17080 @cindex vague linkage
17082 There are several constructs in C++ that require space in the object
17083 file but are not clearly tied to a single translation unit. We say that
17084 these constructs have ``vague linkage''. Typically such constructs are
17085 emitted wherever they are needed, though sometimes we can be more
17089 @item Inline Functions
17090 Inline functions are typically defined in a header file which can be
17091 included in many different compilations. Hopefully they can usually be
17092 inlined, but sometimes an out-of-line copy is necessary, if the address
17093 of the function is taken or if inlining fails. In general, we emit an
17094 out-of-line copy in all translation units where one is needed. As an
17095 exception, we only emit inline virtual functions with the vtable, since
17096 it always requires a copy.
17098 Local static variables and string constants used in an inline function
17099 are also considered to have vague linkage, since they must be shared
17100 between all inlined and out-of-line instances of the function.
17104 C++ virtual functions are implemented in most compilers using a lookup
17105 table, known as a vtable. The vtable contains pointers to the virtual
17106 functions provided by a class, and each object of the class contains a
17107 pointer to its vtable (or vtables, in some multiple-inheritance
17108 situations). If the class declares any non-inline, non-pure virtual
17109 functions, the first one is chosen as the ``key method'' for the class,
17110 and the vtable is only emitted in the translation unit where the key
17113 @emph{Note:} If the chosen key method is later defined as inline, the
17114 vtable is still emitted in every translation unit that defines it.
17115 Make sure that any inline virtuals are declared inline in the class
17116 body, even if they are not defined there.
17118 @item @code{type_info} objects
17119 @cindex @code{type_info}
17121 C++ requires information about types to be written out in order to
17122 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
17123 For polymorphic classes (classes with virtual functions), the @samp{type_info}
17124 object is written out along with the vtable so that @samp{dynamic_cast}
17125 can determine the dynamic type of a class object at run time. For all
17126 other types, we write out the @samp{type_info} object when it is used: when
17127 applying @samp{typeid} to an expression, throwing an object, or
17128 referring to a type in a catch clause or exception specification.
17130 @item Template Instantiations
17131 Most everything in this section also applies to template instantiations,
17132 but there are other options as well.
17133 @xref{Template Instantiation,,Where's the Template?}.
17137 When used with GNU ld version 2.8 or later on an ELF system such as
17138 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
17139 these constructs will be discarded at link time. This is known as
17142 On targets that don't support COMDAT, but do support weak symbols, GCC
17143 uses them. This way one copy overrides all the others, but
17144 the unused copies still take up space in the executable.
17146 For targets that do not support either COMDAT or weak symbols,
17147 most entities with vague linkage are emitted as local symbols to
17148 avoid duplicate definition errors from the linker. This does not happen
17149 for local statics in inlines, however, as having multiple copies
17150 almost certainly breaks things.
17152 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
17153 another way to control placement of these constructs.
17155 @node C++ Interface
17156 @section #pragma interface and implementation
17158 @cindex interface and implementation headers, C++
17159 @cindex C++ interface and implementation headers
17160 @cindex pragmas, interface and implementation
17162 @code{#pragma interface} and @code{#pragma implementation} provide the
17163 user with a way of explicitly directing the compiler to emit entities
17164 with vague linkage (and debugging information) in a particular
17167 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
17168 most cases, because of COMDAT support and the ``key method'' heuristic
17169 mentioned in @ref{Vague Linkage}. Using them can actually cause your
17170 program to grow due to unnecessary out-of-line copies of inline
17171 functions. Currently (3.4) the only benefit of these
17172 @code{#pragma}s is reduced duplication of debugging information, and
17173 that should be addressed soon on DWARF 2 targets with the use of
17177 @item #pragma interface
17178 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
17179 @kindex #pragma interface
17180 Use this directive in @emph{header files} that define object classes, to save
17181 space in most of the object files that use those classes. Normally,
17182 local copies of certain information (backup copies of inline member
17183 functions, debugging information, and the internal tables that implement
17184 virtual functions) must be kept in each object file that includes class
17185 definitions. You can use this pragma to avoid such duplication. When a
17186 header file containing @samp{#pragma interface} is included in a
17187 compilation, this auxiliary information is not generated (unless
17188 the main input source file itself uses @samp{#pragma implementation}).
17189 Instead, the object files contain references to be resolved at link
17192 The second form of this directive is useful for the case where you have
17193 multiple headers with the same name in different directories. If you
17194 use this form, you must specify the same string to @samp{#pragma
17197 @item #pragma implementation
17198 @itemx #pragma implementation "@var{objects}.h"
17199 @kindex #pragma implementation
17200 Use this pragma in a @emph{main input file}, when you want full output from
17201 included header files to be generated (and made globally visible). The
17202 included header file, in turn, should use @samp{#pragma interface}.
17203 Backup copies of inline member functions, debugging information, and the
17204 internal tables used to implement virtual functions are all generated in
17205 implementation files.
17207 @cindex implied @code{#pragma implementation}
17208 @cindex @code{#pragma implementation}, implied
17209 @cindex naming convention, implementation headers
17210 If you use @samp{#pragma implementation} with no argument, it applies to
17211 an include file with the same basename@footnote{A file's @dfn{basename}
17212 is the name stripped of all leading path information and of trailing
17213 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
17214 file. For example, in @file{allclass.cc}, giving just
17215 @samp{#pragma implementation}
17216 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
17218 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
17219 an implementation file whenever you would include it from
17220 @file{allclass.cc} even if you never specified @samp{#pragma
17221 implementation}. This was deemed to be more trouble than it was worth,
17222 however, and disabled.
17224 Use the string argument if you want a single implementation file to
17225 include code from multiple header files. (You must also use
17226 @samp{#include} to include the header file; @samp{#pragma
17227 implementation} only specifies how to use the file---it doesn't actually
17230 There is no way to split up the contents of a single header file into
17231 multiple implementation files.
17234 @cindex inlining and C++ pragmas
17235 @cindex C++ pragmas, effect on inlining
17236 @cindex pragmas in C++, effect on inlining
17237 @samp{#pragma implementation} and @samp{#pragma interface} also have an
17238 effect on function inlining.
17240 If you define a class in a header file marked with @samp{#pragma
17241 interface}, the effect on an inline function defined in that class is
17242 similar to an explicit @code{extern} declaration---the compiler emits
17243 no code at all to define an independent version of the function. Its
17244 definition is used only for inlining with its callers.
17246 @opindex fno-implement-inlines
17247 Conversely, when you include the same header file in a main source file
17248 that declares it as @samp{#pragma implementation}, the compiler emits
17249 code for the function itself; this defines a version of the function
17250 that can be found via pointers (or by callers compiled without
17251 inlining). If all calls to the function can be inlined, you can avoid
17252 emitting the function by compiling with @option{-fno-implement-inlines}.
17253 If any calls are not inlined, you will get linker errors.
17255 @node Template Instantiation
17256 @section Where's the Template?
17257 @cindex template instantiation
17259 C++ templates are the first language feature to require more
17260 intelligence from the environment than one usually finds on a UNIX
17261 system. Somehow the compiler and linker have to make sure that each
17262 template instance occurs exactly once in the executable if it is needed,
17263 and not at all otherwise. There are two basic approaches to this
17264 problem, which are referred to as the Borland model and the Cfront model.
17267 @item Borland model
17268 Borland C++ solved the template instantiation problem by adding the code
17269 equivalent of common blocks to their linker; the compiler emits template
17270 instances in each translation unit that uses them, and the linker
17271 collapses them together. The advantage of this model is that the linker
17272 only has to consider the object files themselves; there is no external
17273 complexity to worry about. This disadvantage is that compilation time
17274 is increased because the template code is being compiled repeatedly.
17275 Code written for this model tends to include definitions of all
17276 templates in the header file, since they must be seen to be
17280 The AT&T C++ translator, Cfront, solved the template instantiation
17281 problem by creating the notion of a template repository, an
17282 automatically maintained place where template instances are stored. A
17283 more modern version of the repository works as follows: As individual
17284 object files are built, the compiler places any template definitions and
17285 instantiations encountered in the repository. At link time, the link
17286 wrapper adds in the objects in the repository and compiles any needed
17287 instances that were not previously emitted. The advantages of this
17288 model are more optimal compilation speed and the ability to use the
17289 system linker; to implement the Borland model a compiler vendor also
17290 needs to replace the linker. The disadvantages are vastly increased
17291 complexity, and thus potential for error; for some code this can be
17292 just as transparent, but in practice it can been very difficult to build
17293 multiple programs in one directory and one program in multiple
17294 directories. Code written for this model tends to separate definitions
17295 of non-inline member templates into a separate file, which should be
17296 compiled separately.
17299 When used with GNU ld version 2.8 or later on an ELF system such as
17300 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
17301 Borland model. On other systems, G++ implements neither automatic
17304 You have the following options for dealing with template instantiations:
17309 Compile your template-using code with @option{-frepo}. The compiler
17310 generates files with the extension @samp{.rpo} listing all of the
17311 template instantiations used in the corresponding object files that
17312 could be instantiated there; the link wrapper, @samp{collect2},
17313 then updates the @samp{.rpo} files to tell the compiler where to place
17314 those instantiations and rebuild any affected object files. The
17315 link-time overhead is negligible after the first pass, as the compiler
17316 continues to place the instantiations in the same files.
17318 This is your best option for application code written for the Borland
17319 model, as it just works. Code written for the Cfront model
17320 needs to be modified so that the template definitions are available at
17321 one or more points of instantiation; usually this is as simple as adding
17322 @code{#include <tmethods.cc>} to the end of each template header.
17324 For library code, if you want the library to provide all of the template
17325 instantiations it needs, just try to link all of its object files
17326 together; the link will fail, but cause the instantiations to be
17327 generated as a side effect. Be warned, however, that this may cause
17328 conflicts if multiple libraries try to provide the same instantiations.
17329 For greater control, use explicit instantiation as described in the next
17333 @opindex fno-implicit-templates
17334 Compile your code with @option{-fno-implicit-templates} to disable the
17335 implicit generation of template instances, and explicitly instantiate
17336 all the ones you use. This approach requires more knowledge of exactly
17337 which instances you need than do the others, but it's less
17338 mysterious and allows greater control. You can scatter the explicit
17339 instantiations throughout your program, perhaps putting them in the
17340 translation units where the instances are used or the translation units
17341 that define the templates themselves; you can put all of the explicit
17342 instantiations you need into one big file; or you can create small files
17349 template class Foo<int>;
17350 template ostream& operator <<
17351 (ostream&, const Foo<int>&);
17355 for each of the instances you need, and create a template instantiation
17356 library from those.
17358 If you are using Cfront-model code, you can probably get away with not
17359 using @option{-fno-implicit-templates} when compiling files that don't
17360 @samp{#include} the member template definitions.
17362 If you use one big file to do the instantiations, you may want to
17363 compile it without @option{-fno-implicit-templates} so you get all of the
17364 instances required by your explicit instantiations (but not by any
17365 other files) without having to specify them as well.
17367 The ISO C++ 2011 standard allows forward declaration of explicit
17368 instantiations (with @code{extern}). G++ supports explicit instantiation
17369 declarations in C++98 mode and has extended the template instantiation
17370 syntax to support instantiation of the compiler support data for a
17371 template class (i.e.@: the vtable) without instantiating any of its
17372 members (with @code{inline}), and instantiation of only the static data
17373 members of a template class, without the support data or member
17374 functions (with (@code{static}):
17377 extern template int max (int, int);
17378 inline template class Foo<int>;
17379 static template class Foo<int>;
17383 Do nothing. Pretend G++ does implement automatic instantiation
17384 management. Code written for the Borland model works fine, but
17385 each translation unit contains instances of each of the templates it
17386 uses. In a large program, this can lead to an unacceptable amount of code
17390 @node Bound member functions
17391 @section Extracting the function pointer from a bound pointer to member function
17393 @cindex pointer to member function
17394 @cindex bound pointer to member function
17396 In C++, pointer to member functions (PMFs) are implemented using a wide
17397 pointer of sorts to handle all the possible call mechanisms; the PMF
17398 needs to store information about how to adjust the @samp{this} pointer,
17399 and if the function pointed to is virtual, where to find the vtable, and
17400 where in the vtable to look for the member function. If you are using
17401 PMFs in an inner loop, you should really reconsider that decision. If
17402 that is not an option, you can extract the pointer to the function that
17403 would be called for a given object/PMF pair and call it directly inside
17404 the inner loop, to save a bit of time.
17406 Note that you still pay the penalty for the call through a
17407 function pointer; on most modern architectures, such a call defeats the
17408 branch prediction features of the CPU@. This is also true of normal
17409 virtual function calls.
17411 The syntax for this extension is
17415 extern int (A::*fp)();
17416 typedef int (*fptr)(A *);
17418 fptr p = (fptr)(a.*fp);
17421 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
17422 no object is needed to obtain the address of the function. They can be
17423 converted to function pointers directly:
17426 fptr p1 = (fptr)(&A::foo);
17429 @opindex Wno-pmf-conversions
17430 You must specify @option{-Wno-pmf-conversions} to use this extension.
17432 @node C++ Attributes
17433 @section C++-Specific Variable, Function, and Type Attributes
17435 Some attributes only make sense for C++ programs.
17438 @item abi_tag ("@var{tag}", ...)
17439 @cindex @code{abi_tag} attribute
17440 The @code{abi_tag} attribute can be applied to a function or class
17441 declaration. It modifies the mangled name of the function or class to
17442 incorporate the tag name, in order to distinguish the function or
17443 class from an earlier version with a different ABI; perhaps the class
17444 has changed size, or the function has a different return type that is
17445 not encoded in the mangled name.
17447 The argument can be a list of strings of arbitrary length. The
17448 strings are sorted on output, so the order of the list is
17451 A redeclaration of a function or class must not add new ABI tags,
17452 since doing so would change the mangled name.
17454 The @option{-Wabi-tag} flag enables a warning about a class which does
17455 not have all the ABI tags used by its subobjects and virtual functions; for users with code
17456 that needs to coexist with an earlier ABI, using this option can help
17457 to find all affected types that need to be tagged.
17459 @item init_priority (@var{priority})
17460 @cindex @code{init_priority} attribute
17463 In Standard C++, objects defined at namespace scope are guaranteed to be
17464 initialized in an order in strict accordance with that of their definitions
17465 @emph{in a given translation unit}. No guarantee is made for initializations
17466 across translation units. However, GNU C++ allows users to control the
17467 order of initialization of objects defined at namespace scope with the
17468 @code{init_priority} attribute by specifying a relative @var{priority},
17469 a constant integral expression currently bounded between 101 and 65535
17470 inclusive. Lower numbers indicate a higher priority.
17472 In the following example, @code{A} would normally be created before
17473 @code{B}, but the @code{init_priority} attribute reverses that order:
17476 Some_Class A __attribute__ ((init_priority (2000)));
17477 Some_Class B __attribute__ ((init_priority (543)));
17481 Note that the particular values of @var{priority} do not matter; only their
17484 @item java_interface
17485 @cindex @code{java_interface} attribute
17487 This type attribute informs C++ that the class is a Java interface. It may
17488 only be applied to classes declared within an @code{extern "Java"} block.
17489 Calls to methods declared in this interface are dispatched using GCJ's
17490 interface table mechanism, instead of regular virtual table dispatch.
17493 @cindex @code{warn_unused} attribute
17495 For C++ types with non-trivial constructors and/or destructors it is
17496 impossible for the compiler to determine whether a variable of this
17497 type is truly unused if it is not referenced. This type attribute
17498 informs the compiler that variables of this type should be warned
17499 about if they appear to be unused, just like variables of fundamental
17502 This attribute is appropriate for types which just represent a value,
17503 such as @code{std::string}; it is not appropriate for types which
17504 control a resource, such as @code{std::mutex}.
17506 This attribute is also accepted in C, but it is unnecessary because C
17507 does not have constructors or destructors.
17511 See also @ref{Namespace Association}.
17513 @node Function Multiversioning
17514 @section Function Multiversioning
17515 @cindex function versions
17517 With the GNU C++ front end, for target i386, you may specify multiple
17518 versions of a function, where each function is specialized for a
17519 specific target feature. At runtime, the appropriate version of the
17520 function is automatically executed depending on the characteristics of
17521 the execution platform. Here is an example.
17524 __attribute__ ((target ("default")))
17527 // The default version of foo.
17531 __attribute__ ((target ("sse4.2")))
17534 // foo version for SSE4.2
17538 __attribute__ ((target ("arch=atom")))
17541 // foo version for the Intel ATOM processor
17545 __attribute__ ((target ("arch=amdfam10")))
17548 // foo version for the AMD Family 0x10 processors.
17555 assert ((*p) () == foo ());
17560 In the above example, four versions of function foo are created. The
17561 first version of foo with the target attribute "default" is the default
17562 version. This version gets executed when no other target specific
17563 version qualifies for execution on a particular platform. A new version
17564 of foo is created by using the same function signature but with a
17565 different target string. Function foo is called or a pointer to it is
17566 taken just like a regular function. GCC takes care of doing the
17567 dispatching to call the right version at runtime. Refer to the
17568 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
17569 Function Multiversioning} for more details.
17571 @node Namespace Association
17572 @section Namespace Association
17574 @strong{Caution:} The semantics of this extension are equivalent
17575 to C++ 2011 inline namespaces. Users should use inline namespaces
17576 instead as this extension will be removed in future versions of G++.
17578 A using-directive with @code{__attribute ((strong))} is stronger
17579 than a normal using-directive in two ways:
17583 Templates from the used namespace can be specialized and explicitly
17584 instantiated as though they were members of the using namespace.
17587 The using namespace is considered an associated namespace of all
17588 templates in the used namespace for purposes of argument-dependent
17592 The used namespace must be nested within the using namespace so that
17593 normal unqualified lookup works properly.
17595 This is useful for composing a namespace transparently from
17596 implementation namespaces. For example:
17601 template <class T> struct A @{ @};
17603 using namespace debug __attribute ((__strong__));
17604 template <> struct A<int> @{ @}; // @r{OK to specialize}
17606 template <class T> void f (A<T>);
17611 f (std::A<float>()); // @r{lookup finds} std::f
17617 @section Type Traits
17619 The C++ front end implements syntactic extensions that allow
17620 compile-time determination of
17621 various characteristics of a type (or of a
17625 @item __has_nothrow_assign (type)
17626 If @code{type} is const qualified or is a reference type then the trait is
17627 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
17628 is true, else if @code{type} is a cv class or union type with copy assignment
17629 operators that are known not to throw an exception then the trait is true,
17630 else it is false. Requires: @code{type} shall be a complete type,
17631 (possibly cv-qualified) @code{void}, or an array of unknown bound.
17633 @item __has_nothrow_copy (type)
17634 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
17635 @code{type} is a cv class or union type with copy constructors that
17636 are known not to throw an exception then the trait is true, else it is false.
17637 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
17638 @code{void}, or an array of unknown bound.
17640 @item __has_nothrow_constructor (type)
17641 If @code{__has_trivial_constructor (type)} is true then the trait is
17642 true, else if @code{type} is a cv class or union type (or array
17643 thereof) with a default constructor that is known not to throw an
17644 exception then the trait is true, else it is false. Requires:
17645 @code{type} shall be a complete type, (possibly cv-qualified)
17646 @code{void}, or an array of unknown bound.
17648 @item __has_trivial_assign (type)
17649 If @code{type} is const qualified or is a reference type then the trait is
17650 false. Otherwise if @code{__is_pod (type)} is true then the trait is
17651 true, else if @code{type} is a cv class or union type with a trivial
17652 copy assignment ([class.copy]) then the trait is true, else it is
17653 false. Requires: @code{type} shall be a complete type, (possibly
17654 cv-qualified) @code{void}, or an array of unknown bound.
17656 @item __has_trivial_copy (type)
17657 If @code{__is_pod (type)} is true or @code{type} is a reference type
17658 then the trait is true, else if @code{type} is a cv class or union type
17659 with a trivial copy constructor ([class.copy]) then the trait
17660 is true, else it is false. Requires: @code{type} shall be a complete
17661 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17663 @item __has_trivial_constructor (type)
17664 If @code{__is_pod (type)} is true then the trait is true, else if
17665 @code{type} is a cv class or union type (or array thereof) with a
17666 trivial default constructor ([class.ctor]) then the trait is true,
17667 else it is false. Requires: @code{type} shall be a complete
17668 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17670 @item __has_trivial_destructor (type)
17671 If @code{__is_pod (type)} is true or @code{type} is a reference type then
17672 the trait is true, else if @code{type} is a cv class or union type (or
17673 array thereof) with a trivial destructor ([class.dtor]) then the trait
17674 is true, else it is false. Requires: @code{type} shall be a complete
17675 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17677 @item __has_virtual_destructor (type)
17678 If @code{type} is a class type with a virtual destructor
17679 ([class.dtor]) then the trait is true, else it is false. Requires:
17680 @code{type} shall be a complete type, (possibly cv-qualified)
17681 @code{void}, or an array of unknown bound.
17683 @item __is_abstract (type)
17684 If @code{type} is an abstract class ([class.abstract]) then the trait
17685 is true, else it is false. Requires: @code{type} shall be a complete
17686 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17688 @item __is_base_of (base_type, derived_type)
17689 If @code{base_type} is a base class of @code{derived_type}
17690 ([class.derived]) then the trait is true, otherwise it is false.
17691 Top-level cv qualifications of @code{base_type} and
17692 @code{derived_type} are ignored. For the purposes of this trait, a
17693 class type is considered is own base. Requires: if @code{__is_class
17694 (base_type)} and @code{__is_class (derived_type)} are true and
17695 @code{base_type} and @code{derived_type} are not the same type
17696 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
17697 type. Diagnostic is produced if this requirement is not met.
17699 @item __is_class (type)
17700 If @code{type} is a cv class type, and not a union type
17701 ([basic.compound]) the trait is true, else it is false.
17703 @item __is_empty (type)
17704 If @code{__is_class (type)} is false then the trait is false.
17705 Otherwise @code{type} is considered empty if and only if: @code{type}
17706 has no non-static data members, or all non-static data members, if
17707 any, are bit-fields of length 0, and @code{type} has no virtual
17708 members, and @code{type} has no virtual base classes, and @code{type}
17709 has no base classes @code{base_type} for which
17710 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
17711 be a complete type, (possibly cv-qualified) @code{void}, or an array
17714 @item __is_enum (type)
17715 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
17716 true, else it is false.
17718 @item __is_literal_type (type)
17719 If @code{type} is a literal type ([basic.types]) the trait is
17720 true, else it is false. Requires: @code{type} shall be a complete type,
17721 (possibly cv-qualified) @code{void}, or an array of unknown bound.
17723 @item __is_pod (type)
17724 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
17725 else it is false. Requires: @code{type} shall be a complete type,
17726 (possibly cv-qualified) @code{void}, or an array of unknown bound.
17728 @item __is_polymorphic (type)
17729 If @code{type} is a polymorphic class ([class.virtual]) then the trait
17730 is true, else it is false. Requires: @code{type} shall be a complete
17731 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17733 @item __is_standard_layout (type)
17734 If @code{type} is a standard-layout type ([basic.types]) the trait is
17735 true, else it is false. Requires: @code{type} shall be a complete
17736 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17738 @item __is_trivial (type)
17739 If @code{type} is a trivial type ([basic.types]) the trait is
17740 true, else it is false. Requires: @code{type} shall be a complete
17741 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
17743 @item __is_union (type)
17744 If @code{type} is a cv union type ([basic.compound]) the trait is
17745 true, else it is false.
17747 @item __underlying_type (type)
17748 The underlying type of @code{type}. Requires: @code{type} shall be
17749 an enumeration type ([dcl.enum]).
17755 @section C++ Concepts
17757 C++ concepts provide much-improved support for generic programming. In
17758 particular, they allow the specification of constraints on template arguments.
17759 The constraints are used to extend the usual overloading and partial
17760 specialization capabilities of the language, allowing generic data structures
17761 and algorithms to be ``refined'' based on their properties rather than their
17764 The following keywords are reserved for concepts.
17768 States an expression as an assumption, and if possible, verifies tht the
17769 assumption is valid. For example, @code{assume(n > 0)}.
17772 Introduces an axiom definition. Axioms introduce requirements on values.
17775 Introduces a universally quantified object in an axiom. For example,
17776 @code{forall (int n) n + 0 == n}).
17779 Introduces a concept definition. Concepts are sets of syntactic and semantic
17780 requirements on types and their values.
17783 Introduces constraints on template arguments or requirements for a member
17784 function of a class template.
17788 The front end also exposes a number of internal mechanism that can be used
17789 to simplify the writing of type traits. Note that some of these traits are
17790 likely to be removed in the future.
17793 @item __is_same (type1, type2)
17794 A binary type trait: true whenever the type arguments are the same.
17796 @item __is_valid_expr (expr)
17797 A syntactic requirement. In a constraint, this trait can be used to determine
17798 if a particular expression is valid for the types of its operands, and if so,
17799 whether the result type is the same as or convertible to some other type.
17801 @item __is_valid_type (type)
17802 A syntactic requirement. In a constraint, this trait can be used to determine
17803 if a type name is well-formed.
17806 A new declaration specifier. This can be used in constraints to introduce
17807 ``declared values'' to be used as a replacement for @code{declval}.
17812 @node Java Exceptions
17813 @section Java Exceptions
17815 The Java language uses a slightly different exception handling model
17816 from C++. Normally, GNU C++ automatically detects when you are
17817 writing C++ code that uses Java exceptions, and handle them
17818 appropriately. However, if C++ code only needs to execute destructors
17819 when Java exceptions are thrown through it, GCC guesses incorrectly.
17820 Sample problematic code is:
17823 struct S @{ ~S(); @};
17824 extern void bar(); // @r{is written in Java, and may throw exceptions}
17833 The usual effect of an incorrect guess is a link failure, complaining of
17834 a missing routine called @samp{__gxx_personality_v0}.
17836 You can inform the compiler that Java exceptions are to be used in a
17837 translation unit, irrespective of what it might think, by writing
17838 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
17839 @samp{#pragma} must appear before any functions that throw or catch
17840 exceptions, or run destructors when exceptions are thrown through them.
17842 You cannot mix Java and C++ exceptions in the same translation unit. It
17843 is believed to be safe to throw a C++ exception from one file through
17844 another file compiled for the Java exception model, or vice versa, but
17845 there may be bugs in this area.
17847 @node Deprecated Features
17848 @section Deprecated Features
17850 In the past, the GNU C++ compiler was extended to experiment with new
17851 features, at a time when the C++ language was still evolving. Now that
17852 the C++ standard is complete, some of those features are superseded by
17853 superior alternatives. Using the old features might cause a warning in
17854 some cases that the feature will be dropped in the future. In other
17855 cases, the feature might be gone already.
17857 While the list below is not exhaustive, it documents some of the options
17858 that are now deprecated:
17861 @item -fexternal-templates
17862 @itemx -falt-external-templates
17863 These are two of the many ways for G++ to implement template
17864 instantiation. @xref{Template Instantiation}. The C++ standard clearly
17865 defines how template definitions have to be organized across
17866 implementation units. G++ has an implicit instantiation mechanism that
17867 should work just fine for standard-conforming code.
17869 @item -fstrict-prototype
17870 @itemx -fno-strict-prototype
17871 Previously it was possible to use an empty prototype parameter list to
17872 indicate an unspecified number of parameters (like C), rather than no
17873 parameters, as C++ demands. This feature has been removed, except where
17874 it is required for backwards compatibility. @xref{Backwards Compatibility}.
17877 G++ allows a virtual function returning @samp{void *} to be overridden
17878 by one returning a different pointer type. This extension to the
17879 covariant return type rules is now deprecated and will be removed from a
17882 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
17883 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
17884 and are now removed from G++. Code using these operators should be
17885 modified to use @code{std::min} and @code{std::max} instead.
17887 The named return value extension has been deprecated, and is now
17890 The use of initializer lists with new expressions has been deprecated,
17891 and is now removed from G++.
17893 Floating and complex non-type template parameters have been deprecated,
17894 and are now removed from G++.
17896 The implicit typename extension has been deprecated and is now
17899 The use of default arguments in function pointers, function typedefs
17900 and other places where they are not permitted by the standard is
17901 deprecated and will be removed from a future version of G++.
17903 G++ allows floating-point literals to appear in integral constant expressions,
17904 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
17905 This extension is deprecated and will be removed from a future version.
17907 G++ allows static data members of const floating-point type to be declared
17908 with an initializer in a class definition. The standard only allows
17909 initializers for static members of const integral types and const
17910 enumeration types so this extension has been deprecated and will be removed
17911 from a future version.
17913 @node Backwards Compatibility
17914 @section Backwards Compatibility
17915 @cindex Backwards Compatibility
17916 @cindex ARM [Annotated C++ Reference Manual]
17918 Now that there is a definitive ISO standard C++, G++ has a specification
17919 to adhere to. The C++ language evolved over time, and features that
17920 used to be acceptable in previous drafts of the standard, such as the ARM
17921 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
17922 compilation of C++ written to such drafts, G++ contains some backwards
17923 compatibilities. @emph{All such backwards compatibility features are
17924 liable to disappear in future versions of G++.} They should be considered
17925 deprecated. @xref{Deprecated Features}.
17929 If a variable is declared at for scope, it used to remain in scope until
17930 the end of the scope that contained the for statement (rather than just
17931 within the for scope). G++ retains this, but issues a warning, if such a
17932 variable is accessed outside the for scope.
17934 @item Implicit C language
17935 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
17936 scope to set the language. On such systems, all header files are
17937 implicitly scoped inside a C language scope. Also, an empty prototype
17938 @code{()} is treated as an unspecified number of arguments, rather
17939 than no arguments, as C++ demands.
17942 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
17943 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr followign